System and method for coordinated control of wind power clusters

Abstract

The application discloses a system and method for wind power cluster collaborative control, and belongs to the technical field of collaborative control. The system comprises the following steps: predicting wind field information in a future target time domain, calculating a dynamic risk index of each wind turbine generator at a current control time; dividing the wind power cluster into at least four different risk regions according to the risk index; generating a differentiated collaborative scheduling and control instruction set according to different risk regions; and issuing the control instructions in the instruction set to each wind turbine generator actuator for execution. The application can more accurately identify the disaster risk differences of different units in the typhoon path evolution process, provide a forward-looking risk judgment basis for wind power cluster disaster resistance control, and significantly improve the early warning ability and environmental adaptability of the system under extreme weather conditions.

Description

Technical Field

[0001] This invention relates to the field of collaborative control technology, and in particular to a system and method for collaborative control of wind power clusters. It is especially relevant to the disaster-resistant collaborative control of wind power clusters during typhoon season. Background Technology

[0002] As an important renewable energy source, wind power has seen rapid large-scale deployment in coastal and offshore areas in recent years, with wind farms exhibiting characteristics of clustered operation of multiple wind turbine units. However, these areas are also typhoon-prone regions. During typhoons, wind speeds are high, wind direction changes rapidly, and gusts and turbulence are significant, resulting in wind fields exhibiting strong nonlinearity, strong time-varying characteristics, and obvious spatial distribution differences. Under these extreme conditions, key components of wind turbines, such as blades, towers, and drive trains, are susceptible to severe dynamic load impacts, leading to increased vibration, accumulated structural fatigue, power output fluctuations, and even triggering protective shutdowns. In severe cases, this can cause equipment damage and operational safety risks, ultimately affecting the overall power generation revenue and grid connection stability of the wind farm.

[0003] Existing wind turbine protection and control technologies under typhoon or extreme wind conditions typically employ strategies such as pitch control, yaw control, power limiting operation, and shutdown protection at the single-unit level to reduce the stress level on the wind turbine and increase safety margins. In addition, some solutions also introduce short-term wind speed forecasting or power dispatching methods to improve power fluctuations and operational stability.

[0004] However, the aforementioned methods are all based on a uniform threshold or a single control strategy, failing to fully consider key factors such as the typhoon's path evolution, the spatial differences of each wind turbine, and load margins. Therefore, they are unable to accurately distinguish the differences in disaster risk among wind turbines in different regions. This deficiency leads to high-risk wind turbines potentially under-responding during typhoon impacts, while low-risk wind turbines excessively reduce their load, ultimately resulting in the overall safety and power stability of wind power clusters being unable to be effectively guaranteed during typhoons. Summary of the Invention

[0005] Purpose of the Invention: To address the above-mentioned problems, the purpose of this invention is to provide a system and method for collaborative control of wind power clusters. The core concept of this invention lies in: utilizing a BFO-BP hybrid algorithm to train a high-precision dynamic wind farm prediction model, and fusing multi-dimensional information such as prediction results, typhoon path information, wind turbine spatial location, and key structural loads to calculate the dynamic risk index for each wind turbine. Based on this dynamic risk index, the wind turbines in the cluster are dynamically divided into a core impact zone, a transitional load reduction zone, a peripheral compensation zone, and a recovery zone. Finally, for different zones, a differentiated multi-objective optimization model is established with the objectives of minimizing the structural safety stress of the wind turbines and maximizing power output, generating corresponding control commands. The risk zones and control strategies are then continuously corrected through execution status feedback, thereby forming a complete closed-loop control.

[0006] It should be added that a wind power cluster consists of multiple wind turbine units, and usually refers to the collection of all wind turbine units in a wind farm.

[0007] Technical solution: In a first aspect, the present invention provides a system for collaborative control of wind power clusters, comprising:

[0008] The data sensing module is configured to collect data on each individual wind turbine unit within the wind farm.

[0009] The cluster collaborative computing module is configured to call the data collected by the data sensing module to predict wind field information in the future target time domain, and construct a risk factor set based on the prediction results; calculate the dynamic risk index of each wind turbine at the current control moment by weighted summation of each risk factor; divide the wind power cluster into at least four different risk areas according to the risk index; and generate differentiated collaborative scheduling and control instruction sets according to different risk areas.

[0010] as well as

[0011] The wind turbine control module is configured to send control commands from the control command set to the actuators of each wind turbine.

[0012] In the above scheme, multi-dimensional data (typhoon path, unit status, spatial location, etc.) is collected through a data sensing module, and future temporal wind fields are predicted. A dynamic risk index is formed by weighted summation of five risk factors: predicted wind speed, sudden changes in wind direction, structural load, typhoon distance, and prediction error. This allows the risk assessment of each unit to not only be based on the current moment but also reflect the typhoon's evolution trend, enabling preventative control to be initiated before high-risk units are impacted, significantly improving the timeliness of early warning response.

[0013] Furthermore, based on the dynamic risk index, the cluster is divided into a core impact zone, a transitional load reduction zone, a peripheral compensation zone, and a recovery zone, and commands for feathering shutdown, pre-yaw load reduction, power compensation, and smooth recovery are generated respectively. After the units in the core zone perform feathering load reduction, the ultimate bending moment on the blades and the fatigue load on the tower are significantly reduced; the units in the peripheral compensation zone increase their power output within the load margin to make up for the power deficit of the load-reduced units. This achieves a synergistic effect of prioritizing the protection of high-risk units and safely supplementing power to low-risk units, thus avoiding structural overload and suppressing power fluctuations across the entire field.

[0014] Secondly, the present invention provides a method for collaborative control of wind power clusters, comprising the following steps:

[0015] Data is acquired by collecting data from each individual wind turbine within the wind farm.

[0016] Based on the risk zoning results, a differentiated set of collaborative scheduling and control instructions is generated; wherein the sub-steps of risk zoning include: predicting wind field information in the future target time domain and constructing a set of risk factors based on the prediction results; calculating the dynamic risk index of each wind turbine at the current control moment by weighted summation of each risk factor; and dividing the wind power cluster into at least four different risk zones based on the risk index.

[0017] The control commands from the control command set are sent to the actuators of each wind turbine.

[0018] Furthermore, after executing control, the actual yaw angle, pitch angle, speed, power, and load are collected to form a feedback vector, which is then weighted and fused with the predicted risk index to dynamically update the risk index and zoning results for the next moment. When the actual response deviates significantly from the prediction, the feedback weight is increased, causing the zoning results to be rolled back and corrected according to typhoon changes and the actual state of the unit. This avoids the accumulation of errors in open-loop prediction, ensures that the control strategy always conforms to the actual operating conditions, and enhances the system's robustness.

[0019] Beneficial effects:

[0020] The significant advantages of this invention compared to existing technologies are:

[0021] 1. This invention can more accurately identify the differences in disaster risk of different units during the evolution of typhoon paths, provide a forward-looking risk assessment basis for wind power cluster disaster resistance control, and significantly improve the system's early warning capability and environmental adaptability under extreme weather conditions.

[0022] 2. This invention generates differentiated control commands such as feathering shutdown, pre-yaw load reduction, power compensation, and smooth recovery for units in different risk areas. This can avoid insufficient protection for high-risk units and reduce excessive load reduction for low-risk units, thereby achieving a balance between unit safety, cluster coordination, and power stability. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the system composition structure of the present invention;

[0024] Figure 2 This is a schematic diagram of the method steps of the present invention;

[0025] Figure 3 This is a comparison chart of the predicted wind speed and the actual observed wind speed according to the present invention;

[0026] Figure 4 This is a comparison chart of the error convergence curves of different prediction algorithms of the present invention;

[0027] Figure 5 This is a comparison diagram of the unit's power response before and after the coordinated control of the present invention. Detailed Implementation

[0028] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and not intended to limit the scope of the invention. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the embodiments of the present invention, and not all structures.

[0029] In the following description, specific details such as target system architecture and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0030] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0031] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0032] Furthermore, in the description of this application and the appended claims, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0033] References to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include the target features, structures, or characteristics described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0034] First, the following technical terms of this invention will be explained:

[0035] Bacterial Foraging Optimization (BFO) is a swarm intelligence optimization algorithm that simulates the foraging behavior of E. coli. It achieves global optimization of the objective function through three operations: tendency, replication, and dispersal.

[0036] Back Propagation Neural Network (BP) is a multi-layer feedforward neural network trained using the back propagation algorithm. It has strong nonlinear fitting ability, but its training effect is sensitive to the initial weights and thresholds and is prone to getting trapped in local optima.

[0037] The core principle of the BFO-BP hybrid algorithm lies in using the global search capability of the BFO algorithm to provide optimal initial weights and thresholds for the BP neural network, and then using the BP algorithm for local fine-tuning. This overcomes the weakness of a single BP network being prone to getting trapped in local optima, thereby improving the convergence speed and accuracy of the prediction model. This hybrid algorithm has already been researched in fields such as wind power prediction and is a known algorithm combination in this field.

[0038] In this invention, the BFO-BP hybrid algorithm is used to optimize and train the parameters of the dynamic wind field prediction model. Its output (i.e., the optimized prediction model) serves as the data basis for subsequent dynamic risk index calculation and collaborative optimization control. Those skilled in the art should understand that the core improvement of this invention lies not in the BFO-BP hybrid algorithm itself, but in fusing the algorithm's prediction results with multi-dimensional information such as typhoon path, wind turbine spatial location, critical structural loads, and prediction errors to construct a dynamic risk index and achieve zoned collaborative control accordingly. In other words, the BFO-BP hybrid algorithm serves as a front-end prediction tool in this invention, providing more reliable predictive input for subsequent risk assessment and differentiated control. The substantial innovation of this invention lies in the risk zoned collaborative control system and method constructed based on this predictive input.

[0039] The following examples will provide a detailed explanation.

[0040] In one embodiment, a wind farm includes multiple wind turbine units, which together form a wind power cluster. Each wind turbine unit is equipped with a wind turbine control module and a disaster mitigation execution module, and is communicatively connected to a cluster collaborative computing module and a data sensing module.

[0041] The data sensing module includes a meteorological monitoring unit, a wind turbine status acquisition unit, a wind farm environment monitoring unit, and a wind turbine location recording unit. The meteorological monitoring unit collects data on wind speed, wind direction, air pressure, typhoon center location, typhoon movement direction, and typhoon speed. The wind turbine status acquisition unit collects data on the rotational speed, output power, yaw angle, pitch angle, and key structural loads of each wind turbine. The wind farm environment monitoring unit acquires wind farm spatial distribution parameters and environmental parameters. The wind turbine location recording unit stores the spatial coordinates of each wind turbine in the wind farm coordinate system.

[0042] Those skilled in the art should understand that all of the above-mentioned units can be implemented using conventional techniques known in the art. For example, wind speed and direction can be obtained through wind measurement towers or lidar; typhoon path information can be accessed from meteorological forecast data; operating parameters such as wind turbine speed, power, yaw angle, and pitch angle can be directly read from the wind turbine main control system or data acquisition and monitoring control system; key structural loads can be measured by installing strain gauges or fiber optic grating sensors on key parts such as the tower and blades; wind farm spatial distribution parameters can be calculated by setting up multiple wind measurement points or by combining wake models; and the spatial coordinates of the wind turbine are determined and recorded in the wind turbine location recording unit during the wind farm construction phase.

[0043] In other words, to achieve data perception, those skilled in the art, after reading the above description, can select appropriate specific implementation methods based on actual working conditions and put them into practice without any creative effort.

[0044] In this embodiment, considering the significant differences in wind speed, wind direction, and load impact experienced by wind turbines at different spatial locations within a wind farm under the influence of a typhoon, data is collected on each individual wind turbine within the wind farm to provide a data foundation for subsequent dynamic risk identification and differentiated control at the wind turbine level. Specifically, data is collected on each individual wind turbine within the wind farm. Taking a typhoon generator as an example, at the current sampling time The collected single-frame input data can be:

[0045] ;

[0046] In the formula, For the first Typhoon turbine units at all times The input data, where the subscript Represents the index of the discrete sampling point in the time series (i.e., the first... (at a certain moment); For a moment Typhoon meteorological and track characteristics, including wind speed, wind direction, air pressure, typhoon center location, and direction of movement; For the first Typhoon turbine units at all times The operating status includes speed, output power, yaw angle, pitch angle and critical structural loads; For a moment Wind field environmental parameters; For the first The spatial location parameters of a typhoon turbine can be its horizontal and vertical coordinates within the wind farm coordinate system; superscript This is the transpose of a vector.

[0047] Because high-frequency disturbances are significant and measurement and communication anomalies are more likely to occur during typhoon conditions, it is necessary to perform filtering and noise reduction, anomaly identification and correction, and normalization on the raw data to improve numerical stability and reduce the risk of error propagation.

[0048] The purpose of normalization is to eliminate the differences in dimensions and orders of magnitude between different physical quantities, so that parameters such as wind speed, power, and load are expressed on the same numerical scale. This avoids gradient update imbalance or slow convergence during subsequent model training due to excessive differences in the orders of magnitude of input features. In this embodiment, minimum-maximum normalization can be used. Its basic principle is to map the original data to the [0,1] interval in a linear proportion. The distribution of the data remains unchanged after the transformation, only the numerical scale is changed. The specific form of minimum-maximum normalization is as follows:

[0049] ;

[0050] In the formula, This represents the normalized output value. For a scalar data to be normalized, that is, the original sampled value to be normalized in a certain data channel; , The minimum and maximum values ​​of this data channel within the training samples or a defined range; when The value is equal to hour, =0; when The value is equal to hour, The value is 1; the remaining values ​​are uniformly mapped between 0 and 1.

[0051] Those skilled in the art should know that filtering and denoising can be achieved using median filtering, low-pass filtering, or wavelet denoising, while anomaly identification and correction can be achieved using statistical thresholding or methods based on standard deviation. Normalization can also be achieved using Z-score standardization or other methods. The specific implementation methods described above are all well-known technologies in the field, and this embodiment does not limit them. Those skilled in the art can choose appropriate means to implement them according to actual working conditions and data characteristics.

[0052] After completing the above processing, standardized input is obtained. The data is written into the wind power cluster's operational status database.

[0053] In other words, each acquisition unit collects data according to a unified sampling period, achieving synchronous acquisition of multi-source data. The standardized data formed after preprocessing of the input data mentioned here is written into the wind power cluster operation status database as a complete input record, and this data also serves as the basic input for subsequent dynamic wind farm prediction models.

[0054] Considering the strong nonlinearity, time-varying nature, and significant spatial differences of typhoon wind fields, relying solely on current measurements makes timely forward-looking control difficult. To address this, this embodiment establishes a dynamic wind field prediction model and employs a BFO-BP hybrid algorithm for parameter optimization training. This model outputs predicted wind field information for the target time domain in advance, thereby obtaining forward-looking wind field change trends during typhoon evolution. Based on this, the predicted information is fused with multi-dimensional data such as typhoon path, wind turbine location, structural load, and prediction errors to calculate the dynamic risk index of each wind turbine and classify risks accordingly. This allows control strategies to be formulated based on predictions of the typhoon's future impact, rather than relying on current measurements.

[0055] Specifically, after the runtime database is built, the cluster collaborative computing module calls the standardized input data in the database to establish a dynamic wind field prediction model. This model is used to output predicted wind field information for the future target time domain:

[0056] ;

[0057] In the formula, For the first Typhoon turbine units at the predicted time The predicted wind field output includes predicted wind speed and predicted wind direction. For dynamic wind field prediction models; This is the model parameter vector; To predict the step size.

[0058] In this embodiment, a hybrid BFO-BP algorithm is used to train the dynamic wind field prediction model for parameter optimization. The BFO algorithm performs a global search optimization of the prediction model parameters by simulating the directional, replication, and dispersal operations of E. coli during foraging, reducing the risk of the model getting trapped in local optima. The BP neural network, based on the prediction error, uses gradient descent to refine the model parameters locally, improving the local convergence accuracy of the prediction results. The two algorithms work together to ensure both the model's global optimization capability and local convergence accuracy.

[0059] To enable the dynamic wind farm prediction model to be trained and evaluated based on standardized input data from the wind farm cluster operation status database, this embodiment uses the error between the predicted wind farm output and the actual observed wind farm output as the optimization objective, defining a prediction error objective function, or loss function, for training the dynamic wind farm prediction model:

[0060] ;

[0061] In the formula, The parameters of the prediction model are The objective function for prediction error at that time, This represents the number of training samples; For the first Input features of each sample; For the first The actual observed wind field output corresponding to each sample; For dynamic wind field prediction models in terms of parameters Next to the The predicted wind field results output for each sample. The predicted wind field results include at least the predicted wind speed and the predicted wind direction.

[0062] In this embodiment, the dynamic wind field prediction model can be implemented using a BP neural network, and the model parameter vector... This includes the connection weights and thresholds between the layers of the BP neural network. The above prediction error objective function... It can be used as both the loss function in the backpropagation training process of the BP neural network and the fitness evaluation function when the BFO algorithm searches for model parameters. In one specific implementation, the parameter optimization training of the dynamic wind field prediction model using the BFO-BP hybrid algorithm includes the following steps:

[0063] First, determine the structure of the input layer, hidden layer, and output layer of the BP neural network. The input layer receives the first... Standardized input data of typhoon generator at the current moment The input data includes typhoon meteorological and path characteristics, wind turbine operating status, wind field environmental parameters, and wind turbine spatial location parameters; the output layer outputs the predicted wind speed and predicted wind direction at the next predicted time.

[0064] Secondly, the connection weights and thresholds in the BP neural network are encoded as a candidate individual in the BFO algorithm, that is, one bacterial individual corresponds to a set of model parameter vectors to be optimized. After initializing the bacterial population, the parameter vector corresponding to each bacterial individual is substituted into the dynamic wind field prediction model, and the prediction error objective function is applied. Calculate the fitness value of this individual bacterium.

[0065] Then, the BFO algorithm performs orientation, replication, and migration operations on bacterial individuals based on their fitness values, so that parameter combinations with smaller prediction errors are retained and expanded, while parameter combinations with larger prediction errors are eliminated or re-searched, thereby obtaining better initial weights and thresholds for the BP neural network globally.

[0066] Furthermore, the optimal or relatively optimal parameter vector obtained by the BFO algorithm is used as the initial parameters of the BP neural network, and the backpropagation algorithm is used to locally refine the parameter vector until the prediction error meets the preset error threshold or the number of iterations reaches the preset upper limit, thus obtaining the trained dynamic wind field prediction model.

[0067] In this way, the BFO algorithm is used to search for better initial weights and thresholds in a large parameter space, reducing the risk of getting trapped in local optima during the training of the BP neural network; the BP neural network is used to correct local errors in the parameter results obtained by the BFO algorithm, improving the ability of the prediction model to fit nonlinear and strongly time-varying wind fields under typhoon conditions.

[0068] After training the dynamic wind field prediction model, the model output can be compared with actual observed wind field data to verify the model's ability to predict wind field trends in the target time domain. Figure 3 As shown, within the same prediction time domain, the predicted wind speed curve and the actual wind speed curve have the same overall trend, which can better track the change process of wind speed first increasing and then decreasing during the typhoon's influence. Figure 3 The mean absolute error (MAE) was 0.260 m / s, and the root mean square error (RMSE) was 0.281 m / s, indicating that the trained dynamic wind field prediction model has a small wind speed prediction error and good trend tracking ability. The predicted wind speed results obtained can serve as the data basis for subsequent calculations of the predicted wind speed risk factor and prediction error risk factor, and can be used together with information such as predicted wind direction, typhoon path, unit spatial location, and key structural loads for the calculation of the dynamic risk index.

[0069] like Figure 4 As shown, the implementation method of this embodiment is compared with the training error convergence process of traditional backpropagation algorithm and other prediction methods. The BFO-BP hybrid algorithm used in this embodiment has a faster error convergence speed and a lower final prediction error during the iteration process. This shows that by using the BFO algorithm to perform global search optimization of model parameters and combining it with the BP neural network for error backpropagation correction, the model training efficiency and prediction accuracy can be effectively improved.

[0070] Furthermore, after obtaining the predicted wind field information, a dynamic risk index is constructed, which unifies the typhoon path evolution, wind turbine spatial location, predicted wind field intensity, key structural loads, and prediction errors into a risk assessment quantity that can be used for scheduling decisions. For the first... The distance between the typhoon generator and the predicted center of the typhoon's influence can be:

[0071] ;

[0072] In the formula, For the first Typhoon turbine units at the predicted time Distance from the predicted center of typhoon impact; For the first The horizontal coordinate of the typhoon turbine in the wind farm coordinate system. For the first The longitudinal coordinate of the typhoon turbine in the wind farm coordinate system; the smaller this distance, the closer the wind turbine is to the center of the typhoon's impact, and the higher the potential impact risk.

[0073] To unify the conversion of predicted wind speed, wind direction changes, structural loads, typhoon distance, and prediction errors into risk assessment metrics, a set of risk factors is constructed:

[0074] ;

[0075] in, For the first Typhoon turbine units at all times The set of risk factors; This is a wind speed risk factor used to reflect the impact of the maximum wind speed within the future forecast time domain; This is a wind direction change risk factor, used to reflect the impact of rapid wind direction changes on yaw control and structural loads; This is a load risk factor used to reflect the degree to which the load on the critical structure of the wind turbine approaches the safety limit; This is a typhoon distance risk factor, used to reflect the degree to which wind turbines are close to the center of typhoon influence; This is a prediction error risk factor used to reflect the uncertainty of the prediction model under the current typhoon conditions.

[0076] Those skilled in the art should understand that the aforementioned risk factors can be calculated from the ratio of predicted wind speed to the safe wind speed limit, the ratio of predicted wind direction change to the wind direction change limit, the ratio of predicted critical structural load to the load safety limit, the degree of closeness between the distance between the wind turbine and the typhoon's impact center relative to the typhoon's impact distance threshold, and the relative error between the predicted output and the actual observed output. The logic is as follows: the higher the wind speed, the more drastic the wind direction change, the closer the load is to the safety limit, the closer the wind turbine is to the typhoon's impact center, and the greater the prediction error, the higher the risk to the wind turbine.

[0077] Based on the aforementioned risk factors, the weighted sum of each risk factor is used to calculate the... Dynamic risk index of typhoon generator units:

[0078] ;

[0079] In the formula, For the first Typhoon turbine units at all times The dynamic risk index; Weighting coefficients for predicting wind speed risk factors; The weighting coefficients for the risk factor of sudden wind direction change; These are the weighting coefficients for the load risk factor; The weighting coefficients for the typhoon distance risk factor; These are the weighting coefficients for the prediction error risk factors.

[0080] The weighting coefficients of each risk factor satisfy the following:

[0081] ;

[0082] ;

[0083] Each weighting coefficient is used to characterize the contribution of different risk factors to the disaster risk of wind turbines. In the scenario of collaborative disaster prevention and control during typhoons, the weighting coefficients can be predetermined based on historical typhoon operation data of the wind farm, design safety limits of wind turbines, expert experience, or offline simulation results. Specifically, prioritizing structural safety, the predicted wind speed risk factor, load risk factor, and wind direction change risk factor can be assigned higher weights, while the typhoon distance risk factor and prediction error risk factor can be used as auxiliary correction weights.

[0084] The aforementioned weighting coefficients can be adjusted based on actual operating data for different wind farms or typhoon levels. For example, when the critical structural load of the wind turbine approaches the safety limit, the weight of the load risk factor can be increased. When the rate of wind direction change is large and the yaw system response pressure increases, the weight of the wind direction change risk factor should be increased. When the prediction error increases continuously, increase the weight of the prediction error risk factor. This enhances the sensitivity of the risk index to forecast uncertainties. Through these methods, the dynamic risk index can reflect both the intensity of the typhoon wind field and take into account the load on the generator structure, its spatial location, and the reliability of model predictions.

[0085] In other words, the dynamic wind field prediction model in this embodiment can not only obtain the future wind field change trend, but also further identify the risk differences of different wind turbines in the typhoon path evolution process, avoiding the problem of relying only on the predicted wind field without wind turbine-level risk assessment.

[0086] In one embodiment, based on the forward-looking prediction information (i.e., predicted wind field information, dynamic risk index, and prediction error) of each wind turbine in the future time domain obtained through the dynamic wind field prediction model described in the previous embodiment, this embodiment further considers the spatial dimension of this prediction information to form a zoning strategy that can guide the execution of specific control commands. It should be noted that although the dynamic risk index can quantify the risk level of each wind turbine, if it only stays at the index level, the control strategy still lacks clear execution boundaries and differentiated action commands. That is, what level of protection measures should be taken for high-risk wind turbines, and to what extent can low-risk wind turbines participate in power support, all require further spatial zoning for decision-making.

[0087] This embodiment sets up four levels: core impact zone, transitional load reduction zone, peripheral compensation zone, and recovery zone. This maps the dynamic risk index of each wind turbine to a spatial region with clear control semantics, so that the prediction information is ultimately transformed into specific action instructions.

[0088] Specifically, after obtaining the dynamic risk index of each wind turbine, the cluster collaborative calculation module divides the wind power cluster into different risk zones based on the risk index. The design concept is that the impact of a typhoon does not act on all wind turbines simultaneously and with equal intensity. Instead, high-risk wind turbines should be prioritized for protection, low-risk wind turbines can undertake power compensation tasks within a safe range, and wind turbines in the recovery zone should be protected from sudden recovery that could cause new load impacts.

[0089] Let the risk threshold of the core impact zone be... The risk threshold for the transition de-loading area is The risk threshold for the recovery zone is And satisfy Among them, the first Typhoon turbine units at all times The risk zoning results are as follows:

[0090] ;

[0091] In the formula, For the first Typhoon turbine units at all times Risk zoning results; The core impact zone; This is a transitional unloading zone; For the outer compensation zone; This is the recovery area. In other words, when... At that time, the wind turbine was classified as part of the core impact zone. ;when At that time, the wind turbine was designated as a transitional load reduction zone. ;when At that time, the wind turbine was included in the outer compensation zone. ;when At that time, the wind turbine was designated as a recovery zone. .

[0092] The core impact zone is used to characterize wind turbine areas with a high risk of direct impact from typhoons; the transitional load reduction zone is used to characterize wind turbine areas where the impact of typhoons has intensified but has not yet reached the highest risk level; the outer compensation zone is used to characterize wind turbine areas with relatively high load margins that can undertake certain power compensation tasks under safety constraints; and the recovery zone is used to characterize wind turbine areas where the risk is gradually decreasing and can gradually return to normal operation.

[0093] After completing the risk zoning, a risk-adaptive multi-objective collaborative optimization model is established. The control variable vector for the i-th wind turbine is:

[0094]

[0095] In the formula, For the first Typhoon turbine units at all times The vector of control variables; This is the yaw angle control value; This is the pitch angle control value. This is the power control command value; superscript For control command values; superscript This is the transpose of a vector.

[0096] In one specific implementation, the risk zoning results are used to determine the basic control principles, control variable constraint boundaries, and initial scheduling commands for each wind turbine. The risk-adaptive multi-objective collaborative optimization model further solves for specific control quantities based on these parameters. Specifically, when the i-th wind turbine is classified as being within the core impact zone, it indicates a high risk of direct typhoon impact. In this case, prioritizing structural safety and load suppression, commands such as early feathering, power-limited operation, or shutdown protection can be generated to adjust the power control command value. Reduce the power to a preset low percentage range of rated power, or reduce it to zero when shutdown conditions are met, while simultaneously increasing the pitch angle control value. This allows the blades to gradually enter a feathering state, and the yaw angle control is adjusted according to the current wind direction and the unit's safety strategy. This is to reduce the extreme loads on the blades and tower.

[0097] When the i-th wind turbine is designated as part of the transitional load reduction zone, it indicates that the wind turbine is being increasingly affected by the typhoon but has not yet reached the highest risk level. At this time, based on the control principles of preventative load reduction and smooth transition, pre-yaw load reduction commands and power limiting commands can be generated to... Reduce the load according to the preset reduction ratio, and adjust appropriately. and This allows for the reduction of structural loads before the typhoon's impact intensifies, thus avoiding delays in control actions.

[0098] When the i-th wind turbine is included in the outer compensation zone, it indicates that the wind turbine has a relatively high load margin and a certain power support capacity. At this time, the power compensation under safety constraints is the control principle. Under the premise of not exceeding the rated power, structural load safety limits and grid connection scheduling constraints, its power control command value can be increased or maintained. It is used to compensate for the power deficit caused by load reduction or shutdown of wind turbines in the core impact zone and transitional load reduction zone, while keeping the pitch angle and yaw angle control values ​​within a stable adjustment range.

[0099] When the i-th wind turbine is included in the recovery zone, it indicates that the impact of the typhoon on that wind turbine is gradually weakening. At this time, based on the control principles of smooth recovery and avoiding secondary load impacts, a gradual power recovery command can be generated to enable... Gradually restore the system to the normal operating target value at the preset ramp rate, while limiting... and The range of change should be controlled to avoid sudden changes in power, pitch angle, or yaw angle that could lead to new structural load impacts.

[0100] The risk-adaptive multi-objective collaborative optimization function is established as follows:

[0101] ;

[0102] In the formula, For a moment The objective function of collaborative control, subscript This objective function is used for cooperative control; For wind power clusters at all times The set of all control variables; For the set of control variables Find the minimum value of the objective function. For the first to the second Summation of typhoon generator units; For the first Target weights of structural loads on typhoon generator units; For the first The typhoon turbine unit at the next predicted time Predicting critical structural loads; For the first Safety limits for critical structural loads of typhoon generator units; For the first Target weights for the power output of typhoon generator units; For the first The typhoon turbine unit at the next predicted time The predicted output power; For the first Rated power of typhoon generator sets; For the first Control smoothing weights for typhoon generator units; For the first The previous control time of the typhoon generator The control variable vector.

[0103] In this objective function, the first term reduces structural load, the second term is preceded by a negative sign, so minimizing the objective function is equivalent to increasing power output, and the third term limits abrupt changes in yaw angle, pitch angle, and power command, making the control process smoother. The purpose of this setup is to ensure that the control strategy does not only pursue single power generation or single load reduction, but also simultaneously considers structural safety, power output, and control smoothness under typhoon conditions.

[0104] To prioritize the protection of high-risk wind turbines and provide appropriate compensation for low-risk wind turbines, the target weights for structural loads and power output are adaptively adjusted based on the risk index. Specifically:

[0105] ;

[0106] ;

[0107] In the formula, For the first Target weights of structural loads on typhoon generator units; The structural load target is the basic weight, with the superscript 0 representing the basic value; This is an adjustment coefficient that increases the target weight of the structural load as the risk index increases; For the first Dynamic risk index of typhoon generator units; For the first Target weights for the power output of typhoon generator units; This represents the minimum allowable value for the power output target weight. The basic weights for the power output target; This is an adjustment coefficient that reduces the target weight of power output as the risk index increases.

[0108] The algorithm works as follows: when the risk index of a wind turbine increases, its load suppression weight is increased while its power output weight is decreased, allowing the wind turbine to prioritize load reduction or protection; when the risk index of a wind turbine is low, its power output weight is relatively high, enabling it to undertake power compensation tasks while meeting load safety constraints. This approach addresses the problems of insufficient protection for high-risk wind turbines and excessive load reduction for low-risk wind turbines in existing unified optimization strategies.

[0109] After completing the multi-objective collaborative optimization solution, differentiated collaborative scheduling and control command sets are generated based on the risk zoning results. Specifically, for wind turbines in the core impact zone, advance feathering or shutdown protection commands are generated; for wind turbines in the transitional load reduction zone, pre-yaw load reduction commands are generated; for wind turbines in the peripheral compensation zone, power compensation commands are generated; and for wind turbines in the recovery zone, smooth recovery commands are generated. In this way, by implementing differentiated control for wind turbines in different risk areas, coordinated control and power allocation of different wind turbines can be achieved.

[0110] In one exemplary implementation, a wind farm includes four wind turbines, and at the current control time... Based on the dynamic risk index calculation results, the cluster collaborative computing module divides the first wind turbine into the core impact zone, the second wind turbine into the transitional load reduction zone, the third wind turbine into the peripheral compensation zone, and the fourth wind turbine into the recovery zone.

[0111] For the first wind turbine located in the core impact zone, the cluster collaborative computing module generates an early feathering or shutdown protection command, which gradually increases the pitch angle control value to the feathering direction and reduces the power control command value to a low power protection state or a zero power state, so as to prioritize reducing the extreme loads borne by the blades and tower.

[0112] For the second wind turbine in the transitional load reduction zone, the cluster collaborative computing module generates a pre-yaw load reduction command and a power-limited operation command, so that its yaw angle control value is adjusted in advance according to the predicted wind direction change, and the power control command value is reduced to a preset proportion of the rated power, so as to release the structural load in advance before the typhoon impact intensifies further.

[0113] For the third wind turbine located in the outer compensation zone, the cluster collaborative computing module generates a compensation power command provided that its critical structural load does not exceed the safety limit. This command maintains or increases the power control command value to a higher level within the safe allowable range, in order to compensate for the power deficit caused by the load reduction or shutdown of the first and second wind turbines.

[0114] For the fourth wind turbine in the recovery zone, the cluster collaborative computing module generates a smooth recovery command, which gradually restores its power control command value to the normal operation target value according to the preset ramp rate, and limits the variation of pitch angle control and yaw angle control to avoid new structural load impacts caused by the rapid recovery process after the typhoon's impact weakens.

[0115] As can be seen from the above examples, the risk zoning results are not only used for risk level marking, but are directly used to determine the yaw angle control value, pitch angle control value and power control command value of each wind turbine, so as to give priority protection to high-risk wind turbines, reduce the load of wind turbines with transitional risk in advance, provide safety compensation for low-risk wind turbines, and smoothly restore wind turbines in the recovery zone.

[0116] Furthermore, after generating the overall control command set, the wind turbine control module sends control commands to the actuators of each wind turbine. The disaster mitigation execution module drives the wind turbines to perform yaw adjustment, pitch adjustment, power limiting operation, power compensation operation, smooth recovery control, or feathering shutdown control. The design concept is based on the fact that typhoon paths and wind field conditions are constantly changing. If risk zones are calculated only at a single moment, they are prone to deviating from the actual operating conditions. Therefore, it is necessary to use execution feedback to continuously correct the risk index and risk zones.

[0117] Specifically, after execution, key states and loads are collected to form a feedback vector:

[0118] ;

[0119] In the formula, For the first Typhoon turbine units at all times The execution status feedback vector; This is the actual yaw angle; This is the actual pitch angle; This refers to the actual rotational speed; This refers to the actual output power. For critical structural loads; superscript The corresponding variable is the actual executed value or the actual measured value; superscript This is the transpose of a vector.

[0120] Based on the execution status feedback vector and the predicted risk index, the dynamic risk index for the next control moment is corrected:

[0121] ;

[0122] In the formula, For the first The typhoon generator unit at the next control moment Revised dynamic risk index; superscript To update the value; This is the predicted risk index calculated based on the predicted wind field and typhoon track; superscript This is a predicted value; The feedback risk index is calculated based on the execution feedback status and the actual load. This is the feedback correction coefficient, used to adjust the proportions of the predicted risk index and the feedback risk index in the updated risk index, and it satisfies... .

[0123] The algorithm logic for the above feedback correction is as follows: when the prediction model is relatively consistent with the actual operating state, the error can be reduced. More forward-looking control is based on prediction results; when the actual load, power, or yaw response deviates significantly from the prediction results, the control can be increased. The risk index and risk zoning are further revised based on feedback results. In one specific implementation, adjustments can be made based on real-time changes in the prediction error. The value of is automatically increased when the variance of the recent prediction error increases. The value is increased to enhance the feedback effect, and vice versa. The value is intended to maintain a forward-looking approach to control. (Revised) Re-enter the risk zoning judgment process to achieve dynamic rolling updates of risk zoning.

[0124] Specifically, at the next control moment The cluster collaborative computing module will correct the dynamic risk index. The risk zoning result for the i-th wind turbine is redefined by comparing it with the preset risk zoning threshold. Let the risk threshold for the core impact zone be... The risk threshold for the transition de-loading area is The risk threshold for the recovery zone is And satisfy Then, the risk zoning result of the i-th wind turbine at the next control moment can be expressed as:

[0125] ;

[0126] In the formula, This represents the risk partitioning result for the i-th wind turbine at the next control time. The core impact zone; This is a transitional unloading zone; For the outer compensation zone; This is the recovery area.

[0127] In one specific implementation, when the dynamic risk index is normalized to the [0,1] interval, it can be set to... , , That is, when At that time, the wind turbine unit was classified as part of the core impact zone; when When, it is classified as a transitional de-loading zone; when When, it is included in the outer compensation zone; when When the time comes, it is classified into the recovery zone. In this way, the corrected dynamic risk index is not only used to update the risk assessment results, but also directly used to determine the risk zoning at the next control time. This allows the risk zoning to be updated in a rolling manner as the typhoon path changes, the prediction error changes, and the actual execution feedback of the wind turbines changes, avoiding the lag in control strategies caused by only using the zoning results of the previous time.

[0128] In one embodiment, Figure 1 A system block diagram is shown, which realizes the collaborative control of wind power clusters based on the BFO-BP hybrid algorithm. Figure 1 The system includes:

[0129] The data sensing module is configured to collect data on each individual wind turbine unit within the wind farm.

[0130] The cluster collaborative computing module is configured to call the data collected by the data sensing module to predict wind field information in the future target time domain, and construct a risk factor set based on the prediction results; calculate the dynamic risk index of each wind turbine at the current control moment by weighted summation of each risk factor; divide the wind power cluster into at least four different risk areas according to the risk index; and generate differentiated collaborative scheduling and control instruction sets according to different risk areas.

[0131] as well as

[0132] The wind turbine control module is configured to send control commands from the control command set to the actuators of each wind turbine.

[0133] Optionally, the data collected by the data sensing module includes typhoon meteorological and path characteristics, wind turbine operating status, wind farm environmental parameters, and wind turbine spatial location parameters.

[0134] Optionally, the set of risk factors includes predicted wind speed risk factors, wind direction change risk factors, load risk factors, typhoon distance risk factors, and prediction error risk factors.

[0135] Optionally, the at least four different risk zones include the core impact zone, the transition de-load zone, the peripheral compensation zone, and the recovery zone.

[0136] Optionally, the cluster collaborative computing module is also configured to perform rolling corrections on the risk index and risk partitions based on the execution feedback of the wind turbine.

[0137] In one embodiment, Figure 2 The specific steps of a method are illustrated in the figure, which includes the following steps:

[0138] Data is acquired by collecting data from each individual wind turbine within the wind farm.

[0139] Based on the risk zoning results, a differentiated set of collaborative scheduling and control instructions is generated; wherein the sub-steps of risk zoning include: predicting wind field information in the future target time domain and constructing a set of risk factors based on the prediction results; calculating the dynamic risk index of each wind turbine at the current control moment by weighted summation of each risk factor; and dividing the wind power cluster into at least four different risk zones based on the risk index.

[0140] The control commands from the control command set are sent to the actuators of each wind turbine.

[0141] It should be noted that since the specific implementation principles of the above steps have been disclosed in the above embodiments, they will not be repeated here to avoid repetition.

[0142] like Figure 5 As shown, after adopting the collaborative control method of this invention, the critical structural load of the wind turbine is significantly reduced compared to before collaborative control, while the power response remains at a relatively optimal level. This demonstrates that this invention can reduce the risk of typhoon impact while maintaining the power generation capacity of the wind power cluster, thereby achieving a coordinated balance between the safety and operational economy of the wind turbine. Therefore, this invention, through a closed-loop control mechanism of data perception, dynamic prediction, collaborative optimization, control execution, and state feedback, can effectively improve the safe and stable operation capability of wind power clusters under extreme typhoon weather conditions.

[0143] Through the above embodiments, this invention enables wind turbine-level risk identification and zoned control of wind power clusters under conditions of continuously changing typhoon paths, uneven spatial distribution of wind fields, and different load states of different wind turbines. Compared with a unified optimization control method for the entire field, this embodiment can provide early protection for wind turbines in high-risk areas, pre-emptively reduce the load on wind turbines in medium-risk areas, compensate the power of wind turbines in low-risk areas, and smoothly restore power after the typhoon's impact weakens. This reduces the load on critical structures while maintaining the overall power output stability of the wind power cluster.

Claims

1. A system for collaborative control of wind power clusters, characterized in that, include: The data sensing module is configured to collect data on each individual wind turbine unit within the wind farm. The cluster collaborative computing module is configured to call the data collected by the data sensing module to predict wind field information in the future target time domain, and construct a set of risk factors based on the prediction results. The dynamic risk index of each wind turbine at the current control moment is calculated by weighted summation of each risk factor. Based on the risk index, the wind power cluster is divided into at least four different risk zones; Generate differentiated collaborative scheduling and control instruction sets based on different risk areas; as well as The wind turbine control module is configured to send control commands from the control command set to the actuators of each wind turbine.

2. The system according to claim 1, characterized in that, The data collected by the data sensing module includes typhoon meteorological and path characteristics, wind turbine operating status, wind farm environmental parameters, and wind turbine spatial location parameters.

3. The system according to claim 1, characterized in that, The set of risk factors includes predicted wind speed risk factors, wind direction change risk factors, load risk factors, typhoon distance risk factors, and prediction error risk factors.

4. The system according to claim 1, characterized in that, The at least four distinct risk zones include the core impact zone, the transition deload zone, the peripheral compensation zone, and the recovery zone.

5. The system according to claim 4, characterized in that, The cluster collaborative computing module is also configured to perform rolling corrections to the risk index and risk partitions based on the execution feedback from the wind turbine units.

6. A method for collaborative control of wind power clusters, characterized in that, Includes the following steps: Data is acquired by collecting data from each individual wind turbine within the wind farm. Generate differentiated collaborative scheduling and control instruction sets based on the risk zoning results; The sub-steps of risk zoning include: predicting wind field information in the future target time domain and constructing a set of risk factors based on the prediction results; calculating the dynamic risk index of each wind turbine at the current control time by weighted summation of each risk factor; and dividing the wind power cluster into at least four different risk zones according to the risk index. The control commands from the control command set are sent to the actuators of each wind turbine.

7. The method according to claim 6, characterized in that, The prediction of wind field information in the future target time domain employs a BFO-BP hybrid algorithm to optimize and train the parameters of the dynamic wind field prediction model. Specific steps include: Establish a dynamic wind field prediction model: ； ； In the formula, For the first Typhoon turbine units at the predicted time The predicted wind field output, of which To predict the step size; For dynamic wind field prediction models; For the first Typhoon turbine units at all times Input data; This is the model parameter vector; For a moment Typhoon meteorological and path characteristics; For the first Typhoon turbine units at all times The running status; For a moment Wind field environmental parameters; For the first Spatial location parameters of typhoon generator units; Define the objective function for prediction error as follows: ； In the formula, The parameters of the prediction model are The objective function for predicting error at that time; This represents the number of training samples; For the first Input features of each sample; For the first The actual observed wind field output corresponding to each sample; For dynamic wind field prediction models in terms of parameters Next to the The predicted wind field results output for each sample.

8. The method according to claim 6, characterized in that, The expression for calculating the dynamic risk index of each wind turbine at the current control moment is as follows: ； In the formula, For the first Typhoon turbine units at all times The dynamic risk index; Weighting coefficients for predicting wind speed risk factors; The weighting coefficients for the risk factor of sudden wind direction change; These are the weighting coefficients for the load risk factor; The weighting coefficients for the typhoon distance risk factor; These are the weighting coefficients for the prediction error risk factors; To predict wind speed risk factors; As a risk factor for sudden changes in wind direction; For load risk factors; The distance risk factor for typhoons; For prediction error risk factors; The weighting coefficients of each risk factor satisfy the following: ； 。 9. The method according to claim 6, characterized in that, The at least four distinct risk zones include the core impact zone, the transition de-loading zone, the peripheral compensation zone, and the recovery zone, and their division is based on the following criteria: ； In the formula, For the first Typhoon turbine units at all times Risk zoning results; The core impact zone; This is a transitional unloading zone; For the outer compensation zone; This is the recovery area; among which, The risk threshold for the core impact zone, As a risk threshold for the transitional de-loading zone, The risk threshold for the recovery zone, and meets the following conditions. .

10. The method according to claim 6, characterized in that, The step of generating differentiated collaborative scheduling and control instruction sets based on different risk areas includes: Establish a risk-adaptive multi-objective collaborative optimization model, where the control variable vector of the i-th wind turbine is: ； In the formula, For the first Typhoon turbine units at all times The vector of control variables; This is the yaw angle control value; This is the pitch angle control value; This is the power control command value, where the superscript... For control command values; superscript Transpose of a vector; Establish a risk-adaptive multi-objective collaborative optimization function: ； In the formula, For a moment The collaborative control objective function; For wind power clusters at all times The set of all control variables; For the set of control variables Find the minimum value of the objective function. For the first Target weights of structural loads on typhoon generator units; For the first The typhoon turbine unit at the next predicted time Predicting critical structural loads; For the first Safety limits for critical structural loads of typhoon generator units; For the first Target weights for the power output of typhoon generator units; For the first The typhoon turbine unit at the next predicted time The predicted output power; For the first Rated power of typhoon generator sets; For the first Control smoothing weights for typhoon generator units; For the first The previous control time of the typhoon generator The vector of control variables; The target weights for structural loads and power output are adaptively adjusted based on the risk index, specifically: ； ； In the formula, The basic weights for structural load targets; This is an adjustment coefficient that increases the target weight of the structural load as the risk index increases; For the first Dynamic risk index of typhoon generator units; This represents the minimum allowable value for the power output target weight. The basic weights for the power output target; The adjustment coefficient is used to reduce the target weight of power output as the risk index increases; After execution, key states and loads are collected to form a feedback vector: ； In the formula, For the first Typhoon turbine units at all times The execution status feedback vector; This is the actual yaw angle; This is the actual pitch angle; This refers to the actual rotational speed; This refers to the actual output power. For critical structural loads, the superscript indicates The corresponding variable is the actual executed value or the actual measured value; superscript This is the transpose of a vector. Based on the execution status feedback vector and the predicted risk index, the dynamic risk index for the next control moment is corrected: ； In the formula, For the first The typhoon generator unit at the next control moment The revised dynamic risk index, where the superscript... To update the value; This is the predicted risk index calculated based on the predicted wind field and typhoon track, where the superscript... This is a predicted value; The feedback risk index is calculated based on the execution feedback status and the actual load, where: When the prediction model closely matches the actual operating state, reduce ; When the actual load, power, or yaw response deviates from the predicted result, increase ; Based on the revised Re-enter the risk zoning assessment process.

Images