Dynamic compensation method, system and medium based on wind turbine yaw error identification

Abstract

This invention discloses a dynamic compensation method, system, and medium based on wind turbine yaw error identification, relating to the field of wind power equipment control. The method includes: tracing back historical error compensation logs of a wind turbine cluster, performing yaw error compensation analysis under operating conditions, and constructing a multi-scale compensation benchmark information database; the lead wind turbine locally integrates real-time yaw error and operating conditions as operating condition feature vectors to query the multi-scale compensation benchmark information database, outputting a first dynamic compensation drive spectrum, and making compensation drive decisions for the lead wind turbine; distributing the first dynamic compensation drive spectrum to multiple following wind turbines; using the first dynamic compensation drive spectrum as a benchmark compensation drive parameter, and correcting and outputting multiple compensation drive decision results based on local wind turbine state snapshots to perform low-latency synchronous compensation. This solves the technical problems of existing yaw error compensation methods, such as inability to adapt to dynamic operating conditions in real time and insufficient compensation accuracy, achieving the technical effect of improving the accuracy and timeliness of yaw error compensation.

Description

Technical Field

[0001] This application relates to the field of wind power equipment control, and in particular to a dynamic compensation method, system and medium based on wind turbine yaw error identification. Background Technology

[0002] In the field of wind power generation, precise control of wind turbine yaw error is crucial for improving power generation efficiency, ensuring stable equipment operation, and extending service life. If yaw error is not compensated for in a timely and effective manner, the wind turbine will not be able to consistently capture wind energy at the optimal angle, resulting in power generation losses. It will also increase wear on critical components, reduce equipment reliability, and increase maintenance costs. Currently, the main method for solving the yaw error compensation problem is to adopt a compensation strategy based on a fixed parameter model. This strategy compensates for the yaw error of the wind turbine based on pre-set fixed parameters and an error calculation model. These fixed parameters are usually determined experimentally during the wind turbine design phase or under specific operating conditions and remain largely unchanged during subsequent operation. However, the actual operating conditions of wind turbines are complex and variable, influenced by factors such as wind speed, wind direction, turbulence intensity, and temperature. The compensation strategy based on the fixed parameter model cannot adapt to these dynamically changing conditions in real time, resulting in insufficient compensation accuracy.

[0003] At present, the yaw error compensation of wind turbines has technical problems such as being unable to adapt to dynamic operating conditions in real time and insufficient compensation accuracy. Summary of the Invention

[0004] This application provides a dynamic compensation method, system, and medium based on wind turbine yaw error identification. It employs methods such as backtracking historical error compensation logs of the wind turbine cluster, analyzing yaw error compensation under different operating conditions, constructing a multi-scale compensation benchmark information database, leading wind turbines to integrate real-time yaw error and operating conditions to form operating condition feature vectors, querying the database to output a first dynamic compensation driving spectrum, making its own compensation driving decisions, and simultaneously distributing the first dynamic compensation driving spectrum to multiple following wind turbines. These following wind turbines use this as a benchmark, combining local state snapshot corrections, outputting multiple compensation driving decision results, and performing low-latency synchronous compensation. These techniques solve the technical problems of existing wind turbine yaw error compensation methods, such as inability to adapt to dynamic operating conditions in real time and insufficient compensation accuracy. The application achieves the technical effect of adjusting the yaw error compensation strategy in real time and accurately according to the dynamic operating conditions of the wind turbines, improving the accuracy and timeliness of yaw error compensation.

[0005] This application provides a dynamic compensation method based on wind turbine yaw error identification, comprising: tracing back the historical error compensation logs of a wind turbine cluster, performing yaw error operating condition compensation analysis, and constructing a multi-scale compensation benchmark information database, wherein the wind turbine cluster includes a leader wind turbine and multiple follower wind turbines; in the wind turbine cluster, the leader wind turbine locally integrates real-time yaw error and real-time operating conditions as operating condition feature vectors to query the multi-scale compensation benchmark information database, outputs a first dynamic compensation driving spectrum, and makes compensation driving decisions for the leader wind turbine; the leader wind turbine distributes the first dynamic compensation driving spectrum to the multiple follower wind turbines; the multiple follower wind turbines use the first dynamic compensation driving spectrum as a benchmark compensation driving parameter, correct and output multiple compensation driving decision results based on local wind turbine state snapshots, and execute low-latency synchronous compensation for the multiple follower wind turbines.

[0006] In one possible implementation, the historical error compensation logs of the wind turbine cluster are traced back to perform yaw error condition compensation analysis, a multi-scale compensation benchmark information database is constructed, and the following processing is performed: Based on a preset field structure, multiple error compensation log fields of the leader wind turbine are retrieved from the historical error compensation logs. These error compensation log fields include sample compensation magnitude, yaw error angle before sample compensation, sample compensation condition feature vector, and sample compensation driving spectrum. The sample compensation driving spectrum includes sample power increment and sample load change, and the sample compensation condition feature vector includes sample compensation wind speed and sample compensation turbulence intensity. Based on the sample compensation magnitude... The fluctuation range is dynamically defined by the compensation scale classification rule; the multiple error compensation log fields are decomposed into multi-scale compensation knowledge units according to the compensation scale classification rule; multi-scale working condition binning modeling of the multi-scale compensation knowledge units is performed based on the preset field structure to construct multiple three-dimensional working condition binning units; multiple sets of follower compensation log fields of the multiple follower wind turbines are retrieved from the historical error compensation logs, and the multiple sets of follower compensation log fields are used to perform fuzzy correction of the compensation error of the multiple three-dimensional working condition binning units to obtain multiple benchmark compensation knowledge units; the multiple benchmark compensation knowledge units are associated and stored to construct the multi-scale compensation benchmark information database.

[0007] In a possible implementation, multi-scale load condition binning modeling of the multi-scale compensation knowledge unit is performed based on a preset field structure to construct multiple three-dimensional load condition binning units. The following processing is then performed: a first interval division rule for wind speed dimension, a second level division rule for turbulence intensity dimension, and a third interval division rule for yaw error angle dimension are set to construct a three-dimensional partition space; the first-scale compensation knowledge unit is discretized and mapped in the three-dimensional partition space to obtain multiple first binning data nodes, wherein the filling information of the first binning data nodes is a sample compensation driving spectrum; a spatial aggregation mechanism for nodes with the same coordinates is triggered to merge the multiple first binning data nodes with the same three-dimensional coordinates to construct a first primary binning unit; and a dynamic interval expansion operation is performed based on the data distribution sparsity of the first primary binning unit to generate a first three-dimensional load condition binning unit.

[0008] In a possible implementation, the following processing is also performed: driving the distributed wind turbines to collect yaw correlation data and outputting time-series slices of distributed yaw correlation data; dividing the distributed wind turbines into clusters based on the time-series slices of distributed yaw correlation data to obtain K wind turbine clusters; and performing collaborative dynamic compensation decision-making for yaw error of the wind turbines in the K wind turbine clusters.

[0009] In a possible implementation, the distributed wind turbine clusters are divided based on the distributed yaw correlation data time-series slices to obtain K wind turbine clusters. The following processing is then performed: A first yaw correlation data time-series slice is extracted from the distributed yaw correlation data time-series slices of the first individual wind turbine in the distributed wind turbine clusters. This first yaw correlation data time-series slice includes a yaw error angle time-series slice, a time-series wind speed slice, a time-series wind direction slice, and a turbulence intensity time-series slice. A first time-series feature vector is extracted based on the first yaw correlation data time-series slice. The first temporal feature vector is composed of a first dynamic morphological similarity, a first wind speed error covariance factor, and a first spatial weight feature. Similarly, M temporal feature vectors for M individual wind turbines in the distributed wind turbine system are constructed. An agglomerative hierarchical clustering algorithm is applied, using the Ward variance minimization criterion as the clustering basis, to perform clustering calculations on the M temporal feature vectors corresponding to the M individual wind turbines, generating multiple initial clusters. Predefined spatial distance constraints are used to traverse the multiple initial clusters, and constraint decoupling operations are performed on the long-distance clusters to output the K wind turbine clusters.

[0010] In a possible implementation, the following processing is also performed: extracting multiple yaw sensor errors, multiple spatial coordinates, and multiple average fault intervals from multiple individual wind turbines in the Kth wind turbine cluster; performing a three-weighted scoring assignment based on the multiple yaw sensor errors, multiple spatial coordinates, and multiple average fault intervals, and outputting multiple three-weighted score values; locating the Kth leader wind turbine and the Kth group of follower wind turbines among the multiple individual wind turbines according to the descending sorting result of the multiple three-weighted score values; and binding the Kth leader wind turbine and the Kth group of follower wind turbines based on communication dependency relationships.

[0011] In a possible implementation, within the wind turbine cluster, the leader wind turbine locally integrates real-time yaw error and real-time operating conditions as operating condition feature vectors to query the multi-scale compensation benchmark information database, outputs a first dynamic compensation driving spectrum, and makes compensation driving decisions for the leader wind turbine, performing the following processing: locating the target operating condition sub-unit in the multi-scale compensation benchmark information database by adopting the real-time yaw error traversal compensation scale classification rules; mapping the real-time operating conditions and real-time yaw error to the target operating condition sub-unit to generate real-time virtual particle points; retrieving N sub-unit data nodes from the target operating condition sub-unit that satisfy a preset association distance threshold with the real-time virtual particle points; and performing spectral fusion of the N sample compensation driving spectra among the N sub-unit data nodes based on the N spatial association distances between the N sub-unit data nodes and the real-time virtual particle points, outputting the first dynamic compensation driving spectrum.

[0012] In a possible implementation, the plurality of following wind turbines use the first dynamic compensation drive spectrum as a reference for compensation drive parameters, and perform low-latency synchronous compensation of the plurality of following wind turbines by correcting and outputting multiple compensation drive decision results based on local wind turbine state snapshots. The following processes are performed: retrieve the W-th following yaw error from the local wind turbine state snapshot of the W-th following wind turbine; calculate the W-th three-dimensional state offset between the real-time yaw error and the W-th following yaw error; correct the first dynamic compensation drive spectrum based on the W-th three-dimensional state offset and output the W-th compensation drive decision result; perform load safety boundary verification on the W-th compensation drive decision result, and then use the W-th compensation drive decision result to perform low-latency synchronous compensation of the W-th following wind turbine.

[0013] This application also provides a dynamic compensation system based on wind turbine yaw error identification, comprising: a multi-scale compensation benchmark information database construction module, used to backtrack the historical error compensation logs of the wind turbine cluster, perform yaw error operating condition compensation analysis, and construct a multi-scale compensation benchmark information database, wherein the wind turbine cluster includes a leader wind turbine and multiple follower wind turbines; a first dynamic compensation drive spectrum output module, used in the wind turbine cluster, whereby the leader wind turbine locally integrates real-time yaw error and real-time operating condition as operating condition feature vectors to query the multi-scale compensation benchmark information database, outputs a first dynamic compensation drive spectrum, and makes compensation drive decisions for the leader wind turbine; a compensation drive spectrum distribution module, used by the leader wind turbine to distribute the first dynamic compensation drive spectrum to multiple follower wind turbines; and a low-latency synchronization compensation module, used by the multiple follower wind turbines to perform low-latency synchronization compensation by using the first dynamic compensation drive spectrum as a benchmark compensation drive parameter, correcting and outputting multiple compensation drive decision results based on local wind turbine state snapshots.

[0014] This application also provides a computer-readable storage medium, including: a computer program stored thereon, which, when executed by a processor, implements a dynamic compensation method based on wind turbine yaw error identification.

[0015] This application proposes a dynamic compensation method, system, and medium based on wind turbine yaw error identification. First, it backtracks the historical error compensation logs of the wind turbine cluster to perform yaw error operating condition compensation analysis, constructing a multi-scale compensation benchmark information database. The wind turbine cluster includes a leader wind turbine and multiple follower wind turbines. Then, within the cluster, the leader wind turbine locally integrates real-time yaw error and real-time operating conditions as operating condition feature vectors to query the multi-scale compensation benchmark information database, outputting a first dynamic compensation driving spectrum for compensation driving decisions. Next, the leader wind turbine distributes the first dynamic compensation driving spectrum to the multiple follower wind turbines. Finally, the multiple follower wind turbines use the first dynamic compensation driving spectrum as the benchmark compensation driving parameter, correcting and outputting multiple compensation driving decision results based on local wind turbine state snapshots, and executing low-latency synchronous compensation. This achieves the technical effect of adjusting the yaw error compensation strategy in real-time and accurately according to the dynamic operating conditions of the wind turbines, improving the accuracy and timeliness of yaw error compensation. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a flowchart illustrating the dynamic compensation method based on wind turbine yaw error identification provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of the structure of a dynamic compensation system based on wind turbine yaw error identification provided in an embodiment of this application.

[0019] Figure labeling: Multi-scale compensation reference information database construction module 10, first dynamic compensation driving spectrum output module 20, compensation driving spectrum distribution module 30, low-latency synchronous compensation module 40. Detailed Implementation

[0020] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0023] This application provides a dynamic compensation method based on wind turbine yaw error identification, such as... Figure 1 As shown, the method includes:

[0024] Step S100: Retrospectively analyze the historical error compensation logs of the wind turbine cluster, perform yaw error operating condition compensation analysis, and construct a multi-scale compensation benchmark information database. The wind turbine cluster includes a lead wind turbine and multiple follower wind turbines.

[0025] Specifically, historical yaw error compensation logs are extracted from the data acquisition and monitoring control system of the wind turbine cluster. These logs include yaw angle error, wind speed, wind direction, turbulence intensity, compensation action execution time, and changes in power generation after compensation. The historical error compensation logs include those of the lead wind turbine and follower wind turbines. The lead wind turbine, possessing stronger computing power, higher-precision sensors, and superior communication control, is responsible for global condition awareness, compensation strategy decision-making, and issuing instructions to other turbines. Follower wind turbines are ordinary turbines in the cluster that perform compensation actions, adjusting their yaw angles according to the instructions from the lead turbine and adaptively correcting based on local conditions to ensure safe and efficient compensation. Data cleaning algorithms, such as moving average filtering and outlier removal, are used to remove noisy data and ensure data quality.

[0026] Based on the degree of change in historical compensation data, the compensation work is divided into categories with varying levels of detail. From key aspects such as wind speed, wind instability, and the magnitude of yaw angle deviation, different compensation categories are further classified in detail, constructing a classification system that reflects various operating conditions and forming a multi-scale compensation benchmark information database.

[0027] In one possible implementation, the historical error compensation logs of the wind turbine cluster are traced back to perform yaw error compensation analysis and construct a multi-scale compensation benchmark information database. Step S100 further includes step S110, which, based on a preset field structure, retrieves multiple error compensation log fields of the leader wind turbine from the historical error compensation logs. These error compensation log fields include sample compensation amplitude, yaw error angle before sample compensation, sample compensation condition feature vector, and sample compensation driving spectrum. The sample compensation driving spectrum includes sample power increment and sample load change, and the sample compensation condition feature vector includes sample compensation wind speed and sample compensation turbulence intensity. Specifically, according to a pre-defined field structure specification, multiple error compensation log fields corresponding to the leader wind turbine are accurately extracted from the historical error compensation logs stored in the wind turbine cluster's data acquisition and monitoring control system. Among them, the sample compensation amplitude reflects the magnitude of historical compensation operations; the yaw error angle before sample compensation records the deviation between the wind turbine's yaw angle and the ideal angle before compensation; the sample compensation operating condition feature vector includes key information such as sample compensation wind speed and sample compensation turbulence intensity. The sample compensation wind speed refers to the wind speed of the wind turbine's environment during compensation, and the sample compensation turbulence intensity reflects the severity of wind speed fluctuations; the sample power increment in the sample compensation drive spectrum represents the increase in power generation after compensation, and the sample load change reflects the impact of compensation on the wind turbine's structural load.

[0028] Step S120: Dynamically define compensation scale grading rules based on the fluctuation range of the sample compensation amplitude. Specifically, perform statistical analysis on the sample compensation amplitude data extracted from historical error compensation logs to determine its maximum, minimum, and distribution values. Based on these statistical characteristics, dynamically classify the compensation scale levels according to certain logic and standards. For example, the sample compensation amplitude can be divided into several intervals in ascending order, with each interval corresponding to a compensation scale level, such as small-amplitude compensation, medium-amplitude compensation, and large-amplitude compensation.

[0029] Step S130: Decompose the multiple error compensation log fields into multi-scale compensation knowledge units according to the compensation scale classification rules. Specifically, according to the compensation scale classification rules defined in step S120, the multiple error compensation log fields extracted in step S110 are decomposed. Each error compensation log field is divided according to its corresponding compensation scale level, and log fields with the same compensation scale characteristics are combined to form independent multi-scale compensation knowledge units. For example, log fields with sample compensation amplitude in the small-amplitude compensation range are integrated into a small-amplitude compensation knowledge unit by combining their corresponding sample pre-compensation yaw error angle, sample compensation condition feature vector, and other information. This process is repeated to obtain multiple compensation knowledge units under different compensation scales.

[0030] Step S140: Based on the preset field structure, multi-scale working condition binning modeling of the multi-scale compensation knowledge unit is performed, constructing multiple three-dimensional working condition binning units. Specifically, according to the preset field structure specifications, the three key factors of wind speed dimension, turbulence intensity dimension, and yaw error angle dimension are used as the reference coordinates for modeling. First, for the wind speed dimension, based on the actual wind speed range of the wind turbine operation and the distribution characteristics of wind speed in historical data, the wind speed is divided into multiple intervals, such as a low wind speed interval of 0-3 m / s, a medium wind speed interval of 3-8 m / s, and a high wind speed interval of above 8 m / s. The specific interval division can be flexibly adjusted according to the actual wind field conditions and data analysis results. For the turbulence intensity dimension, the distribution of turbulence intensity in historical data is also referenced, combined with the wind turbine's adaptability range to turbulence intensity, and the turbulence intensity is divided into different level intervals. For example, turbulence intensity less than 5% is classified as low turbulence intensity, 5%-15% as medium turbulence intensity, and greater than 15% as high turbulence intensity. Regarding the yaw error angle dimension, based on the allowable range of yaw error angle during normal wind turbine operation and the distribution of yaw error angles in historical compensation data, yaw error angles are divided into intervals. For example, yaw error angles between -5° and 5° are classified as small error intervals, -10° to -5° and 5° to 10° as medium error intervals, and less than -10° or greater than 10° as large error intervals. Then, based on these intervals defined by these three dimensions, cross-combinations are performed. Each interval of wind speed, each interval of turbulence intensity, and each interval of yaw error angle are paired to form independent three-dimensional spatial regions. Each such three-dimensional spatial region constitutes a three-dimensional operating condition unit. For example, a three-dimensional working condition sub-unit composed of a low wind speed range, a low turbulence intensity range, and a small error range represents a specific working condition combination where the wind speed is between 0 and 3 m / s, the turbulence intensity is less than 5%, and the yaw error angle is between -5° and 5°.

[0031] Step S150: Retrieve multiple sets of follower compensation log fields from the historical error compensation log for the multiple follower wind turbines, and use these multiple sets of follower compensation log fields to perform fuzzy correction of the compensation errors of the multiple three-dimensional operating condition sub-units, obtaining multiple benchmark compensation knowledge units. Specifically, multiple sets of follower compensation log fields for the multiple follower wind turbines are extracted from the historical error compensation log. These fields also contain compensation-related information similar to that of the leader wind turbine. Since there is a deviation response delay between the follower wind turbine and the leader wind turbine when responding to compensation commands, this deviation response delay is used as a confidence index. Using the multiple sets of follower compensation log fields, combined with the confidence index, the multiple three-dimensional operating condition sub-units constructed in step S140 are subjected to fuzzy correction of compensation errors. For example, for a certain three-dimensional operating condition sub-unit, based on the difference between the actual compensation situation of the follower wind turbine and the expected compensation situation of the leader wind turbine, and the confidence index corresponding to the deviation response delay, the compensation knowledge within the sub-unit is adjusted and optimized, thereby obtaining multiple more accurate and reliable benchmark compensation knowledge units.

[0032] Step S160 involves associating and storing the multiple benchmark compensation knowledge units to construct the multi-scale compensation benchmark information database. Specifically, the multiple benchmark compensation knowledge units obtained in step S150 are associating and storing them according to certain logical relationships. For example, a database can be used, assigning a unique identifier to each benchmark compensation knowledge unit and establishing an index relationship between them to enable fast and accurate querying and retrieval. Through this associative storage method, all benchmark compensation knowledge units are integrated together to construct a complete multi-scale compensation benchmark information database, providing comprehensive data support and decision-making basis for subsequent yaw error compensation of wind turbines.

[0033] In one possible implementation, multi-scale operating condition binning modeling of the multi-scale compensation knowledge unit is performed based on a preset field structure, constructing multiple three-dimensional operating condition binning units. Step S140 further includes step S141, setting a first interval division rule for the wind speed dimension, a second level division rule for the turbulence intensity dimension, and a third interval division rule for the yaw error angle dimension to construct a three-dimensional partitioning space. Specifically, the first interval division rule for the wind speed dimension is set based on the actual wind speed range of the wind turbine and the distribution characteristics of wind speed in historical data. The second level division rule for the turbulence intensity dimension is formulated by referring to the distribution of turbulence intensity in historical data and combining it with the wind turbine's adaptability range to turbulence intensity. The third interval division rule for the yaw error angle dimension is set based on the allowable range of the yaw error angle during normal wind turbine operation and the distribution of the yaw error angle in historical compensation data. Using the three key factors—wind speed dimension, turbulence intensity dimension, and yaw error angle dimension—as reference coordinates, a three-dimensional partitioning space is constructed through the aforementioned set interval and level division rules. This three-dimensional space can encompass a variety of possible combinations of working conditions, providing a basic framework for multi-scale working condition box modeling.

[0034] Step S142 involves performing discretization mapping of the first-scale compensation knowledge unit in the three-dimensional partitioned space to obtain multiple first bin data nodes, where the filling information of the first bin data nodes is the sample compensation driving spectrum. Specifically, the first-scale compensation knowledge unit is any one of the multi-scale compensation knowledge units in step S130. The first-scale compensation knowledge unit is mapped to the three-dimensional partitioned space constructed in step S141 according to its corresponding wind speed, turbulence intensity, and yaw error angle information. Since the three-dimensional partitioned space is a discrete space composed of various intervals and levels, this mapping process actually discretizes the continuous compensation knowledge unit information into corresponding interval and level combinations. During the discretization mapping process, each first-scale compensation knowledge unit corresponds to a specific position in the three-dimensional partitioned space, and this position forms a first bin data node. For example, if a compensation knowledge unit has a wind speed in the medium wind speed interval, a turbulence intensity in the medium turbulence intensity interval, and a yaw error angle in the medium error interval, then it will be mapped to the position corresponding to the combination of these three intervals, forming a data node. Each first bin data node is filled with information about the sample compensation driving spectrum. By filling the data node with the information about the sample compensation driving spectrum, each data node not only represents a specific combination of operating conditions, but also contains information about the effect of the compensation operation under that operating condition.

[0035] Step S143 triggers the spatial aggregation mechanism for nodes with the same coordinates, merging the multiple first binning data nodes with the same three-dimensional coordinates to construct a first primary binning unit. Specifically, in the three-dimensional partitioned space, multiple first binning data nodes have the same three-dimensional coordinates, i.e., the same combination of wind speed range, turbulence intensity level, and yaw error angle range. The spatial aggregation mechanism for nodes with the same coordinates is triggered, merging all first binning data nodes located at the same three-dimensional coordinate position together. For example, if three first binning data nodes are all located at a combination of medium wind speed range, medium turbulence intensity range, and medium error angle range, then these three nodes will be merged together. After the merging operation, a first primary binning unit containing information from multiple original first binning data nodes is formed. This primary binning unit represents the set of all relevant compensation knowledge units under a specific working condition combination.

[0036] Step S144: Based on the sparsity of the data distribution in the first primary binning unit, a dynamic interval expansion operation is performed to generate a first three-dimensional load case binning unit. Specifically, the distribution of data nodes in the first primary binning unit is analyzed. If the number of data nodes corresponding to a certain three-dimensional coordinate combination is small, it indicates that the data under that load case combination is relatively sparse and insufficient to accurately reflect the compensation characteristics under that load case. For the first primary binning unit with sparse data distribution, a dynamic interval expansion operation is performed. For example, if there are few data nodes for a certain load case combination, the interval range of a certain dimension in that combination can be appropriately expanded, such as expanding the medium wind speed range to 2-9 m / s, or expanding the medium turbulence intensity range to 4%-16%, etc., to include more data nodes and make the data under that load case combination richer and more representative. After the dynamic interval expansion operation, a first three-dimensional load case binning unit containing more data information is obtained. This binning unit can more comprehensively reflect the compensation knowledge under the expanded load case combination, providing more accurate data support for subsequent yaw error compensation.

[0037] In step S200, within the wind turbine cluster, the leader wind turbine locally integrates real-time yaw error and real-time operating conditions as operating condition feature vectors to query the multi-scale compensation benchmark information database, outputs the first dynamic compensation driving spectrum, and makes compensation driving decisions for the leader wind turbine.

[0038] Specifically, during the operation of a wind turbine cluster, the leader wind turbine plays a core role. The leader wind turbine uses its onboard sensors to collect real-time yaw error data and current operating condition information, integrating this data to form an operating condition feature vector. The leader wind turbine uses this locally integrated real-time operating condition feature vector as a query condition to access a pre-built multi-scale compensation benchmark information database. This database stores a large amount of compensation knowledge for various operating conditions based on historical error compensation log analysis. By querying and matching in the database, the leader wind turbine finds the data record that best matches the current real-time operating condition feature vector and outputs the corresponding first dynamic compensation drive spectrum. Based on the output first dynamic compensation drive spectrum, the leader wind turbine makes its own yaw error compensation drive decisions, that is, based on the information provided by the drive spectrum, it determines how to adjust its yaw angle to achieve more precise yaw control, improve power generation efficiency, and ensure the safe and stable operation of the wind turbine.

[0039] In one possible implementation, within the wind turbine cluster, the leader wind turbine locally integrates real-time yaw error and real-time operating conditions as operating condition feature vectors to query the multi-scale compensation benchmark information database, outputting a first dynamic compensation drive spectrum, and making compensation drive decisions for the leader wind turbine. Step S200 further includes step S210, locating the target operating condition sub-unit in the multi-scale compensation benchmark information database by employing the real-time yaw error traversal compensation scale classification rules. Specifically, during operation, the leader wind turbine continuously acquires its own real-time yaw error data. Since the compensation scale classification rules have been dynamically defined based on the fluctuation range of sample compensation amplitude when constructing the multi-scale compensation benchmark information database, these rules divide and define different compensation scales. The leader wind turbine compares and traverses the acquired real-time yaw error data one by one according to this predetermined compensation scale classification rule. Through the traversal operation, the target operating condition sub-unit corresponding to the current real-time yaw error is accurately located in the multi-scale compensation benchmark information database. This target condition bin represents the various combinations of operating conditions that may occur under the current yaw error scale, as well as the corresponding set of compensation knowledge, which is used to provide the basic data range for compensation decisions.

[0040] Step S220: Map the real-time operating conditions and real-time yaw error to the target operating condition sub-unit to generate real-time virtual particle points. Specifically, after locating the target operating condition sub-unit, the leading wind turbine further maps the currently acquired real-time operating condition information and real-time yaw error data to the three-dimensional spatial model constructed by the target operating condition sub-unit. In this three-dimensional spatial model, each dimension corresponds to key factors such as wind speed, turbulence intensity, and yaw error angle. Through the mapping operation, the real-time operating conditions and real-time yaw error are transformed into a specific location point in this three-dimensional space. This location point is the real-time virtual particle point, which reflects the actual operating condition and yaw error of the wind turbine and exists in the spatial model of the target operating condition sub-unit in the form of spatial location.

[0041] Step S230: Retrieve N binning data nodes from the target working condition binning unit that satisfy a preset association distance threshold with the real-time virtual particle point. Specifically, the target working condition binning unit stores a large number of binning data nodes constructed based on historical data. Each node contains information such as the sample compensation driving spectrum under a specific working condition combination. To obtain the compensation information that best matches the current real-time working condition, a preset association distance threshold is set. This threshold is used to measure the spatial distance relationship between the real-time virtual particle point and each binning data node. By calculating the distance between the real-time virtual particle point and all binning data nodes in the target working condition binning unit, N binning data nodes whose distances satisfy the preset association distance threshold are selected. The working condition combinations represented by these N binning data nodes have a high similarity to the current real-time working condition, and the sample compensation driving spectrum information they contain has important reference value for determining the compensation strategy under the current real-time working condition.

[0042] Step S240: Based on the N spatial association distances between the N sub-bin data nodes and the real-time virtual particle points, perform spectral fusion of the N sample compensation driving spectra among the N sub-bin data nodes to output the first dynamic compensation driving spectrum. Specifically, after obtaining N sub-bin data nodes that meet the preset association distance thresholds with the real-time virtual particle points, since each sub-bin data node carries a sample compensation driving spectrum, and these nodes have different spatial association distances with the real-time virtual particle points, in order to comprehensively utilize the compensation information under these similar working conditions, perform a spectral fusion operation based on the N spatial association distances between these N sub-bin data nodes and the real-time virtual particle points. Specifically, the closer the spatial association distance of the sub-bin data node, the greater the weight of its sample compensation driving spectrum in the spectral fusion process; conversely, the farther the spatial association distance of the sub-bin data node, the smaller the weight of its sample compensation driving spectrum. By performing spectral fusion based on spatial association distance on the N sample compensation driving spectra, a first dynamic compensation driving spectrum that integrates compensation information from multiple similar working conditions is output. This dynamic compensation drive spectrum is more closely aligned with the actual operating conditions and yaw error of current wind turbines, providing more accurate and effective compensation drive decision-making basis for wind turbines, thereby achieving precise compensation for yaw error and improving the power generation efficiency and operational stability of wind turbines.

[0043] In step S300, the leader wind turbine sends the first dynamic compensation drive spectrum to multiple follower wind turbines.

[0044] Specifically, the first dynamic compensation drive spectrum includes the optimal compensation drive information derived after comprehensive analysis and processing, considering the current operating conditions of the wind turbine cluster and the yaw error of the lead wind turbine itself. For the coordinated and optimized operation of the entire wind turbine cluster, the lead wind turbine assumes the role of information transmission and coordination. The lead wind turbine distributes the generated first dynamic compensation drive spectrum to multiple follower wind turbines in the cluster through a pre-set communication link. This provides crucial information for the follower wind turbines to make compensation drive decisions, ensuring that the entire cluster can respond quickly and make reasonable adjustments when facing complex and changing operating conditions.

[0045] In step S400, the plurality of following wind turbines use the first dynamic compensation drive spectrum as a reference to compensate drive parameters, and perform low-latency synchronous compensation of the plurality of following wind turbines by correcting and outputting multiple compensation drive decision results based on the local wind turbine state snapshot.

[0046] Specifically, because each follower wind turbine has its own unique operating state and characteristics, after receiving the first dynamic compensation drive spectrum from the leader wind turbine, the multiple follower wind turbines do not directly execute it as the final compensation drive command. Instead, they first acquire a local wind turbine state snapshot, which refers to the real-time state collected by the follower wind turbines through local sensors, including various key operating parameters of the wind turbine at a certain moment. The follower wind turbines use the received first dynamic compensation drive spectrum as the benchmark compensation drive parameters, and combine the information from the local state snapshot to make targeted corrections to the first dynamic compensation drive spectrum. Through this correction, the compensation drive decision is made more closely aligned with the actual operating conditions of each follower wind turbine, ensuring the effectiveness and accuracy of the compensation measures. After correction, each follower wind turbine outputs a compensation drive decision result that meets its own actual needs. To achieve synchronized and optimized operation of the entire wind turbine cluster, the follower wind turbines employ low-latency synchronization control technology, such as using hardware timers or a real-time operating system, to ensure that their respective compensation drive decision results are executed at almost the same time. Among them, the low-latency synchronization compensation mechanism is used to minimize the performance fluctuations of the cluster caused by the difference in compensation time of different wind turbines, so that the entire wind turbine cluster can operate as a whole efficiently and stably when facing complex operating conditions, thereby improving the power generation efficiency and operational reliability of the entire cluster.

[0047] In one possible implementation, the multiple following wind turbines use the first dynamic compensation drive spectrum as a reference for compensation drive parameters, and perform low-latency synchronous compensation of the multiple following wind turbines by correcting and outputting multiple compensation drive decision results based on the local wind turbine state snapshot. Step S400 further includes step S410, retrieving the Wth following yaw error from the local wind turbine state snapshot of the Wth following wind turbine. Specifically, in the wind turbine cluster operation system, each following wind turbine is equipped with a sensor network. These sensors collect various operating parameters of the wind turbine in real time, including but not limited to wind speed sensors, wind direction sensors, speed sensors, and yaw error sensors. These sensors transmit the collected data to the local controller of the wind turbine. The controller packages the data according to a preset time interval to form a local wind turbine state snapshot. For the Wth following wind turbine, when compensation drive decision correction is required, the local controller retrieves the Wth following yaw error data from the state snapshot data packet based on a specific data identifier.

[0048] Step S420: Calculate the Wth three-dimensional state offset between the real-time yaw error and the Wth following yaw error. Specifically, after the Wth following wind turbine obtains the real-time yaw error data of the leader wind turbine through the high-speed communication network within the cluster, it compares it with the Wth following yaw error it retrieves to calculate the three-dimensional state offset. The specific calculation process is as follows: First, a three-dimensional spatial model is constructed using wind speed, wind direction, and yaw error as the three coordinate axes. The wind speed and wind direction parameters corresponding to the real-time yaw error of the leader wind turbine and the wind speed and wind direction parameters corresponding to the Wth following yaw error of the Wth following wind turbine are mapped onto this three-dimensional space, forming two spatial points. Then, the distance between these two spatial points is calculated using the Euclidean distance formula, and weighted by combining the cosine value of the wind direction angle, finally yielding the Wth three-dimensional state offset.

[0049] Step S430: Based on the Wth three-dimensional state offset, correct the first dynamic compensation drive spectrum and output the Wth compensation drive decision result. Specifically, after obtaining the Wth three-dimensional state offset, the Wth following wind turbine uses a fuzzy logic-based correction algorithm to correct the first dynamic compensation drive spectrum. The specific correction process is as follows: First, the Wth three-dimensional state offset is input into a pre-constructed fuzzy logic correction model. This model converts the offset into membership values ​​in fuzzy sets through fuzzification processing based on the magnitude and direction of the offset. For example, the yaw error offset is divided into three fuzzy sets: "small," "medium," and "large," and the membership degree of each fuzzy set is determined based on the actual offset. Then, based on a preset fuzzy rule base, inference calculations are performed on the membership values ​​to obtain the fuzzy output of the corrected compensation drive spectrum parameters. Finally, through defuzzification processing, the fuzzy output is converted into specific numerical values, thereby realizing the correction of parameters such as compensation amplitude and compensation frequency in the first dynamic compensation drive spectrum and outputting the Wth compensation drive decision result. This result can better adapt to the actual operating state of the Wth following wind turbine and improve the accuracy and effectiveness of compensation. The preset fuzzy rule base is derived from a large amount of experimental data and expert experience. For example, "if the yaw error offset is large and the wind speed offset is small, the compensation range should be appropriately increased."

[0050] Step S440: After performing load safety boundary verification on the Wth compensation driving decision result, the low-latency synchronous compensation of the Wth following wind turbine is performed using the Wth compensation driving decision result. Specifically, before applying the Wth compensation driving decision result to the Wth following wind turbine, load safety boundary verification needs to be performed to ensure that the compensation operation will not threaten the structural safety of the wind turbine. The specific verification process is as follows: First, using finite element analysis software, a three-dimensional finite element model of the wind turbine is established based on the structural parameters of the Wth following wind turbine and the current operating conditions. Then, the Wth compensation driving decision result is input into the model to simulate the load changes generated by the compensation operation on each component of the wind turbine. By comparing and analyzing with the pre-set load safety boundary value, it is determined whether the compensation operation will cause the component load to exceed the safe range. If the verification is successful, that is, the load of all components is within the safe boundary, the Wth compensation driving decision result is used to perform the low-latency synchronous compensation operation through the actuator of the wind turbine.

[0051] In one possible implementation, the method further includes: driving distributed wind turbines to collect yaw correlation data and outputting distributed yaw correlation data time-series slices; dividing the distributed wind turbines into clusters based on the distributed yaw correlation data time-series slices to obtain K wind turbine clusters; and performing collaborative wind turbine yaw error dynamic compensation decision-making in the K wind turbine clusters.

[0052] Specifically, in a distributed wind turbine system, each wind turbine is equipped with a data acquisition device, which includes various types of sensors, such as wind direction sensors, yaw position sensors, and wind speed sensors. These sensors continuously collect data related to the yaw of the wind turbine, including wind direction, yaw angle, and wind speed. The collected data is transmitted in real time to a high-speed data storage and processing unit on the wind turbine. This unit slices the collected data according to preset time intervals, dividing the continuous data stream into independent, time-stamped data segments, i.e., distributed yaw-related data time-series slices. Each time-series slice contains key yaw-related data such as wind direction, yaw angle, and wind speed for a specific wind turbine within that time period.

[0053] After obtaining the time-series slices of distributed yaw correlation data, clustering of distributed wind turbines is performed through data analysis. For example, a clustering algorithm is first used to process data such as wind direction, yaw angle change patterns, and wind speed characteristics in the time-series slices of distributed yaw correlation data. During the clustering process, the similarity between each data point and other data points is calculated, and data points with high similarity are grouped into the same cluster. By continuously iterating and optimizing the cluster centers, data points within the same cluster have high similarity, while data points between different clusters have significant differences. Here, each data point represents the yaw correlation data of a wind turbine within a specific time period. After clustering analysis, the originally scattered distributed wind turbines are divided into K different clusters, each cluster corresponding to a wind turbine cluster. The wind turbines in these clusters have similar yaw characteristics, for example, they may be in similar wind direction environments or have similar yaw angle change trends. For the K wind turbine clusters obtained, a collaborative wind turbine yaw error dynamic compensation decision is performed using a method similar to steps 100-400.

[0054] In one possible implementation, the distributed wind turbine clusters are divided based on the distributed yaw correlation data time-series slices to obtain K wind turbine clusters. This further includes: extracting a first yaw correlation data time-series slice from the distributed yaw correlation data time-series slices of the first individual wind turbine in the distributed wind turbine clusters, wherein the first yaw correlation data time-series slice includes a yaw error angle time-series slice, a time-series wind speed slice, a time-series wind direction slice, and a turbulence intensity time-series slice; and extracting a first time-series feature vector based on the first yaw correlation data time-series slice. In this process, the first temporal feature vector is composed of a first dynamic morphological similarity, a first wind speed error covariance factor, and a first spatial weight feature; by analogy, M temporal feature vectors of M individual wind turbines in the distributed wind turbine are constructed; agglomerative hierarchical clustering algorithm is applied, and the Ward variance minimization criterion is used as the clustering basis to perform clustering calculations on the M temporal feature vectors corresponding to the M individual wind turbines, generating multiple initial clusters; predefined spatial distance constraints are used to traverse the multiple initial clusters, and constraint decoupling operations are performed on the long-distance clusters to output the K wind turbine clusters.

[0055] Specifically, for the first individual wind turbine in the distributed wind turbine system, i.e., any single wind turbine, a corresponding first yaw-related data time-series slice is extracted. Specifically, the yaw error angle time-series slice records the yaw angle deviation of the wind turbine relative to the ideal wind direction at different times. The yaw angle of the wind turbine is acquired in real time using a yaw angle sensor and recorded at preset time intervals to form the yaw error angle time-series slice. The time-series wind speed slice reflects the change in wind speed at the location of the wind turbine over time. Wind speed is measured using a wind speed sensor, and a time-series wind speed slice is generated. The time-series wind direction slice records the dynamic changes in wind direction, which is sensed and recorded in real time using a wind direction sensor. The turbulence intensity time-series slice describes the severity of wind speed fluctuations. Turbulence intensity is calculated through statistical analysis of wind speed data, such as calculating the ratio of the standard deviation of wind speed to the average wind speed, and recorded in a time series to form a turbulence intensity time-series slice. These four time-series slices together constitute the first yaw-related data time-series slice.

[0056] To more effectively classify and cluster wind turbines, representative features are extracted from the time-series slices of the first yaw correlation data to construct a first time-series feature vector. The first dynamic morphological similarity is an indicator that measures the similarity of the morphological changes of the first individual wind turbine's yaw error angle time-series slices across different time periods. By employing a dynamic time warping algorithm, the yaw error angle time-series data from different time periods are aligned and compared, and their similarity values ​​are calculated. The closer this value is to 1, the more similar the yaw error angle changes across different time periods, reflecting the stability and regularity of the wind turbine's yaw behavior. The first wind speed error covariance factor describes the cooperative relationship between the time-series wind speed slices and the yaw error angle time-series slices. The wind speed error covariance factor is obtained by calculating the covariance between these two time-series data and standardizing it. This factor reflects the degree of influence of wind speed changes on the wind turbine's yaw error; a larger covariance factor indicates a strong correlation between wind speed changes and yaw error. The first spatial weight feature is used to characterize the geographical location information of the first individual wind turbine in the distributed system. Based on the spatial distance between wind turbines, a Gaussian kernel function is used to calculate spatial weights, converting spatial distance into weight values. The spatial weight feature reflects the degree of spatial interaction between wind turbines; wind turbines that are closer together have greater spatial weights and exhibit higher similarity in cluster partitioning. These three features together constitute the first temporal feature vector, used to comprehensively and accurately describe the yaw correlation characteristics of the first individual wind turbine.

[0057] After constructing the first time-series feature vector for the first individual wind turbine, the same method and process were applied to the other M-1 individual wind turbines in the distributed wind turbine system. For each wind turbine, the steps of extracting yaw correlation data time-series slices from distributed yaw correlation data time-series slices were followed to obtain its corresponding yaw error angle time-series slice, time-series wind speed slice, time-series wind direction slice, and turbulence intensity time-series slice. Then, using the same feature extraction method as for the first individual wind turbine, the dynamic morphological similarity, wind speed error covariance factor, and spatial weight features of each wind turbine were calculated to construct M time-series feature vectors for the M individual wind turbines. These time-series feature vectors have the same dimension and structure, facilitating unified cluster analysis.

[0058] A cohesive hierarchical clustering algorithm is employed. This algorithm is a bottom-up clustering method that starts with each data point, i.e., each wind turbine's time-series feature vector, as a separate cluster, and gradually merges similar clusters until a preset stopping condition is met. During the clustering process, the Ward variance minimization criterion is used as the basis for clustering. This criterion aims to minimize the increase in intra-cluster variance after each cluster merge. Specifically, for each possible cluster merge, the increase in intra-cluster variance after the merge is calculated, and the cluster pair that minimizes the increase in variance is selected for merging. Through continuous iterative merging, similar time-series feature vectors are grouped together, gradually forming multiple initial clusters. The wind turbines within these initial clusters exhibit certain similarities in yaw correlation characteristics.

[0059] Since the spatial relationship between wind turbines in a distributed wind turbine system is crucial for cluster partitioning, spatial distance constraints are first predefined. These constraints are set based on the actual wind turbine layout and operational requirements, for example, stipulating that the maximum spatial distance between wind turbines within the same cluster must not exceed a certain threshold, such as 500 meters. Then, multiple generated initial clusters are traversed, checking whether the spatial distance between wind turbines within each cluster meets the predefined spatial distance constraints. For initial clusters with out-of-distance conditions—that is, where the distance between wind turbines within a cluster exceeds the set threshold—constraint decoupling operations are performed. Specifically, the out-of-distance wind turbines are separated from their original clusters and reassigned to suitable clusters based on their spatial distance and temporal similarity to other clusters, or treated as a separate new cluster. Through continuous adjustment and optimization, until all wind turbines within all clusters meet the spatial distance constraints, K compliant wind turbine clusters are finally output. These clusters better reflect the yaw correlation characteristics and spatial similarity of wind turbines in the distributed wind turbine system.

[0060] In one possible implementation, the method further includes: extracting multiple yaw sensor errors, multiple spatial coordinates, and multiple average fault intervals from multiple individual wind turbines in the Kth wind turbine cluster; performing a three-weighted scoring assignment based on the multiple yaw sensor errors, multiple spatial coordinates, and multiple average fault intervals, and outputting multiple three-weighted score values; locating the Kth leader wind turbine and the Kth group of follower wind turbines among the multiple individual wind turbines according to the descending sorting result of the multiple three-weighted score values; and binding the Kth leader wind turbine and the Kth group of follower wind turbines based on communication dependency relationships.

[0061] Specifically, each individual wind turbine is equipped with a yaw sensor. However, due to factors such as manufacturing precision, installation errors, and wear during long-term operation, there is a certain deviation between the sensor's measured value and the true value, known as yaw sensor error. Using sensor calibration equipment and data analysis algorithms, the yaw sensors of each wind turbine are periodically calibrated and their errors calculated, resulting in multiple yaw sensor error data points for multiple individual wind turbines. Positioning equipment is installed on each wind turbine to acquire its precise geographical location information in real time, obtaining multiple spatial coordinates of the individual wind turbines, expressed in longitude and latitude. The mean fault interval reflects the reliability and operational stability of the wind turbine. By analyzing the historical fault records of each wind turbine, the number of faults occurring within a certain time range and the time interval between each fault are statistically analyzed. The average of these time intervals is calculated to obtain the mean fault interval for each wind turbine. A longer mean fault interval indicates more stable wind turbine operation and a lower frequency of faults.

[0062] A three-weighted scoring method was used to evaluate multiple individual wind turbines in the Kth wind turbine cluster. First, based on actual operational requirements and expert experience, appropriate weights were assigned to each factor. For example, yaw sensor error has a significant impact on the yaw control accuracy of the wind turbine and can be assigned a high weight, such as 0.5; spatial location coordinates are important for communication and coordination in collaborative control and are assigned a weight of 0.3; mean fault interval reflects the reliability of the wind turbine and is assigned a weight of 0.2. Then, each factor was standardized, converting data of different dimensions into a unified scoring range, such as 0-100 points. For yaw sensor error, the smaller the error, the higher the score; for spatial location coordinates, the score was based on their reasonable distribution within the cluster, with positions that better meet collaborative control requirements receiving higher scores; for mean fault interval, longer intervals received higher scores. Finally, based on the assigned weights and the standardized scores, a three-weighted score was calculated for each wind turbine. The specific calculation formula is as follows: Three-weighted score = Yaw sensor error score × Yaw sensor error weight + Spatial position coordinate score × Spatial position coordinate weight + Average fault interval score × Average fault interval weight. In this way, multiple three-weighted scores are output for multiple individual wind turbines in the Kth wind turbine cluster. These scores comprehensively and objectively reflect the overall performance and collaborative control potential of each wind turbine in the cluster.

[0063] After obtaining multiple three-weighted scores for individual wind turbines in the Kth wind turbine cluster, these scores are sorted in descending order. The wind turbine ranking first, with the highest three-weighted score, indicates excellent performance in yaw control accuracy, spatial positioning rationality, and operational reliability; therefore, it is designated as the Kth leader wind turbine. The remaining wind turbines are designated as the Kth group of follower wind turbines. A binding mechanism based on communication dependency is established between the Kth leader wind turbine and the Kth group of follower wind turbines. On the hardware side, high-speed and stable communication equipment is provided for both the leader and follower wind turbines. Through this equipment, the leader wind turbine can establish a two-way communication link with the follower wind turbines, enabling real-time data transmission and accurate command issuance. On the software side, a communication protocol and cooperative control algorithm are configured. The communication protocol defines the format, frequency, and encryption method of data transmission between the leader and follower wind turbines, ensuring communication security and reliability. The collaborative control algorithm generates corresponding instructions based on the wind field information and cluster operation status collected by the leader wind turbine, and sends them to the follower wind turbines via a communication link. After receiving the instructions, the follower wind turbines adjust their own operating parameters according to the algorithm requirements to achieve collaborative operation with the leader wind turbine.

[0064] This application's embodiments employ historical error compensation logs of wind turbine clusters to analyze yaw error compensation under various operating conditions, constructing a multi-scale compensation benchmark information database. The leading wind turbine integrates real-time yaw error and operating conditions to form an operating condition feature vector, queries the database to output a first dynamic compensation driving spectrum, makes its own compensation driving decision, and simultaneously distributes the first dynamic compensation driving spectrum to multiple following wind turbines. These following wind turbines use this as a benchmark, combine it with local state snapshot corrections, output multiple compensation driving decision results, and perform low-latency synchronous compensation. These techniques solve the technical problems of existing wind turbine yaw error compensation methods, such as the inability to adapt to dynamic operating conditions in real time and insufficient compensation accuracy. This achieves the technical effect of adjusting the yaw error compensation strategy in real time and accurately according to the dynamic operating conditions of the wind turbine, improving the accuracy and timeliness of yaw error compensation.

[0065] In the above text, refer to Figure 1 A dynamic compensation method based on wind turbine yaw error identification according to an embodiment of the present invention is described in detail. Next, reference will be made to... Figure 2 A dynamic compensation system based on wind turbine yaw error identification is described according to an embodiment of the present invention.

[0066] The dynamic compensation system based on wind turbine yaw error identification according to embodiments of the present invention addresses the technical problems of existing wind turbine yaw error compensation methods, such as inability to adapt to dynamic operating conditions in real time and insufficient compensation accuracy. It achieves the technical effect of adjusting the yaw error compensation strategy in real time and accurately according to the dynamic operating conditions of the wind turbine, thereby improving the accuracy and timeliness of yaw error compensation. The dynamic compensation system based on wind turbine yaw error identification includes: a multi-scale compensation benchmark information database construction module 10, a first dynamic compensation driving spectrum output module 20, a compensation driving spectrum distribution module 30, and a low-latency synchronous compensation module 40.

[0067] A multi-scale compensation benchmark information database construction module 10 is used to backtrack the historical error compensation logs of the wind turbine cluster, perform yaw error operating condition compensation analysis, and construct a multi-scale compensation benchmark information database. The wind turbine cluster includes a leader wind turbine and multiple follower wind turbines. A first dynamic compensation drive spectrum output module 20 is used in the wind turbine cluster to query the multi-scale compensation benchmark information database by integrating real-time yaw error and real-time operating conditions locally as operating condition feature vectors, outputting a first dynamic compensation drive spectrum, and making compensation drive decisions for the leader wind turbine. A compensation drive spectrum distribution module 30 is used by the leader wind turbine to distribute the first dynamic compensation drive spectrum to multiple follower wind turbines. A low-latency synchronization compensation module 40 is used by the multiple follower wind turbines to perform low-latency synchronization compensation by using the first dynamic compensation drive spectrum as a benchmark compensation drive parameter, correcting and outputting multiple compensation drive decision results based on local wind turbine state snapshots.

[0068] The detailed configuration of the multi-scale compensation benchmark information database construction module 10 is explained below: As mentioned above, the historical error compensation logs of the wind turbine cluster are traced back to perform yaw error condition compensation analysis and construct a multi-scale compensation benchmark information database. The multi-scale compensation benchmark information database construction module 10 may further include: an error compensation log field retrieval unit used to retrieve multiple error compensation log fields of the leader wind turbine from the historical error compensation logs based on a preset field structure. The error compensation log fields include sample compensation amplitude, yaw error angle before sample compensation, sample compensation condition feature vector, and sample compensation driving spectrum. The sample compensation driving spectrum includes sample power increment and sample load change. The sample compensation condition feature vector includes sample compensation wind speed and sample compensation turbulence intensity. The compensation scale... The grading rule definition unit is used to dynamically define the grading rules for compensation scale based on the fluctuation range of the sample compensation amplitude; the decomposition unit is used to decompose the multiple error compensation log fields into multi-scale compensation knowledge units according to the compensation scale grading rules; the multi-scale operating condition bin modeling unit is used to perform multi-scale operating condition bin modeling of the multi-scale compensation knowledge units based on a preset field structure, and construct multiple three-dimensional operating condition bin units; the compensation error fuzzy correction unit is used to retrieve multiple sets of follower compensation log fields of the multiple follower wind turbines from the historical error compensation logs, and use the multiple sets of follower compensation log fields to perform fuzzy correction of the compensation error of the multiple three-dimensional operating condition bin units, and obtain multiple benchmark compensation knowledge units; the associated storage unit is used to associate and store the multiple benchmark compensation knowledge units to construct the multi-scale compensation benchmark information database.

[0069] The multi-scale load cell modeling of the multi-scale compensation knowledge unit is based on a preset field structure, constructing multiple three-dimensional load cell units. The multi-scale load cell modeling unit may further include: a three-dimensional partitioning space construction subunit for setting a first interval partitioning rule for wind speed dimension, a second-level partitioning rule for turbulence intensity dimension, and a third interval partitioning rule for yaw error angle dimension, constructing a three-dimensional partitioning space; a discretization mapping subunit for performing discretization mapping of the first-scale compensation knowledge unit in the three-dimensional partitioning space, obtaining multiple first-bin data nodes, wherein the filling information of the first-bin data nodes is a sample compensation driving spectrum; a merging subunit for triggering a spatial aggregation mechanism for nodes with the same coordinates, merging the multiple first-bin data nodes with the same three-dimensional coordinates, constructing a first primary load cell unit; and a dynamic interval expansion operation subunit for performing a dynamic interval expansion operation based on the data distribution sparsity of the first primary load cell unit, generating a first three-dimensional load cell unit.

[0070] The system may further include: a yaw correlation data acquisition module for driving distributed wind turbines to acquire yaw correlation data and outputting distributed yaw correlation data time-series slices; a cluster partitioning module for partitioning the distributed wind turbines into clusters based on the distributed yaw correlation data time-series slices, resulting in K wind turbine clusters; and a collaborative wind turbine yaw error dynamic compensation decision module for executing collaborative wind turbine yaw error dynamic compensation decisions in the K wind turbine clusters.

[0071] The detailed description of the cluster partitioning module is as follows: As mentioned above, the distributed wind turbines are partitioned into clusters based on the distributed yaw correlation data time-series slices to obtain K wind turbine clusters. The cluster partitioning module may further include: a first yaw correlation data time-series slice extraction unit for extracting the first yaw correlation data time-series slice of the first individual wind turbine in the distributed wind turbines from the distributed yaw correlation data time-series slices, wherein the first yaw correlation data time-series slice includes a yaw error angle time-series slice, a time-series wind speed slice, a time-series wind direction slice, and a turbulence intensity time-series slice; and a first time-series feature vector extraction unit for extracting the first yaw correlation data... A time-series slice extracts a first time-series feature vector, which is composed of a first dynamic morphological similarity, a first wind speed error covariance factor, and a first spatial weight feature. An analogy construction unit is used to analogically construct M time-series feature vectors for M individual wind turbines in the distributed wind turbines. A clustering calculation unit is used to apply agglomerative hierarchical clustering algorithm, using the Ward variance minimization criterion as the clustering basis, to perform clustering calculations on the M time-series feature vectors corresponding to the M individual wind turbines, generating multiple initial clusters. A constraint decoupling unit is used to predefine spatial distance constraints, traverse the multiple initial clusters, perform constraint decoupling operations on the long-distance clusters, and output the K wind turbine clusters.

[0072] The system may further include: a data extraction module for extracting multiple yaw sensor errors, multiple spatial coordinates, and multiple average fault intervals from multiple individual wind turbines in the Kth wind turbine cluster; a three-weighted scoring assignment module for performing three-weighted scoring assignment based on the multiple yaw sensor errors, multiple spatial coordinates, and multiple average fault intervals, and outputting multiple three-weighted score values; a wind turbine positioning module for locating the Kth leader wind turbine and the Kth group of follower wind turbines among the multiple individual wind turbines according to the descending sorting result of the multiple three-weighted score values; and a wind turbine binding module for binding the Kth leader wind turbine and the Kth group of follower wind turbines based on communication dependency relationships.

[0073] The detailed description of the specific configuration of the first dynamic compensation drive spectrum output module 20 is explained as follows: As mentioned above, in the wind turbine cluster, the leader wind turbine locally integrates real-time yaw error and real-time operating conditions as operating condition feature vectors to query the multi-scale compensation benchmark information library, outputs the first dynamic compensation drive spectrum, and makes compensation drive decisions for the leader wind turbine. The first dynamic compensation drive spectrum output module 20 may further include: a target operating condition sub-bin unit positioning unit used to locate the target operating condition sub-bin unit in the multi-scale compensation benchmark information library by adopting the real-time yaw error traversal compensation scale classification rules; a real-time virtual particle point generation unit used to map the real-time operating conditions and real-time yaw error to the target operating condition sub-bin unit to generate real-time virtual particle points; a sub-bin data node retrieval unit used to retrieve N sub-bin data nodes from the target operating condition sub-bin unit that satisfy a preset association distance threshold with the real-time virtual particle points; and a spectrum fusion unit used to perform spectrum fusion of the N sample compensation drive spectra in the N sub-bin data nodes based on the N spatial association distances between the N sub-bin data nodes and the real-time virtual particle points, and output the first dynamic compensation drive spectrum.

[0074] The detailed description of the specific configuration of the low-latency synchronization compensation module 40 is explained as follows: As mentioned above, multiple following wind turbines use the first dynamic compensation drive spectrum as a reference for compensation drive parameters, and correct and output multiple compensation drive decision results based on the local wind turbine state snapshot to perform low-latency synchronization compensation for the multiple following wind turbines. The low-latency synchronization compensation module 40 may further include: a following yaw error retrieval unit for retrieving the Wth following yaw error from the local wind turbine state snapshot of the Wth following wind turbine; a three-dimensional state offset calculation unit for calculating the Wth three-dimensional state offset between the real-time yaw error and the Wth following yaw error; a first dynamic compensation drive spectrum correction unit for correcting the first dynamic compensation drive spectrum based on the Wth three-dimensional state offset and outputting the Wth compensation drive decision result; and a load safety boundary verification unit for performing load safety boundary verification on the Wth compensation drive decision result and then using the Wth compensation drive decision result to perform low-latency synchronization compensation for the Wth following wind turbine.

[0075] The dynamic compensation system based on wind turbine yaw error identification provided in the embodiments of the present invention can execute the dynamic compensation method based on wind turbine yaw error identification provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0076] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0077] Based on the foregoing embodiments, this application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor of an electronic device, it can implement the dynamic compensation method based on wind turbine yaw error identification as described in any of the foregoing embodiments.

[0078] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A dynamic compensation method based on wind turbine yaw error identification, characterized in that, The method includes: By reviewing the historical error compensation logs of the wind turbine cluster, yaw error compensation analysis is performed, and a multi-scale compensation benchmark information database is constructed. The wind turbine cluster includes a leader wind turbine and multiple follower wind turbines. In the wind turbine cluster, the leader wind turbine locally integrates real-time yaw error and real-time operating conditions as operating condition feature vectors to query the multi-scale compensation benchmark information database, outputs the first dynamic compensation drive spectrum, and makes compensation drive decisions for the leader wind turbine. The leader wind turbine sends the first dynamic compensation drive spectrum to multiple follower wind turbines; The plurality of following wind turbines use the first dynamic compensation drive spectrum as a reference for compensation drive parameters, and correct and output multiple compensation drive decision results based on the local wind turbine state snapshot to perform low-latency synchronous compensation of the plurality of following wind turbines. The method further includes: Drive distributed wind turbines to collect yaw correlation data and output time-series slices of distributed yaw correlation data; Based on the time-series slices of the distributed yaw correlation data, the distributed wind turbine clusters are divided to obtain K wind turbine clusters; Cooperative dynamic compensation decision-making for yaw error of wind turbines is performed in the K wind turbine clusters; Based on the time-series slices of the distributed yaw correlation data, the distributed wind turbine clusters are divided to obtain K wind turbine clusters. The method includes: The first yaw correlation data time-series slice of the first individual wind turbine in the distributed wind turbine is extracted from the distributed yaw correlation data time-series slice, wherein the first yaw correlation data time-series slice includes yaw error angle time-series slice, time-series wind speed slice, time-series wind direction slice and turbulence intensity time-series slice; Based on the first yaw correlation data time-series slice, a first time-series feature vector is extracted. The first time-series feature vector is composed of a first dynamic morphological similarity, a first wind speed error covariance factor, and a first spatial weight feature. The first dynamic morphological similarity is an indicator that measures the similarity of the morphological changes of the first individual wind turbine yaw error angle time-series slice in different time periods. The first wind speed error covariance factor is used to describe the cooperative change relationship between the time-series wind speed slice and the yaw error angle time-series slice. The first spatial weight feature is used to characterize the geographical location information of the first individual wind turbine in the distributed system. By analogy, construct M time-series feature vectors for the M individual wind turbines in the distributed wind turbine system; A condensed hierarchical clustering algorithm is applied, and the Ward variance minimization criterion is used as the clustering basis to perform clustering calculations on the M time-series feature vectors corresponding to the M individual wind turbines, generating multiple initial clusters. The multiple initial clusters are traversed using predefined spatial distance constraints. Constraint decoupling operations are performed on the super-distance clusters, and the K wind turbine clusters are output. The method further includes: Extract the average values ​​of multiple yaw sensor errors, multiple spatial position coordinates, and multiple fault intervals from multiple individual wind turbines in the Kth wind turbine cluster; Based on the errors of the multiple yaw sensors, multiple spatial position coordinates, and multiple fault intervals, a three-weighted scoring assignment is performed, and multiple three-weighted scoring values ​​are output. Based on the descending sorting results of the multiple three-weighted score values, the Kth leader wind turbine and the Kth group of follower wind turbines are located among the multiple individual wind turbines; The Kth leader wind turbine and the Kth group of follower wind turbines are bound together based on the communication hierarchy.

2. The dynamic compensation method based on wind turbine yaw error identification as described in claim 1, characterized in that, The method involves tracing back historical error compensation logs of a wind turbine cluster, performing yaw error compensation analysis under operating conditions, and constructing a multi-scale compensation benchmark information database. Based on a preset field structure, multiple error compensation log fields of the leader wind turbine are retrieved from the historical error compensation log. The error compensation log fields include sample compensation magnitude, yaw error angle before sample compensation, sample compensation condition feature vector, and sample compensation drive spectrum. The sample compensation drive spectrum includes sample power increment and sample load change. The sample compensation condition feature vector includes sample compensation wind speed and sample compensation turbulence intensity. The compensation scale grading rules are dynamically defined based on the fluctuation range of the sample compensation amplitude. Based on the aforementioned compensation scale classification rules, the multiple error compensation log fields are decomposed into multi-scale compensation knowledge units; Based on the preset field structure, multi-scale working condition binning modeling of the multi-scale compensation knowledge unit is performed to construct multiple three-dimensional working condition binning units. The historical error compensation log retrieves multiple sets of follower compensation log fields from the multiple follower wind turbines, and uses the multiple sets of follower compensation log fields to perform fuzzy correction of the compensation error of the multiple three-dimensional working condition sub-units, thereby obtaining multiple benchmark compensation knowledge units. The multiple benchmark compensation knowledge units are associated and stored to construct the multi-scale compensation benchmark information database.

3. The dynamic compensation method based on wind turbine yaw error identification as described in claim 2, characterized in that, The method involves multi-scale load cell modeling of the multi-scale compensation knowledge unit based on a preset field structure, constructing multiple three-dimensional load cell units, and includes: A three-dimensional partitioning space is constructed by defining the first interval division rule for wind speed dimension, the second level division rule for turbulence intensity dimension, and the third interval division rule for yaw error angle dimension. The discretization mapping of the first-scale compensation knowledge unit in the three-dimensional partition space is performed to obtain multiple first-bin data nodes, wherein the filling information of the first-bin data nodes is the sample compensation driving spectrum. Trigger the spatial aggregation mechanism of nodes with the same coordinates, merge the multiple first bin data nodes with the same three-dimensional coordinates, and construct the first primary bin unit; Based on the data sparsity of the first primary binning unit, a dynamic interval expansion operation is performed to generate the first three-dimensional working condition binning unit.

4. The dynamic compensation method based on wind turbine yaw error identification as described in claim 3, characterized in that, In the wind turbine cluster, the leader wind turbine locally integrates real-time yaw error and real-time operating conditions as operating condition feature vectors to query the multi-scale compensation benchmark information database, outputs a first dynamic compensation drive spectrum, and makes compensation drive decisions for the leader wind turbine. The method includes: By employing the real-time yaw error traversal compensation scale classification rule, the target working condition sub-unit is located in the multi-scale compensation benchmark information database. Map the real-time operating conditions and real-time yaw error to the target operating condition sub-unit to generate real-time virtual particle points; N data nodes of the sub-bin that satisfy a preset association distance threshold with the real-time virtual particle points are retrieved from the target working condition sub-bin unit; Based on the N spatial association distances between the N binning data nodes and the real-time virtual particle points, the spectrum fusion of the N sample compensation driving spectra in the N binning data nodes is performed, and the first dynamic compensation driving spectrum is output.

5. The dynamic compensation method based on wind turbine yaw error identification as described in claim 1, characterized in that, The plurality of follower wind turbines use the first dynamic compensation drive spectrum as a reference for compensation drive parameters, and perform low-latency synchronization compensation of the plurality of follower wind turbines based on local wind turbine state snapshots and correction output of multiple compensation drive decision results. The method includes: Retrieve the Wth following yaw error from the local wind turbine status snapshot of the Wth following wind turbine. Calculate the Wth three-dimensional state offset between the real-time yaw error and the Wth following yaw error; Based on the Wth three-dimensional state offset, the first dynamic compensation driving spectrum is corrected, and the Wth compensation driving decision result is output. After performing load safety boundary verification on the Wth compensation-driven decision result, the low-latency synchronization compensation of the Wth following wind turbine is performed using the Wth compensation-driven decision result.

6. A dynamic compensation system based on wind turbine yaw error identification, characterized in that, The system is used to implement the dynamic compensation method based on wind turbine yaw error identification as described in any one of claims 1-5, and the system comprises: A multi-scale compensation benchmark information database construction module is used to backtrack the historical error compensation logs of the wind turbine cluster, perform yaw error operating condition compensation analysis, and construct a multi-scale compensation benchmark information database. The wind turbine cluster includes a leader wind turbine and multiple follower wind turbines. The first dynamic compensation drive spectrum output module is used in the wind turbine cluster to query the multi-scale compensation benchmark information library by integrating the real-time yaw error and real-time operating conditions of the leader wind turbine as operating condition feature vectors, outputting the first dynamic compensation drive spectrum, and making compensation drive decisions for the leader wind turbine. The compensation drive spectrum distribution module is used by the leader wind turbine to distribute the first dynamic compensation drive spectrum to multiple follower wind turbines; The low-latency synchronization compensation module is used to perform low-latency synchronization compensation for the multiple following wind turbines by using the first dynamic compensation drive spectrum as a reference to compensate drive parameters and correcting and outputting multiple compensation drive decision results based on the local wind turbine state snapshot.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the dynamic compensation method based on wind turbine yaw error identification as described in any one of claims 1-5.

Images