Control method, device and equipment of wind turbine generator system

Abstract

The application provides a wind turbine control method, device and equipment, and relates to the technical field of wind power generation. The method comprises the following steps: obtaining operation data and wind speed data of a wind turbine, performing prediction processing on the operation data and the wind speed data by using a pre-obtained load prediction model, obtaining predicted load data of a blade of the wind turbine and a target wind-receiving moment; determining a target wind-approaching angle of the blade according to the predicted load data; and adjusting a current wind-approaching angle of the blade to the target wind-approaching angle before the target wind-receiving moment, so as to control the operation of the wind turbine. The above method realizes the advanced adjustment of the wind-approaching angle of the blade, effectively improves the wind energy utilization rate, improves the power generation of the wind turbine under the premise of ensuring the safe operation of the wind turbine, and improves the economy and reliability of wind power generation.

Description

Technical Field

[0001] This application relates to the field of wind power generation technology, and in particular to a control method, device and equipment for a wind turbine generator set. Background Technology

[0002] In wind farms, wind turbine generators are usually installed in open or complex terrain areas. The wind conditions in these areas are highly unstable, random, and abrupt, and extreme wind conditions such as gusts, turbulence, and wind shear are prone to occur, which can cause impact overload on the blades of the wind turbine generators.

[0003] In existing technologies, the control of wind turbine generators relies on wind speed data detected by lidar. When the current wind speed is detected to be greater than a wind speed threshold, it is determined that the blades will bear excessive load. To avoid blade damage, the blade's angle of attack is usually adjusted to zero to reduce the impact of sudden wind speed changes. However, this method of directly adjusting the blade's angle of attack to zero after detecting that the current wind speed is greater than the wind speed threshold leads to a reduction in the power generation of the wind turbine generator. Summary of the Invention

[0004] This application provides a control method, device, and equipment for wind turbine generator sets to improve the power generation of wind turbine generator sets.

[0005] In a first aspect, this application provides a control method for a wind turbine generator set, the wind turbine generator set comprising multiple blades, the method comprising:

[0006] The wind turbine generator set's operating data and wind speed data are acquired, wherein the wind speed data includes wind speed sequences corresponding to multiple sampling points located at different distances in front of the wind turbine generator set;

[0007] By using a pre-acquired load prediction model, the operating data and the wind speed data are processed to predict the load data of the blade and the target wind exposure time. The predicted load corresponding to the target wind exposure time is greater than or equal to the preset load.

[0008] Based on the predicted load data, the target windward angle of the blade is determined;

[0009] Before the target wind arrival time, the current windward angle of the blades is adjusted to the target windward angle to control the operation of the wind turbine generator set.

[0010] In one possible implementation, determining the target windward angle of the blade based on the predicted load data includes:

[0011] Determine the target load range where the predicted load is located at the time when the target is exposed to wind;

[0012] Based on the preset correspondence between load range and windward angle, the target windward angle corresponding to the target load range is determined.

[0013] In one possible implementation, adjusting the current windward angle of the blade to the target windward angle before the target wind arrival time includes:

[0014] Based on the wind speed data, the expected wind exposure time of the blades is determined using a dynamic time warping algorithm;

[0015] If the difference between the target wind arrival time and the expected wind arrival time is less than a preset threshold, then at a time that is a preset time earlier than the target wind arrival time, the current windward angle of the blade is adjusted to the target windward angle.

[0016] In one possible implementation, determining the expected wind arrival time of the blades based on the wind speed data using a dynamic time warping algorithm includes:

[0017] The dynamic time warping algorithm is used to align the wind speed sequences corresponding to multiple sampling points in the wind speed data in time, so as to obtain the propagation time difference of wind events between adjacent sampling points.

[0018] Calculate multiple propagation velocities based on the propagation time difference between adjacent sampling points and the corresponding sampling point spacing;

[0019] The multiple propagation speeds are fused together to obtain the target propagation speed;

[0020] The expected wind arrival time is calculated based on the target propagation speed and the distance between the reference sampling point and the blade, wherein the distance between the reference sampling point and the blade is less than the distance between any other sampling point and the blade.

[0021] In one possible implementation, before performing prediction processing on the operating data and the wind speed data using a pre-acquired load prediction model to obtain the predicted load data of the blades and the target wind arrival time, the method further includes:

[0022] Multiple training samples are obtained. Each training sample includes the historical operating data, historical wind speed data, corresponding historical load data, and historical wind exposure time of the wind turbine generator set. The load corresponding to the historical wind exposure time is greater than or equal to the preset load.

[0023] The initial load prediction model based on the neural network is trained using the multiple training samples to obtain the load prediction model.

[0024] In one possible implementation, obtaining training samples includes:

[0025] Based on a sliding time window of preset length, the original load data is obtained from the unit status database of the wind turbine generator set;

[0026] Based on the turbulence freezing theorem, the original wind speed data and original operating data corresponding to the original load data are determined.

[0027] The original load data, original wind speed data, and original operation data are expanded to obtain historical load data, historical operation data, and historical wind speed data within a historical time range. The time range of each expanded data includes the time interval covered by the sliding time window.

[0028] The moment when the load in the historical load data first exceeds or equals the preset load is taken as the historical wind arrival moment;

[0029] Training samples are constructed based on the historical load data, the historical operation data, the historical wind speed data, and the historical wind exposure times.

[0030] In one possible implementation, the process of obtaining raw load data from the wind turbine generator's status database based on a preset length sliding time window includes:

[0031] Obtain the original load sequence within the sliding time window from the unit status database;

[0032] The original payload sequence is processed to obtain the payload data envelopment;

[0033] The instantaneous load fluctuation time series with different gradients are extracted from the load data envelopment, and the instantaneous load fluctuation time series are determined as the original load data.

[0034] Secondly, this application provides a control device for a wind turbine generator set, the device comprising:

[0035] The acquisition module is used to acquire the operating data and wind speed data of the wind turbine generator set. The wind speed data includes wind speed sequences corresponding to multiple sampling points located at different distances in front of the wind turbine generator set.

[0036] The first processing module is used to perform predictive processing on the operating data and the wind speed data through a pre-acquired load prediction model to obtain the predicted load data of the blade and the target wind exposure time, wherein the predicted load corresponding to the target wind exposure time is greater than or equal to the preset load.

[0037] The second processing module is used to determine the target windward angle of the blade based on the predicted load data.

[0038] The control module is used to adjust the current windward angle of the blades to the target windward angle before the target wind arrival time, so as to control the operation of the wind turbine generator set.

[0039] Thirdly, this application provides a controller, including: a memory and a processor;

[0040] The memory stores computer-executed instructions;

[0041] The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any of the first aspects.

[0042] Fourthly, this application provides a wind turbine generator set, including a controller as described in the third aspect.

[0043] Fifthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method described in any of the first aspects above.

[0044] In a sixth aspect, this application provides a computer program product, including a computer program that, when executed by a processor, is used to implement the method described in any of the first aspects above.

[0045] This application provides a control method, apparatus, and equipment for a wind turbine generator set, including acquiring operating data and wind speed data of the wind turbine generator set, wherein the wind speed data includes wind speed sequences corresponding to multiple sampling points located at different distances in front of the generator set; performing predictive processing on the operating data and wind speed data using a pre-trained load prediction model to obtain predicted load data for the blades and the target wind arrival time when the predicted load is greater than or equal to a preset load; determining the target windward angle of the blades based on the predicted load data, and adjusting the current windward angle of the blades to the target windward angle before the target wind arrival time arrives, thereby controlling the operation of the wind turbine generator set. This method, by constructing a control logic that coordinates the perception of the forward wind field sequence and load prediction, can predict the risk of blade overload in advance and accurately adjust the windward angle, avoiding the impact of extreme wind conditions on the blades. While ensuring the safe and stable operation of the generator set, it reduces power generation loss and improves the accuracy and operational stability of the wind turbine generator set control. Attached Figure Description

[0046] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0047] Figure 1 A schematic diagram illustrating the application scenarios provided in the embodiments of this application;

[0048] Figure 2 A flowchart illustrating an embodiment of the control method for a wind turbine generator provided in this application;

[0049] Figure 3 A flowchart illustrating Embodiment 2 of the control method for the wind turbine generator provided in this application;

[0050] Figure 4 A flowchart illustrating Embodiment 3 of the control method for the wind turbine generator provided in this application;

[0051] Figure 5 A flowchart illustrating an example of a control method for a wind turbine generator provided in this application embodiment;

[0052] Figure 6 A schematic diagram of the structure of the control device for the wind turbine generator set provided in the embodiments of this application;

[0053] Figure 7 This is a schematic diagram of the controller provided in an embodiment of this application.

[0054] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0055] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application.

[0056] As the wind power industry develops towards larger capacity, offshore operation, and higher reliability, higher requirements are placed on the operational safety and power generation efficiency of wind turbine generators under complex wind conditions.

[0057] Blade load control is a core component for ensuring the structural safety of the wind turbine and reducing fatigue damage. Its control accuracy and responsiveness directly determine the service life, power generation revenue, and operational stability of the wind turbine. Under variable wind conditions such as gusts, turbulence, and wind shear, identifying load risks in advance and adjusting the blade angle of attack appropriately are crucial for improving the safety and economy of the wind turbine.

[0058] In existing technologies, the control of wind turbine generators largely relies on real-time wind speed data detected by lidar. When the current wind speed exceeds a threshold, the blade angle is directly adjusted to zero to reduce the impact of sudden wind speed changes on the blades. However, this passive response control method has significant shortcomings: firstly, relying solely on real-time wind speed makes it impossible to anticipate the risk of blade overload; secondly, the control logic of directly reducing the blade angle to zero is too simplistic. While it protects the blades, it significantly reduces the power generation of the wind turbine generator, making it difficult to balance safety and power generation efficiency.

[0059] To address the aforementioned problems, the inventors considered establishing an advanced load prediction and blade angle adjustment mechanism using wind speed data and turbine operation data in front of the wind turbine, aiming to improve power generation while ensuring blade safety. Based on this, after numerous experiments, the inventors discovered that it is possible to acquire wind turbine operation data and wind speed data, including wind speed sequences corresponding to multiple sampling points at different distances in front of the wind turbine. By using a pre-acquired load prediction model to predict the operation data and wind speed data, the inventors obtained predicted load data for the blades and the target windfall time. Based on the predicted load data, the inventors determined the target windfall angle of the blades and adjusted the current windfall angle to the target windfall angle before the target windfall time. This ensures smooth operation of the wind turbine and avoids power generation loss caused by directly adjusting the blade windfall angle to zero. This effectively improves the power generation efficiency of the wind turbine while ensuring the blades are not subjected to impact overload. Based on this, this application proposes a wind turbine control method that aims to balance blade load safety and turbine power generation efficiency, thereby improving the operational stability of the wind turbine.

[0060] Figure 1 This is a schematic diagram illustrating an application scenario provided in an embodiment of this application. Please refer to [link / reference]. Figure 1 Sensors deployed on wind turbine generators can collect the generator's operating data and wind speed data from multiple sampling points at different distances in front of the generator. The collected operating data and wind speed data are then transmitted to the controller. After receiving the data, the controller can generate an adjustment command for the blade's angle of attack (i.e., angle of attack) based on the operating data and wind speed data, and then send the adjustment command to the wind turbine generator to achieve operation control of the generator.

[0061] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0062] Figure 2This is a flowchart illustrating an embodiment of the control method for a wind turbine generator provided in this application. Please refer to [link / reference]. Figure 2 The method includes:

[0063] S201. Obtain the operating data and wind speed data of the wind turbine generator set.

[0064] The execution entity in this application embodiment can be a controller or a control device for a wind turbine generator set installed in the controller. The control device for the wind turbine generator set can be implemented through software or a combination of software and hardware. The control device for the wind turbine generator set can be a processor within the controller. For ease of understanding, the technical solution of this application will be described below using a controller as an example.

[0065] In this step, the controller can acquire operating data and wind speed data of the wind turbine generator through sensors installed on the wind turbine. The wind speed data includes a sequence of wind speeds corresponding to multiple sampling points located at different distances in front of the wind turbine generator.

[0066] Operational data can be various parameters collected continuously over a period of time that reflect the current operating status, structural status, and working conditions of the wind turbine generator set. Specifically, operational data may include at least one of the following: generator speed, generator power, current blade angle of attack (pitch angle), nacelle temperature, nacelle wind speed, nacelle acceleration, etc.

[0067] Wind speed data refers to the wind speed information collected by lidar at sampling points at different distances in front of the wind turbine generator. Each sampling point corresponds to an independent wind speed sequence, which consists of multiple sets of wind speed data obtained by lidar from continuous collection of data at that sampling point. This sequence can accurately reflect the wind speed change pattern at that sampling point over a future period of time.

[0068] In one specific implementation, sampling points at different distances in front of the wind turbine generator can be preset according to the actual environment where the generator is deployed, the generator model, and the control accuracy requirements. For example, sampling points at different distances such as 50m, 100m, and 150m in front of the generator can be preset to flexibly adapt to different application scenarios.

[0069] S202. Using the pre-acquired load prediction model, the operating data and wind speed data are processed to obtain the predicted load data of the blades and the target wind arrival time.

[0070] In this step, the controller can synchronously input the acquired operational data and wind speed data into a pre-trained load prediction model. The load prediction model then processes the two types of data to obtain the predicted load data for the blades and the target wind exposure time. The predicted load corresponding to the target wind exposure time is greater than or equal to the preset load.

[0071] Specifically, the predicted load data includes the load magnitude and trend of the blade at multiple consecutive moments in the future, which is used to determine whether the blade is at risk of impact overload.

[0072] In one alternative implementation, the load prediction model can be trained based on a neural network. The neural network can be a network structure capable of processing time-series data, such as a Recurrent Neural Network (RNN), a Long Short-Term Memory (LSTM) network, a Transformer network, or a Convolutional Neural Network (CNN).

[0073] The preset load can be a critical load value pre-set based on parameters such as the wind turbine model, blade material, structural strength, and design specifications. It is used to define the range of safe blade operation. When the predicted load corresponding to the target wind arrival time is greater than or equal to the preset load, it indicates that the blade will face the risk of overload at that time, and the windward angle needs to be adjusted in advance to ensure structural safety.

[0074] It should be noted that in scenarios such as gusts, turbulence, or extreme wind conditions, wind speed can change drastically in a short period of time, which can easily lead to sudden changes in blade load. By determining the moment when the predicted load is greater than or equal to the preset load as the target wind arrival time, overload risks can be identified before load changes occur, providing a reliable basis for advance adjustment of the blade's windward angle.

[0075] In one alternative implementation, the target wind arrival time can also be the moment when the load value in the predicted load data reaches its maximum value, that is, the critical moment when the blade is under the greatest stress and the structure is most dangerous in the future. Using this as a time reference for adjusting the blade's windward angle can ensure the blade's operational safety to the greatest extent.

[0076] S203. Determine the target windward angle of the blades based on the predicted load data.

[0077] In this step, the target windward angle of the blade that matches the predicted load can be determined based on the predicted load corresponding to the target wind exposure time. Specifically, the target load range where the predicted load at the target wind exposure time falls can be determined, and then the target windward angle corresponding to that target load range can be determined based on a preset correspondence between load ranges and windward angles. The preset correspondence between load ranges and windward angles can include multiple load ranges and the corresponding windward angle for each load range.

[0078] In one specific implementation, different load ranges correspond to different windward angles, and this correspondence can be a pre-defined mapping rule. The control logic is as follows: the larger the predicted load, the smaller the target windward angle, causing the blades to tend to feather, thereby reducing wind capture capacity and reducing impact load; the smaller the predicted load, the larger the target windward angle, in order to maintain a higher wind energy utilization rate.

[0079] In this embodiment, the aforementioned preset load range and windward angle correspondence can be established according to different operating conditions of the wind turbine generator set. Operating conditions may include, but are not limited to: different turbulence intensity levels, different generator speed ranges, different power generation ranges, different air densities or wind shear strengths, and wind resource characteristics corresponding to wind turbine generator sets with similar installation locations.

[0080] For each operating condition, the applicable load range and windward angle mapping relationship can be calibrated through simulation calculations or field tests. For example, under low wind speed conditions, in order to balance power generation efficiency, the load range can be divided into finer intervals, and the windward angle can be appropriately reduced only when the load increases significantly. Under high turbulence or frequent gusts, a more conservative mapping relationship can be set, that is, a small load increment will trigger a larger angle of blade retraction to prioritize blade safety.

[0081] Optionally, the above-mentioned correspondence between the preset load range and the windward angle does not limit whether to distinguish between working conditions. That is, the preset relationship can be a single mapping table that is uniformly applicable to all working conditions, or it can be multiple sets of mapping tables preset for different working conditions.

[0082] S204. Before the target wind arrival time, adjust the current windward angle of the blades to the target windward angle to control the operation of the wind turbine generator set.

[0083] In this step, the controller can adjust the current windward angle of the blades to the target windward angle before the target wind arrival time, based on the determined target windward angle, and control the blades to operate at the target windward angle, thereby achieving stable control of the wind turbine generator.

[0084] In one specific implementation, the controller can send the target windward angle to the angle actuator of the wind turbine generator, which then adjusts the blade attitude based on the target windward angle. The angle actuator can be an electric actuator, a hydraulic actuator, or a servo drive mechanism, capable of receiving control commands from the controller and stably driving the blades to adjust and position the windward angle, ensuring the response speed and control accuracy of the blade attitude adjustment.

[0085] During the adjustment process, the controller can acquire the actual windward angle of the blade in real time and compare the actual windward angle with the target windward angle. Based on the comparison results, the controller performs closed-loop correction of the adjustment process. If there is a deviation between the actual windward angle and the target windward angle, the controller can adjust the control command in time to ensure that the blade can accurately reach and maintain the target windward angle, and avoid the blade from bearing additional load due to adjustment deviation.

[0086] Optionally, after adjusting the windward angle, the controller can continuously monitor the unit's operating status and make real-time corrections to the blade's windward angle based on subsequently updated predicted load data, further improving control accuracy and operational reliability.

[0087] In another embodiment of this application, the load prediction model can be used only to output predicted load data, while the target wind arrival time is determined by the wind field evolution and propagation model. Specifically, a dynamic time warping algorithm can be used to time-align the wind speed sequences of multiple sampling points ahead to obtain the propagation time of gusts at different cross-sections. Based on the different nacelle position sectors and wind speed information, a wind field evolution and propagation model is constructed, including the nacelle position sector, wind speed, and wind field propagation time at each cross-section. Based on this model, the target wind arrival time, i.e., the time when the predicted load reaches the preset load, is determined according to the current wind speed and nacelle position sector. The predicted load data output by the load prediction model is used to determine the target windward angle, and the target wind arrival time output by the wind field evolution and propagation model is used to determine the timing of pitch control, thereby achieving advance pitch control.

[0088] In this embodiment, the controller acquires the operating data and wind speed data of the wind turbine generator set, and then uses a pre-acquired load prediction model to predict the operating data and wind speed data to obtain the predicted load data and target wind arrival time of the wind turbine generator set blades. Subsequently, based on the predicted load data, the target windward angle of the blades is determined. Finally, before the target wind arrival time, the current windward angle of the blades is adjusted to the target windward angle corresponding to the target wind arrival time, thereby controlling the operation of the wind turbine generator set. The wind speed data includes a sequence of wind speeds corresponding to multiple sampling points at different distances in front of the wind turbine generator set, and the predicted load corresponding to the target wind arrival time is greater than or equal to a preset load. In the above process, the controller predicts the load in advance and adjusts the target windward angle of the blades ahead of time through the load prediction model, which can maximize wind energy utilization and effectively increase the power generation of the wind turbine generator set while ensuring the safe operation of the unit.

[0089] exist Figure 2 Based on the illustrated embodiment, the following, in conjunction with Figure 3 The control methods for the aforementioned wind turbine generator sets will be further explained in detail.

[0090] Figure 3 This is a flowchart illustrating a second embodiment of the control method for the wind turbine generator provided in this application. Please refer to [link / reference]. Figure 3 The method may include:

[0091] S301. Obtain the operating data and wind speed data of the wind turbine generator set. The wind speed data includes the wind speed sequence corresponding to multiple sampling points located at different distances in front of the wind turbine generator set.

[0092] For example, the controller can simultaneously collect operating data and wind speed data through the speed sensor, temperature sensor and lidar on the unit; among them, the operating data are the unit speed (range R1-R2 revolutions per minute) and the nacelle temperature (T1-T2℃) collected continuously for 30 seconds, with a collection frequency of once per second, and a total of 30 sets of data are collected within 30 seconds to form a continuous operating data time sequence.

[0093] The wind speed data consists of wind speed sequences corresponding to three preset sampling points at 50m, 100m, and 150m in front of the unit. Each sampling point is collected synchronously with the operational data, once per second, forming 30 sets of wind speed data within 30 seconds. For example, the wind speed sequence at the 50m sampling point can be: V1m / s, ..., V30m / s (30 sets collected consecutively).

[0094] S302. Using the pre-acquired load prediction model, the operating data and wind speed data are processed to obtain the predicted load data of the blades and the target wind arrival time.

[0095] For example, the controller can simultaneously input the collected 30 sets of operational data time series, as well as the wind speed sequences corresponding to the three sampling points of 50m, 100m, and 150m, into a pre-trained load prediction model. The load prediction model then performs fusion prediction processing on the above multi-source time series data and outputs the predicted load data of the blades for the next 30 seconds.

[0096] Furthermore, the controller can determine the time when the predicted load is greater than or equal to the preset load as the target wind arrival time. For example, the preset load can be set to F0. If the predicted load F1 corresponding to the 20th second in the future is greater than the preset load F0, then that time can be recorded as the target wind arrival time.

[0097] S303. Determine the target load range corresponding to the predicted load at the moment when the target is exposed to wind.

[0098] In this step, the controller can match the predicted load corresponding to the target wind exposure time with each load interval based on multiple pre-divided and stored load intervals to determine the target load interval where the predicted load is located.

[0099] For example, the load can be divided into low load, medium load, and high load ranges based on the wind turbine model, blade structural strength, and design safety threshold. The low load range is [F1, F2], the medium load range is [F2, F3], and the high load range is [F3, F4]. The controller can quickly locate the corresponding target load range by comparing the predicted load with the upper and lower limits of each range.

[0100] S304. Based on the preset correspondence between load range and windward angle, determine the target windward angle corresponding to the target load range.

[0101] In this step, the controller can obtain the target windward angle corresponding to the target load range by looking up a table based on the pre-calibrated mapping relationship between the load range and the windward angle, according to the determined target load range.

[0102] For example, the low load range [F1, F2) corresponds to the windward angle A1, the medium load range [F2, F3) corresponds to the windward angle A2, and the high load range [F3, F4] corresponds to the windward angle A3, where A1 is greater than A2 and A2 is greater than A3. The controller can determine the target windward angle corresponding to the target load range by looking up the table.

[0103] S305: Based on wind speed data, the expected wind exposure time of the blades is determined through a dynamic time warping algorithm.

[0104] In this step, the controller can use the wind speed sequence of multiple sampling points ahead to calculate the wind event propagation speed based on the dynamic time warping algorithm, and then estimate the expected wind arrival time of the wind field to the blades.

[0105] In its implementation, the controller uses a dynamic time warping algorithm to time-align the wind speed sequences corresponding to multiple sampling points in the wind speed data, obtaining the propagation time difference of wind events between adjacent sampling points. Based on the propagation time difference between adjacent sampling points and the corresponding sampling point spacing, multiple propagation velocities are calculated. These multiple propagation velocities are then fused to obtain the target propagation velocity. Finally, based on the target propagation velocity and the distance between the reference sampling point and the blade, the expected wind arrival time is calculated. The distance between the reference sampling point and the blade is less than the distance between any other sampling point and the blade.

[0106] Optionally, when fusing multiple propagation velocities, methods such as averaging, weighted averaging, or median can be used. In this embodiment, the average value is used as the target propagation velocity. The predicted wind arrival time calculated based on the target propagation velocity and the distance between the reference sampling point and the blade refers to the time required for the wind event to propagate to the blade from the current moment.

[0107] For example, three sampling points are set in front of the wind turbine, at distances of 50m, 100m, and 150m from the blade, respectively. The lidar simultaneously collects wind speed sequences at each point. A dynamic time warping algorithm identifies the times when the same wind speed peak occurs at the 150m, 10m, and 50m sampling points, respectively, 15s, 10s, and 5s before the current moment. The propagation time difference between adjacent sampling points is then calculated to be 5s. Since the distance between adjacent sampling points is 50m, multiple propagation velocities of 10m / s can be calculated. The average of these multiple propagation velocities yields the target propagation velocity of 10m / s. The sampling point 50m closest to the blade is selected as the baseline sampling point. Based on the target propagation velocity, the expected wind arrival time is calculated to be 5s, meaning the wind will reach the blade 5 seconds after the current moment.

[0108] It should be noted that the predicted wind arrival time and the target wind arrival time output by the load prediction model describe the same wind event. The wind speed peak (or rising edge) identified by the dynamic time warping algorithm is the same wind process that causes the blade load to increase and may trigger the preset load threshold. The predicted wind arrival time represents the time when the wind event reaches the blade, while the target wind arrival time represents the time when the blade load first reaches or exceeds the preset load after the wind event.

[0109] S306. If the difference between the target wind arrival time and the expected wind arrival time is less than a preset threshold, then at a time that is a preset time earlier than the target wind arrival time, the current windward angle of the blades will be adjusted to the target windward angle to control the operation of the wind turbine generator set.

[0110] In this step, the controller calculates the difference between the target wind arrival time and the expected wind arrival time, and compares this difference with a preset threshold. If the difference is less than the preset threshold, it indicates that the target wind arrival time output by the load prediction model is basically consistent with the expected wind arrival time calculated based on wind speed data, and the prediction result can be considered reliable. At this time, the controller can determine a time that is a preset amount of time earlier than the target wind arrival time as the adjustment start time, and at this start time, adjust the current windward angle of the blades to the target windward angle to control the operation of the wind turbine generator.

[0111] For example, the preset threshold can be set to 2 seconds, and the preset duration can be set to 1 second. If the target wind arrival time is 10.5 seconds and the expected wind arrival time is 10.0 seconds, the difference is 0.5 seconds, which is less than the preset threshold of 2 seconds. In this case, the controller determines that the prediction result is reliable, and can start adjusting the current windward angle of the blades to the target windward angle 1 second earlier than the target wind arrival time (i.e., 9.5 seconds), ensuring that the adjustment is completed before 10.5 seconds, thereby effectively reducing the impact of wind load.

[0112] Optionally, if the difference between the target wind arrival time and the expected wind arrival time is greater than or equal to a preset threshold, it indicates that there is a large deviation between the output of the load prediction model and the physical calculation result. At this time, the controller can determine that the prediction is unreliable, temporarily refrain from pitch adjustment, and issue a warning signal or re-collect and predict data to avoid unnecessary pitch or safety risks caused by model misreporting.

[0113] In this embodiment, the controller first acquires the operating data of the wind turbine generator and wind speed data including a sequence of wind speeds from multiple sampling points ahead. Then, it processes this data using a load prediction model to obtain the predicted load data for the blades and the target wind arrival time. Next, it determines the target load interval corresponding to the target wind arrival time and, based on a preset mapping relationship between the load interval and the windward angle, obtains the corresponding target windward angle. Simultaneously, based on the wind speed data, a dynamic time warping algorithm can be used to determine the expected wind arrival time of the blades. Finally, when the difference between the target wind arrival time and the expected wind arrival time is less than a preset threshold, the current windward angle of the blades can be adjusted to the target windward angle at a time preset before the target wind arrival time, thereby controlling the wind turbine generator. Through this method, the controller can predict the load in advance and verify it in conjunction with the wind arrival time. After meeting the condition of being less than the preset threshold, it can adjust the windward angle of the blades ahead of time, thereby ensuring the safe operation of the unit while increasing the power generation of the wind turbine generator.

[0114] Figure 4 This is a flowchart illustrating a third embodiment of the control method for the wind turbine generator provided in this application. Please refer to [link / reference]. Figure 4Based on any of the above embodiments, before performing predictive processing on the operating data and wind speed data using a pre-acquired load prediction model to obtain the predicted load data of the blades and the target wind arrival time, the method further includes:

[0115] S401. Obtain multiple training samples.

[0116] In this step, the controller can acquire multiple training samples. Each training sample includes historical operating data of the wind turbine generator, historical wind speed data, corresponding historical load data, and historical wind arrival time. The load corresponding to the historical wind arrival time is greater than or equal to the preset load.

[0117] In one specific implementation, for any one of a plurality of training samples, the following steps can be taken: Obtain.

[0118] step Based on a preset sliding time window, the original load data is obtained from the wind turbine generator's status database.

[0119] Specifically, the controller can read the raw load sequence within a preset length (e.g., 30 seconds) sliding time window from the unit status database. Here, the raw load sequence refers to the unprocessed load value sequence collected by the load sensor.

[0120] To extract information reflecting load fluctuation characteristics from the original load sequence, the controller can perform signal processing on the original load sequence to obtain a load data envelope. For example, a peak extraction method can be used to identify each peak point in the original load sequence and connect these peak points sequentially to form a smooth contour curve, i.e., the load data envelope. This envelope curve reflects the overall trend of load variation and removes high-frequency noise.

[0121] Then, the controller can extract the instantaneous load fluctuation time series of different gradients from the load data envelope. The gradient represents the drastic degree of load change. For example, the envelope curve can be divided into low gradient (gradual change), medium gradient (moderate change), and high gradient (rapid change) according to the slope, and the corresponding instantaneous load fluctuation time series can be extracted respectively. These instantaneous load fluctuation time series of different gradients are used as the original load data.

[0122] step Based on the turbulence freezing theorem, the original wind speed data and original operating data corresponding to the original load data are determined.

[0123] Specifically, the controller can simultaneously collect raw wind speed data from multiple sampling points at different distances in front of the wind turbine generator using lidar. Based on the turbulence freeze theorem, which assumes that the turbulent structure of the wind field maintains its shape as it propagates downstream, the raw load data obtained in step ① is correlated and matched with the raw wind speed sequence of each sampling point on the time axis.

[0124] To quantitatively determine the matching relationship between the two, the controller can apply a series of candidate time offsets to the wind speed sequence at each sampling point (the offset range is estimated based on the distance between the sampling point and the blade and typical wind speed), and calculate the temporal similarity between the offset wind speed sequence and the original load data, for example, using a cross-correlation function or dynamic time warping distance as a metric. The time offset that maximizes the similarity and the corresponding wind speed segment are selected as the original wind speed data with the strongest causal correlation to the original load data. This time offset also characterizes the temporal alignment relationship between load events and wind speed events.

[0125] Furthermore, the controller can retrieve the original operating data corresponding to the original wind speed data in time from the unit status database based on the obtained time offset. That is, the time window of the original operating data is consistent with the time window of the matched original wind speed data. Through the above matching process, a one-to-one correspondence can be established between load events, specific wind events ahead, and synchronous operating states, ensuring that the input features and output labels in the training samples are completely aligned in terms of time axis and physical causality.

[0126] step The original load data, original wind speed data, and original operational data are expanded to obtain historical load data, historical operational data, and historical wind speed data within the historical time range.

[0127] The extended time ranges for each data point include the time intervals covered by the sliding time window.

[0128] Specifically, the controller can perform independent time axis expansion operations on the original load data, as well as the matched original wind speed data and original operational data, to make the time range of each data sequence larger than the original sliding time window. This provides the neural network with more comprehensive historical context information, enhancing the model's ability to learn long-term dependencies on load evolution trends. For example, if the original sliding time window is 30 seconds, it can be extended forward and backward by several seconds, making the total time length 35 seconds. The expanded time interval completely encompasses the original 30-second window.

[0129] Optionally, the controller can employ differentiated expansion strategies to maintain physical rationality for different data types. For raw load data, spline interpolation or autoregressive models can be used to extrapolate based on its fluctuation trend, ensuring that the load changes in the expanded portion conform to the energy decay law. For raw wind speed data, the turbulence freeze theorem can be used to calculate the propagation speed based on the temporal relationship of the wind speed sequences at each sampling point, thereby converting the wind speed sequences of discrete sampling points into a continuous time sequence and extrapolating them forward and backward, ensuring that the expanded wind speed sequence physically conforms to the propagation characteristics of the wind field. For raw operating data, zero-order hold (i.e., the expanded portion retains the values ​​at the window boundaries) or linear interpolation methods can be used to maintain the continuity of the operating state.

[0130] step The moment when the load in the historical load data first exceeds or equals the preset load is taken as the historical wind arrival moment.

[0131] Specifically, the controller can scan the expanded historical load data sequence point by point in ascending order of time, comparing the load value of each sampling point with the preset load. When the load value is first detected to be equal to or exceed the preset load, the time coordinate of that moment is immediately recorded and identified as the historical wind exposure moment. This historical wind exposure moment is one of the key supervision labels in the training samples, corresponding to the target wind exposure moment that the load prediction model needs to output in the prediction phase.

[0132] Optionally, if the load value consistently falls below a preset load throughout the entire historical timeframe, it indicates that the sample does not contain valid wind-affected time labels. In this case, the controller can discard the sample or mark it as a negative sample used only for load numerical learning, preventing invalid samples from interfering with the model's learning of overload moments. By setting a preset load threshold, it can be ensured that the training set focuses on typical events with overload risk, improving the model's ability to identify and predict hazardous scenarios.

[0133] step Training samples are constructed based on historical load data, historical operation data, historical wind speed data, and historical wind exposure times.

[0134] After completing the above expansion and determining the wind exposure time, the controller can integrate the four parts of data obtained into a complete training sample. Specifically, historical operating data and historical wind speed data can be used as input features of the model, together forming a multi-source temporal input tensor; historical load data and historical wind exposure time can be used as output labels of the model, i.e., the target values ​​for supervised training. Among them, historical load data can be a complete load sequence or a temporal representation of its gradient features, and historical wind exposure time is a scalar value, representing the time point when the load first exceeds the preset load (if it does not exist, it is set to a null value or a special marker).

[0135] Each training sample can be represented as a quadruple: input features (historical operating data, historical wind speed data) and output labels (historical load data, historical wind arrival time). The input features reflect the operating status of the wind turbine generator within the sliding time window and the changing patterns of the wind field ahead, while the output labels record the actual load response and the moment of its first exceedance within that time period. In this way, the input and output in the training samples are aligned both temporally and physically, providing a reliable data foundation for subsequent supervised learning.

[0136] Specifically, the controller repeatedly executes steps ① to ⑤ above, using different starting positions of the sliding time window (e.g., sliding one step at a time), to extract multiple data segments from the unit status database, ultimately forming a dataset containing a large number of training samples. This dataset covers historical load events under various wind conditions and operating conditions, possessing sufficient diversity and representativeness.

[0137] S402. Based on multiple training samples, train the initial load prediction model constructed based on the neural network to obtain the load prediction model.

[0138] In this step, the controller can iteratively train the initial load prediction model based on the neural network using multiple training samples, gradually optimizing the model parameters so that the model can accurately learn the complex nonlinear mapping relationship between the input data (historical operation data and historical wind speed data) and the output labels (historical load data and historical wind arrival time).

[0139] Specifically, the controller feeds the input features (historical operating data and historical wind speed data) from the training sample set into the initial neural network model in batches. The model calculates the predicted load data and predicted wind arrival time through forward propagation. The controller adopts a multi-task learning framework to simultaneously optimize both load prediction and time prediction tasks. The loss function can be designed as a weighted sum of the load prediction error and the wind arrival time prediction error, as shown in formula (1):

[0140] Formula (1)

[0141] in, Let be the predicted load sequence of the model for the i-th sample. Let be the predicted wind arrival time for the i-th sample by the model. For the i-th sample, the true historical payload sequence is... Let N be the actual historical wind exposure time corresponding to the i-th sample, and N be the batch size. and Weighting coefficients (e.g.) It can take the value 0.7. (Possible value: 0.3)

[0142] In an alternative implementation, the controller can use mini-batch gradient descent for parameter optimization. During each iteration, a batch of samples is randomly selected from the training sample set; the batch size m can be preset to 32 or 64. The input features (historical operating data and historical wind speed data) from the current batch are fed into the neural network, and the predicted load sequence and predicted wind arrival time are obtained through forward propagation. Then, the loss value L is calculated according to formula (1), and the gradient of the loss function with respect to each learnable parameter in the network is calculated using the backpropagation algorithm.

[0143] In another alternative implementation, the controller can employ the Adam optimizer to update the network parameters. The Adam optimizer adaptively adjusts the learning rate based on the first and second moments of the gradient, making the training process smoother. Typically, the initial learning rate is set to 0.001, and the learning rate is multiplied by 0.9 every 10 epochs during training, gradually reducing the update step size to help the model converge more finely.

[0144] Furthermore, to prevent overfitting, the controller can allocate approximately 20% of the training samples as a validation set. This data is not used for parameter updates and is only used to evaluate the model's generalization ability. After each training epoch, the controller calculates the loss value on the validation set. If the validation loss does not decrease for 10 consecutive epochs, an early stopping mechanism is triggered, immediately halting training and rolling back to the model parameters at the point of lowest validation loss, thus preventing the model from overlearning noise in the training data.

[0145] Optionally, the controller can also introduce Dropout layers between the fully connected layers of the neural network, randomly dropping some neurons during training with a target probability (e.g., 0.2), forcing the network to learn more robust feature representations and reducing dependence on individual features.

[0146] In this embodiment, the controller can train an initial load prediction model based on a neural network using multiple training samples to obtain the load prediction model. By fully learning and fitting historical operating data, historical wind speed data, and corresponding historical load data and historical wind arrival times, the trained load prediction model can accurately reflect the intrinsic relationship between wind field conditions, turbine operating status, and blade load, possessing strong temporal feature extraction and generalization capabilities. In subsequent actual control processes, this load prediction model can directly and quickly predict the real-time collected operating data and wind speed data, outputting stable and reliable predicted load data and target wind arrival times, providing precise data support for the advanced control of wind turbine generators, thereby improving the safety, stability, and power generation efficiency of the turbine operation.

[0147] Figure 5This is a flowchart illustrating an example of a control method for a wind turbine generator provided in this application. Please refer to... Figure 5 The process includes:

[0148] S501. Based on the historical operating data, historical wind speed data, and corresponding historical load data and historical wind arrival times of the wind turbine generator, multiple training samples are constructed.

[0149] Specifically, the historical operational data, historical wind speed data, and historical load data that have undergone data preprocessing, as well as the matched historical load data and the time when the preset load was first reached (i.e. the historical wind arrival time), can be linked and integrated to form multiple sets of training samples with complete structure and corresponding time sequence.

[0150] S502. Based on multiple training samples, train the initial load prediction model constructed based on the neural network to obtain the load prediction model.

[0151] Specifically, historical operating data and historical wind speed data from the training samples can be used as model inputs, and corresponding historical load data and historical wind exposure times can be used as model output labels. The initial load prediction model can be trained iteratively through multiple tasks to gradually optimize the model parameters, so as to obtain a load prediction model that can simultaneously output predicted load data and target wind exposure times.

[0152] S503. Using the trained load prediction model, predictive processing is performed on the operating data and wind speed data to obtain the predicted load data of the blades and the target wind arrival time.

[0153] Specifically, real-time collected unit operation data and wind speed data from multiple sampling points can be input into the trained load prediction model. The model will then output the predicted load data for the blades in the future, as well as the target wind time when the predicted load first reaches or exceeds the preset load.

[0154] S504. Adjust the blades of the wind turbine generator before the target wind arrival time.

[0155] Specifically, the appropriate blade adjustment parameters (i.e., the target windward angle) can be determined based on the predicted load corresponding to the target windfall time. Before the target windfall time arrives, the current windward angle of the blades can be adjusted to the target windward angle to complete the advance adjustment of the blade attitude and control the operation of the wind turbine generator.

[0156] The wind turbine control method provided in this application first constructs multiple training samples based on the historical operating data, historical wind speed data, corresponding historical load data, and historical wind arrival times of the wind turbine. Then, based on these training samples, an initial load prediction model constructed using a neural network is trained to obtain a load prediction model. Subsequently, the trained load prediction model is used to predict the operating data and wind speed data to obtain the predicted load data for the blades and the target wind arrival time. Finally, before the target wind arrival time, the blades of the wind turbine are adjusted to control the operation of the wind turbine. In the above process, by constructing training samples containing times when the load exceeds the limit and training the load prediction model, accurate prediction of blade load and its exceeding time can be achieved. Combined with advance adjustment of the blade attitude at the target wind arrival time, the wind turbine's windward state can be optimized while ensuring safe and stable operation, maximizing wind energy utilization and effectively improving the unit's power generation and operational economy.

[0157] Figure 6 A schematic diagram of the control device for a wind turbine generator set provided in this embodiment of the application. Please refer to... Figure 6 The control device 10 for the wind turbine generator set includes:

[0158] The acquisition module 11 is used to acquire the operating data and wind speed data of the wind turbine generator set. The wind speed data includes the wind speed sequence corresponding to multiple sampling points located at different distances in front of the wind turbine generator set.

[0159] The first processing module 12 is used to perform predictive processing on the operating data and wind speed data through a pre-acquired load prediction model to obtain the predicted load data of the blade and the target wind exposure time. The predicted load corresponding to the target wind exposure time is greater than or equal to the preset load.

[0160] The second processing module 13 is used to determine the target windward angle of the blade based on the predicted load data;

[0161] The control module 14 is used to adjust the current windward angle of the blades to the target windward angle before the target wind arrival time, so as to control the operation of the wind turbine generator set.

[0162] The control device for the wind turbine generator provided in this application embodiment can execute the technical solution shown in the above method embodiment. Its implementation principle and beneficial effects are similar, and will not be described again here.

[0163] In one possible implementation, the second processing module 13 is specifically used for:

[0164] Determine the target load range corresponding to the predicted load at the moment when the target is exposed to wind;

[0165] Based on the preset correspondence between load range and windward angle, determine the target windward angle corresponding to the target load range.

[0166] In one possible implementation, the control module 14 is specifically used for:

[0167] Based on wind speed data, the expected wind exposure time of the blades is determined through a dynamic time warping algorithm;

[0168] If the difference between the target wind arrival time and the expected wind arrival time is less than a preset threshold, then at a time that is a preset time earlier than the target wind arrival time, the current windward angle of the blades will be adjusted to the target windward angle.

[0169] In one possible implementation, the control module 14 is specifically used for:

[0170] By using a dynamic time warping algorithm, the wind speed sequences corresponding to multiple sampling points in the wind speed data are time-aligned to obtain the propagation time difference of wind events between adjacent sampling points;

[0171] Calculate multiple propagation velocities based on the propagation time difference between adjacent sampling points and the corresponding sampling point spacing;

[0172] The target propagation speed is obtained by fusing multiple propagation speeds.

[0173] Based on the target propagation speed and the distance between the reference sampling point and the blade, the expected wind arrival time is calculated. The distance between the reference sampling point and the blade is less than the distance between any other sampling point and the blade.

[0174] In one possible implementation, the first processing module 12 is further configured to:

[0175] Multiple training samples are obtained. Each training sample includes historical operating data of the wind turbine generator, historical wind speed data, corresponding historical load data, and historical wind exposure time. The load corresponding to the historical wind exposure time is greater than or equal to the preset load.

[0176] The initial load prediction model based on a neural network is trained using multiple training samples to obtain the load prediction model.

[0177] In one possible implementation, the first processing module 12 is specifically used for:

[0178] Based on a sliding time window of preset length, the original load data is obtained from the wind turbine generator's unit status database;

[0179] Based on the turbulence freezing theorem, the original wind speed data and original operational data corresponding to the original load data are determined.

[0180] The original load data, original wind speed data, and original operational data are expanded to obtain historical load data, historical operational data, and historical wind speed data within the historical time range. The time range of each expanded data includes the time interval covered by the sliding time window.

[0181] The moment when the load in the historical load data first exceeds or equals the preset load is taken as the historical wind arrival moment;

[0182] Training samples are constructed based on historical load data, historical operation data, historical wind speed data, and historical wind exposure times.

[0183] In one possible implementation, the first processing module 12 is specifically used for:

[0184] Obtain the original load sequence within the sliding time window from the unit status database;

[0185] Signal processing is performed on the original payload sequence to obtain the payload data envelopment;

[0186] The instantaneous load fluctuation time series with different gradients are extracted from the load data envelopment, and the instantaneous load fluctuation time series are determined as the original load data.

[0187] The control device for the wind turbine generator provided in this application embodiment can execute the technical solution shown in the above method embodiment. Its implementation principle and beneficial effects are similar, and will not be described again here.

[0188] Figure 7 This is a schematic diagram of the controller provided in an embodiment of this application. Please refer to... Figure 7 The controller 20 provided in this embodiment can be a main controller installed in the nacelle of the wind turbine generator set, or an embedded controller integrated into the wind turbine generator set monitoring system, or a dedicated control device independently deployed next to the generator set and communicating with each sensor and actuator of the generator set.

[0189] The controller 20 may include at least one processor 21 and a memory 22. Optionally, the controller 20 may also include a communication component 24. The processor 21, the memory 22, and the communication component 24 are connected via a bus 23.

[0190] In the specific implementation process, at least one processor 21 executes computer execution instructions stored in memory 22, causing at least one processor 21 to perform the above-described method.

[0191] The specific implementation process of processor 21 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0192] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0193] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0194] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0195] This application also provides a wind turbine generator set, including as follows: Figure 7 The controller shown is further shown. The wind turbine also includes a variety of sensors deployed at key locations of the unit, including speed sensors, temperature sensors, and vibration sensors for collecting operating data such as unit speed, power, temperature, and vibration; lidar for collecting wind speed sequences at multiple sampling points at different distances in front of the unit; and load sensors deployed on the blades for collecting blade load data.

[0196] The wind turbine also includes an actuator for responding to controller commands and adjusting the blade angle. Both the aforementioned sensors and actuators are communicatively connected to the controller, and through their coordinated operation, the wind turbine control method described in the above embodiment can be implemented.

[0197] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0198] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0199] The aforementioned readable storage medium 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 readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0200] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0201] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0202] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0203] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0204] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0205] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0206] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A control method for a wind turbine generator set, characterized in that, The wind turbine generator set includes multiple blades, and the method includes: The wind turbine generator set's operating data and wind speed data are acquired, wherein the wind speed data includes wind speed sequences corresponding to multiple sampling points located at different distances in front of the wind turbine generator set; By using a pre-acquired load prediction model, the operating data and the wind speed data are processed to predict the load data of the blade and the target wind exposure time. The predicted load corresponding to the target wind exposure time is greater than or equal to the preset load. Based on the predicted load data, the target windward angle of the blade is determined; Before the target wind arrival time, the current windward angle of the blades is adjusted to the target windward angle to control the operation of the wind turbine generator set.

2. The method according to claim 1, characterized in that, Determining the target windward angle of the blade based on the predicted load data includes: Determine the target load range where the predicted load is located at the time when the target is exposed to wind; Based on the preset correspondence between load range and windward angle, the target windward angle corresponding to the target load range is determined.

3. The method according to claim 1 or 2, characterized in that, The step of adjusting the current windward angle of the blade to the target windward angle before the target wind arrival time includes: Based on the wind speed data, the expected wind exposure time of the blades is determined using a dynamic time warping algorithm; If the difference between the target wind arrival time and the expected wind arrival time is less than a preset threshold, then at a time that is a preset time earlier than the target wind arrival time, the current windward angle of the blade is adjusted to the target windward angle.

4. The method according to claim 3, characterized in that, The step of determining the expected wind exposure time of the blades based on the wind speed data using a dynamic time warping algorithm includes: The dynamic time warping algorithm is used to align the wind speed sequences corresponding to multiple sampling points in the wind speed data in time, so as to obtain the propagation time difference of wind events between adjacent sampling points. Calculate multiple propagation velocities based on the propagation time difference between adjacent sampling points and the corresponding sampling point spacing; The multiple propagation speeds are fused together to obtain the target propagation speed; The expected wind arrival time is calculated based on the target propagation speed and the distance between the reference sampling point and the blade, wherein the distance between the reference sampling point and the blade is less than the distance between any other sampling point and the blade.

5. The method according to claim 1 or 2, characterized in that, Before obtaining the predicted load data of the blade and the target wind arrival time by performing prediction processing on the operating data and the wind speed data using a pre-acquired load prediction model, the method further includes: Multiple training samples are obtained. Each training sample includes the historical operating data, historical wind speed data, corresponding historical load data, and historical wind exposure time of the wind turbine generator set. The load corresponding to the historical wind exposure time is greater than or equal to the preset load. The initial load prediction model based on the neural network is trained using the multiple training samples to obtain the load prediction model.

6. The method according to claim 5, characterized in that, The acquisition of training samples includes: Based on a sliding time window of preset length, the original load data is obtained from the unit status database of the wind turbine generator set; Based on the turbulence freezing theorem, the original wind speed data and original operating data corresponding to the original load data are determined. The original load data, original wind speed data, and original operation data are expanded to obtain historical load data, historical operation data, and historical wind speed data within a historical time range. The time range of each expanded data includes the time interval covered by the sliding time window. The moment when the load in the historical load data first exceeds or equals the preset load is taken as the historical wind arrival moment; Training samples are constructed based on the historical load data, the historical operation data, the historical wind speed data, and the historical wind exposure times.

7. The method according to claim 6, characterized in that, The sliding time window based on a preset length obtains raw load data from the wind turbine generator's status database, including: Obtain the original load sequence within the sliding time window from the unit status database; The original payload sequence is processed to obtain the payload data envelopment; The instantaneous load fluctuation time series with different gradients are extracted from the load data envelopment, and the instantaneous load fluctuation time series are determined as the original load data.

8. A control device for a wind turbine generator set, characterized in that, The device includes: The acquisition module is used to acquire the operating data and wind speed data of the wind turbine generator set. The wind speed data includes wind speed sequences corresponding to multiple sampling points located at different distances in front of the wind turbine generator set. The first processing module is used to perform predictive processing on the operating data and the wind speed data through a pre-acquired load prediction model to obtain the predicted load data of the blade and the target wind exposure time, wherein the predicted load corresponding to the target wind exposure time is greater than or equal to the preset load. The second processing module is used to determine the target windward angle of the blade based on the predicted load data. The control module is used to adjust the current windward angle of the blades to the target windward angle before the target wind arrival time, so as to control the operation of the wind turbine generator set.

9. A controller, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-7.

10. A wind turbine generator set, characterized in that, Includes the controller as described in claim 9.

Images