Wind turbine generator cooperative control optimization method and device based on time sequence analysis

Abstract

The invention discloses a wind turbine generator cooperative control optimization method and device based on time sequence analysis, and relates to the data processing related field, and the method comprises the steps: carrying out the operation data collection, and constructing a multivariate time sequence analysis model; receiving and analyzing the K operation time sequence data in real time, and positioning a target fault unit; performing risk assessment, and configuring a cooperative control window; power demand prediction is obtained, and power distribution is carried out on the K-1 wind turbine generators; environment data collection is carried out, K-1 real-time wind power environments and K-1 output power constraints are synchronized to the control strategy analysis model, and K-1 short-term control strategies are obtained; and performing fault isolation on the target fault unit, and executing cooperative control optimization by adopting K-1 short-term control strategies. The technical problem that the overall operation efficiency and stability of a wind power plant are poor due to the fact that all wind turbine generators in existing wind turbine generator control lack effective cooperation is solved, and the technical effect of improving the overall operation efficiency and stability of the wind power plant is achieved.

Description

A Method and Device for Cooperative Control Optimization of Wind Turbine Units Based on Time Series Analysis Technical Field

[0001] This application relates to the field of data processing, and in particular to a method and apparatus for optimizing the collaborative control of wind turbine units based on time series analysis. Background Technology

[0002] Wind farms, as a primary form of wind energy utilization, typically consist of multiple wind turbines that work together to convert wind energy into electrical energy to supply the power grid. Traditional wind farm management relies heavily on the independent control of individual wind turbines. When a turbine malfunctions, manual intervention is required for troubleshooting and repair, a time-consuming process that can lead to significant fluctuations in the wind farm's output power and impact grid stability.

[0003] At present, the control of wind turbine units suffers from a lack of effective coordination between the individual wind turbine units, resulting in poor overall operating efficiency and stability of the wind farm. Summary of the Invention

[0004] This application provides a wind turbine collaborative control optimization method and device based on time-series analysis. It employs real-time acquisition of operating data from each wind turbine in a wind farm to construct a multivariate time-series analysis model. Once a target faulty turbine is identified, a risk assessment is immediately conducted and a collaborative control window is configured. Based on the power demand forecast from the grid operator, power is allocated to the remaining wind turbines. Furthermore, short-term control strategies are generated by combining real-time wind power environmental data to optimize the collaborative control of the wind farm. Through these collaborative control strategies, effective coordination among the wind turbines within the wind farm is achieved, resulting in improved overall operating efficiency and stability.

[0005] This application provides a wind turbine collaborative control optimization method based on time-series analysis, comprising: collecting operational data from a wind farm and constructing a multivariate time-series analysis model based on the data collection results, wherein the wind farm includes K wind turbines; the multivariate time-series analysis model receives and analyzes K operational sequence data from the K wind turbines in real time to locate the target faulty turbine; performing a risk assessment on the target faulty turbine and configuring a collaborative control window based on the assessment results; obtaining power demand forecasts from the grid operator and allocating power to K-1 wind turbines based on the power demand forecasts to obtain K-1 output power constraints; collecting environmental data from the K-1 wind turbines and synchronizing the obtained K-1 real-time wind power environment data and K-1 output power constraints to a pre-constructed control strategy analysis model to obtain K-1 short-term control strategies; using the collaborative control window as a constraint, isolating the target faulty turbine and performing collaborative control optimization on the K-1 wind turbines using the K-1 short-term control strategies.

[0006] In a possible implementation, operational data of the wind farm is collected, and a multivariate time-series analysis model is constructed based on the data collection results. The following processes are performed: the wind farm's operational log is retrieved, and K historical operational data streams of the K wind turbines are obtained based on the operational logs; operational features are extracted from the K historical operational data streams to obtain K equipment operational feature data; K multivariate time-series analysis branches are constructed using the K equipment operational feature data; the K multivariate time-series analysis branches are identified by the K wind turbines, and then connected in parallel to complete the construction of the multivariate time-series analysis model.

[0007] In a possible implementation, operational features are extracted from the K historical operational data streams to obtain K equipment operational feature data, and the following processing is performed: A first historical operational data stream of a first wind turbine is retrieved from the K historical operational data streams, wherein the first wind turbine is any one of the K wind turbines; the first historical operational data stream is segmented based on environmental change cycles to obtain multiple periodic operational data streams; the wind farm's operation and maintenance logs are interacted with to obtain a first operation and maintenance sequence; the multiple periodic operational data streams and the first operation and maintenance sequence are aligned, and fault correlation index analysis is performed based on the alignment results to obtain a first multi-dimensional operational feature index; first equipment operational feature data is separated from the first historical operational data stream based on the first multi-dimensional operational feature index; and so on, operational features are extracted from the K historical operational data streams to obtain the K equipment operational feature data.

[0008] In a possible implementation, the multiple periodic operation data streams and the first operation and maintenance sequence are aligned, and fault correlation index analysis is performed based on the alignment results to obtain a first multi-dimensional operation characteristic index. The following processing is performed: the multiple periodic operation data streams and the first operation and maintenance sequence are aligned to obtain multiple sets of fault operation nodes; a fault correlation window and a fault index frequency threshold are preset; data is collected from the multiple periodic operation data at the multiple sets of fault operation nodes based on the fault correlation window to obtain multiple sets of fault data streams; fault index identification and collection are performed based on the multiple sets of fault data streams to obtain multiple sets of fault indicators; after aggregating the multiple sets of fault indicators, the aggregation results are filtered based on the fault index frequency threshold to obtain the first multi-dimensional operation characteristic index.

[0009] In a possible implementation, a risk assessment is performed on the target faulty unit, and a collaborative control window is configured based on the assessment results. The following processes are then performed: the multiple sets of fault data streams are deleted from the first equipment operating characteristic data to obtain first compliant operating characteristic data; the first compliant operating characteristic data is decomposed using the first multi-dimensional operating characteristic index to obtain M compliant operating characteristic arrays, wherein the first multi-dimensional operating characteristic index includes M first related characteristic indicators; extreme value calls are performed based on the M compliant operating characteristic arrays to obtain M compliant operating intervals, and the M compliant operating intervals are used as first compliant operating thresholds; and so on, constructing K compliant operating thresholds for the K wind turbine units; target time-series data of the target faulty unit is called from the K operating time-series data; target compliant operating thresholds of the target faulty unit are called from the K compliant operating thresholds, and the target time-series data and target compliant operating thresholds are quantitatively evaluated based on Euclidean distance to obtain a first operating risk coefficient; the first operating risk coefficient is used to traverse a pre-constructed risk investigation time table to obtain the collaborative control window.

[0010] In a possible implementation, the grid operator obtains a power demand forecast and allocates power to K-1 wind turbines based on the power demand forecast to obtain K-1 output power constraints. The following processing is then performed: variance calculation is performed on K compliant operating thresholds to obtain K operating variances; K-1 operating variances of the K-1 wind turbines are extracted from the K operating variances, and K-1 power allocation weights are calculated based on the K-1 operating variances; power is allocated to the K-1 wind turbines according to the K-1 power allocation weights and the power demand forecast to obtain the K-1 output power constraints.

[0011] In a possible implementation, environmental data is collected from the K-1 wind turbines, and the obtained K-1 real-time wind power environment data and K-1 output power constraints are synchronized to a pre-constructed control strategy analysis model to obtain K-1 short-term control strategies. The following processing is then performed: interacting with the wind farm to obtain the first unit control log of the first wind turbine; using the multiple sets of faulty operating nodes to remove invalid data from the first unit control log to obtain a first optimized control log, wherein the first optimized control log includes multiple sample environment arrays, multiple sample control arrays, and multiple sample power outputs; using the first optimized control log as training data to construct a first control strategy analysis branch; and so on, constructing K control strategy analysis branches for the K wind turbines, and obtaining the control strategy analysis model by connecting the K control strategy analysis branches in parallel; after setting the control strategy analysis branch of the target faulty unit as an analysis taboo, the K-1 real-time wind power environment data and K-1 output power constraints are synchronized to the control strategy analysis model to obtain the K-1 short-term control strategies.

[0012] This application also provides a wind turbine collaborative control optimization device based on time-series analysis, comprising: a multivariate time-series analysis model construction module, which is used to collect operational data from a wind farm and construct a multivariate time-series analysis model based on the data collection results, wherein the wind farm includes K wind turbines; a target faulty turbine location module, which is used by the multivariate time-series analysis model to receive and analyze K operational sequence data from the K wind turbines in real time to locate the target faulty turbine; a collaborative control window configuration module, which is used to perform risk assessment on the target faulty turbine and configure a collaborative control window based on the assessment results; and a power allocation module, which... The power allocation module is used to obtain power demand forecasts from the grid operator and allocate power to K-1 wind turbines based on the power demand forecasts to obtain K-1 output power constraints. The short-term control strategy acquisition module is used to collect environmental data from the K-1 wind turbines and synchronize the obtained K-1 real-time wind power environment data and K-1 output power constraints to a pre-built control strategy analysis model to obtain K-1 short-term control strategies. The collaborative control optimization module is used to isolate the target faulty turbines using the collaborative control window as a constraint and to perform collaborative control optimization on the K-1 wind turbines using the K-1 short-term control strategies.

[0013] The proposed wind turbine collaborative control optimization method and device based on time-series analysis first collects operational data from the wind farm and constructs a multivariate time-series analysis model based on the data collection results. The wind farm includes K wind turbines. The multivariate time-series analysis model then receives and analyzes K operational sequence data from the K wind turbines in real time to locate the target faulty turbine. A risk assessment is then performed on the target faulty turbine, and a collaborative control window is configured based on the assessment results. Next, power demand forecasts are obtained from the grid operator, and power allocation is performed on K-1 wind turbines based on these forecasts, resulting in K-1 output power constraints. Then, environmental data is collected from the K-1 wind turbines, and the obtained K-1 real-time wind power environment data and K-1 output power constraints are synchronized to a pre-constructed control strategy analysis model to obtain K-1 short-term control strategies. Finally, using the collaborative control window as a constraint, the target faulty turbine is isolated, and the K-1 short-term control strategies are used to perform collaborative control optimization on the K-1 wind turbines, achieving the technical effect of improving the overall operating efficiency and stability of the wind farm. Attached Figure Description

[0014] 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 apparatus 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.

[0015] Figure 1 is a flowchart illustrating the wind turbine collaborative control optimization method based on time-series analysis provided in an embodiment of this application.

[0016] Figure 2 is a schematic diagram of the structure of the wind turbine collaborative control optimization device based on time analysis provided in the embodiment of this application.

[0017] Figure labeling: Multivariate time series analysis model construction module 10, target fault unit location module 20, cooperative control window configuration module 30, power allocation module 40, short-term control strategy acquisition module 50, cooperative control optimization module 60. Detailed Implementation

[0018] The above description is only 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, the following are specific embodiments of this application.

[0019] 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.

[0020] 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, apparatus, 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 apparatuses. 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.

[0021] This application provides a wind turbine collaborative control optimization method based on time series analysis, as shown in Figure 1. The method includes: Step S100, collecting operational data from the wind farm and constructing a multivariate time series analysis model based on the data collection results, wherein the wind farm includes K wind turbines. Specifically, a sensor network (such as anemometers, temperature sensors, vibration sensors, etc.) is used to collect real-time operational data from each of the K wind turbines in the wind farm, including but not limited to wind speed, wind direction, generator speed, power output, and temperature. The collected raw data is cleaned (noise and outliers are removed), transformed (e.g., wind speed is converted from meters per second to standard units), and standardized to ensure data consistency and accuracy. Based on the preprocessed data, a multivariate time series analysis model that reflects the interaction between wind turbines and environmental impacts is constructed using time series analysis methods. That is, the multivariate time series analysis model can simultaneously process operational data from multiple wind turbines and perform rapid analysis to locate the target faulty turbine.

[0022] In one possible implementation, operational data of the wind farm is collected, and a multivariate time-series analysis model is constructed based on the data collection results. Step S100 further includes step S110, which involves calling the wind farm's operational logs and obtaining K historical operational data streams for the K wind turbine units based on the operational logs. Specifically, the wind farm's operational log system is accessed, and these logs record historical data such as the operating status, faults, and maintenance of wind farm equipment. Historical operational data related to the K wind turbine units is filtered from the operational logs, including information from multiple dimensions such as timestamps, wind speed, wind direction, generator speed, and output power. The filtered data is classified according to the wind turbine unit's identifier (e.g., ID) to form K independent historical operational data streams, where each historical operational data stream is a data set reflecting the past operating status of the wind turbine units, arranged in chronological order. Step S120 involves extracting operational features from the K historical operational data streams to obtain K equipment operational feature data. Specifically, based on the characteristics of wind farm operation, representative features of wind turbine operating status are selected, such as average wind speed, power fluctuation rate, and failure rate. Through data aggregation (e.g., calculating average and standard deviation) and data smoothing (noise removal), the original historical operating data stream is converted into a dataset containing the selected features. Equipment operating feature data is then extracted from the converted dataset. Step S130: K multivariate time-series analysis branches are constructed using the K equipment operating feature data. Specifically, an independent multivariate time-series analysis branch is initialized for each wind turbine, and these branches are used to analyze the equipment operating feature data of each wind turbine individually. Using time series analysis methods, the corresponding multivariate time-series analysis branches are trained using the equipment operating feature data of each wind turbine, enabling them to learn and predict the future operating status of the wind turbine. Step S140: After identifying the K multivariate time-series analysis branches using the K wind turbines, the K multivariate time-series analysis branches are connected in parallel to complete the construction of the multivariate time-series analysis model. Specifically, each multivariate time-series analysis branch is assigned a unique identifier, which is associated with the corresponding wind turbine. By concatenating K pre-trained multivariate time-series analysis branches in parallel, a unified multivariate time-series analysis model is formed. This model can simultaneously process time-series data from K wind turbines. This implementation method constructs the multivariate time-series analysis model by calling operational logs and historical operational data streams, making full use of the existing data resources of the wind farm. Furthermore, by constructing an independent multivariate time-series analysis branch for each wind turbine, it considers the individual differences among wind turbines in the wind farm, accurately capturing the operational characteristics and time-series features of each wind turbine, thus achieving the technical effect of improving the accuracy and reliability of the model.

[0023] In one possible implementation, operational features are extracted from the K historical operational data streams to obtain K equipment operational feature data. Step S120 further includes step S121, which involves calling the first historical operational data stream of a first wind turbine from the K historical operational data streams, wherein the first wind turbine is any one of the K wind turbines. Specifically, a processing object is arbitrarily selected from the K wind turbines as the first wind turbine. From the dataset storing the K historical operational data streams, the corresponding historical operational data stream is retrieved and extracted based on the identifier (e.g., ID) of the first wind turbine. Step S122 involves segmenting the first historical operational data stream based on the environmental change cycle to obtain multiple periodic operational data streams. Specifically, environmental parameters (e.g., wind speed, temperature, humidity, etc.) in the first historical operational data stream are analyzed to identify the periodic patterns of environmental changes (the patterns of natural environmental conditions changing periodically over time). Based on the identified environmental change cycle, the first historical operational data stream is segmented into multiple time periods, each time period corresponding to a complete environmental change cycle, resulting in periodic operational data streams corresponding to each time period. Step S123: Interact with the wind farm's operation and maintenance logs to obtain a first operation and maintenance sequence. Specifically, access the wind farm's operation and maintenance log system to obtain operation and maintenance records related to the first wind turbine. Extract operation and maintenance events and operation records that match the time range of the first historical operation data stream from the operation and maintenance logs (a historical data set recording wind farm equipment maintenance, repair, fault handling, and other operation and maintenance activities) to form the first operation and maintenance sequence. The first operation and maintenance sequence is a data sequence reflecting the operation and maintenance activities of the first wind turbine. Step S124: Align the multiple periodic operation data streams with the first operation and maintenance sequence, and perform fault correlation index analysis based on the alignment results to obtain a first multivariate operation characteristic index. Specifically, align the timestamps of multiple periodic operation data streams with the first operation and maintenance sequence to ensure they are compared within the same time frame. Analyze the aligned data to identify potential correlations between operation and maintenance events and operation data, i.e., the changing relationship between fault events and operation characteristic parameters. Based on the fault correlation analysis results, multivariate operational characteristic indicators that reflect the operating status and potential fault risks of the first wind turbine are extracted. Specifically, the first multivariate operational characteristic indicators are a set of characteristic indicators that reflect the correlation between the operating characteristics of the first wind turbine and potential faults. These indicators reveal the changing patterns of certain specific operating parameters before and after a fault occurs, or the mutual influence and correlation between different operating parameters. Step S125: Based on the first multivariate operational characteristic indicators, the operating characteristic data of the first equipment is separated from the first historical operating data stream.Specifically, data directly related to the first multi-dimensional operational characteristic index is selected from the first historical operational data stream to form the first equipment operational characteristic data. This first equipment operational characteristic data is derived through fault correlation analysis and can more accurately reflect the operating status and potential problems of the first wind turbine unit, exhibiting high sensitivity and specificity. Step S126, and so on, extracts operational characteristics from the K historical operational data streams to obtain the K equipment operational characteristic data. Specifically, steps S121 to S125 are repeated, but each time the processing object is the next unit among the K wind turbine units, until all historical operational data streams of the K wind turbine units have been processed. This implementation method, by combining environmental change cycles and maintenance log information, more accurately identifies the correlation between operational data and fault events, thereby achieving the technical effect of improving the accuracy of equipment operational characteristic extraction.

[0024] In one possible implementation, data alignment is performed on the multiple periodic operation data streams and the first operation and maintenance sequence, and fault correlation index analysis is performed based on the alignment results to obtain a first multi-dimensional operation characteristic index. Step S124 further includes step S1241, which involves aligning the multiple periodic operation data streams and the first operation and maintenance sequence to obtain multiple sets of fault operation nodes. Specifically, the timestamp of each data point in the multiple periodic operation data streams is matched with the timestamp in the first operation and maintenance sequence to ensure that they are within the same time frame. In the operation and maintenance sequence, fault-related events (such as maintenance records, fault reports, etc.) are identified, and the time points corresponding to these events are recorded as fault operation nodes. The data segments before and after the corresponding fault operation nodes in the periodic operation data streams are marked to form a set of data segments associated with operation and maintenance fault events, i.e., multiple sets of fault operation nodes. Step S1242 involves presetting a fault correlation window and a fault index frequency threshold. Specifically, based on experience, a fault correlation time window (such as a few hours or days before and after a fault occurs) is preset to collect data associated with fault events in the periodic operation data streams. A fault indicator frequency threshold is set to filter fault indicators that appear frequently enough in subsequent analysis. Step S1243: Based on the fault association window, data is collected from multiple sets of periodic operating data at multiple fault operating nodes to obtain multiple sets of fault data streams. Specifically, for each fault operating node, data within the corresponding time range before and after that node is collected from the periodic operating data stream according to the preset fault association window. The collected data is grouped according to the fault operating node to form multiple sets of periodic operating data stream segments associated with specific fault events. Step S1244: Fault indicator identification and collection are performed based on the multiple sets of fault data streams to obtain multiple sets of fault indicators. Specifically, feature parameters (such as temperature, vibration, power, etc.) that may reflect fault characteristics are extracted from each set of fault data streams. Based on the extracted feature parameters, combined with domain knowledge and experience, fault indicators that can describe fault characteristics are generated, such as the temperature difference before and after the fault occurs, vibration amplitude, etc. Step S1245: After aggregating the multiple sets of fault indicators, the aggregation results are filtered based on the fault indicator frequency threshold to obtain the first multi-dimensional operating characteristic indicator. Specifically, all fault indicators are summarized and organized to form a comprehensive set of fault indicators. Based on a preset fault indicator frequency threshold, fault indicators that occur most frequently in all fault events are selected. These selected high-frequency fault indicators are used as the first multi-dimensional operational characteristic indicators, which can comprehensively reflect the operational characteristics and correlations of the first wind turbine unit under fault conditions.This approach ensures that the collected data is closely related to the fault events through data alignment and the setting of fault correlation windows, reducing interference from irrelevant data. By filtering through preset fault index frequency thresholds, it further highlights those representative fault indicators that frequently appear in all fault events, thus achieving the technical effect of improving the accuracy and reliability of fault correlation index analysis.

[0025] In step S200, the multivariate time-series analysis model receives and analyzes K operational sequence data from the K wind turbines in real time to locate the target faulty turbine. Specifically, the multivariate time-series analysis model receives operational sequence data from each wind turbine in real time, and by comparing the current data with historical data, identifies abnormal or deviating data points to locate the wind turbine that has experienced a fault or anomaly.

[0026] Step S300 involves conducting a risk assessment of the target faulty turbine unit and configuring a collaborative control window based on the assessment results. Specifically, based on the fault type, severity, and historical data, the impact of the fault on the overall operation of the wind farm is assessed, including power generation loss and maintenance costs. Based on the assessment results, a time window is determined during the fault handling period when other non-faulty turbine units need to work collaboratively to ensure the stability and efficiency of the overall wind farm operation. That is, the collaborative control window refers to the period during which non-faulty turbine units in the wind farm are collaboratively controlled to address the turbine fault.

[0027] In one possible implementation, a risk assessment is performed on the target faulty unit, and a collaborative control window is configured based on the assessment results. Step S300 further includes step S310, deleting the multiple sets of fault data streams from the first equipment operating characteristic data to obtain first compliant operating characteristic data. Specifically, multiple sets of fault data streams associated with the fault are identified and deleted from the first equipment operating characteristic data, and the remaining data is the first compliant operating characteristic data, which reflects the operating status of the first wind turbine unit within the normal range. Step S320, the first compliant operating characteristic data is decomposed using the first multi-dimensional operating characteristic index to obtain M compliant operating characteristic arrays, wherein the first multi-dimensional operating characteristic index includes M first associated characteristic indicators. Specifically, the first compliant operating characteristic data is decomposed using the first multi-dimensional operating characteristic index (a multi-dimensional set of indicators used to describe the operating status of the first wind turbine unit, containing M first associated characteristic indicators), and the decomposed data is organized into M compliant operating characteristic arrays, each array corresponding to one associated characteristic indicator. Step S330: Based on the M compliant operation feature arrays, extreme value calls are performed to obtain M compliant operation intervals, and these M compliant operation intervals are used as the first compliant operation threshold. Specifically, maximum and minimum value calls are performed on each compliant operation feature array to determine the normal operating range of the feature, and these ranges are used as the M compliant operation intervals. That is, the compliant operation interval refers to the allowable range of variation of a certain feature indicator under normal operating conditions of the first wind turbine. The M compliant operation intervals together constitute the first compliant operation threshold. Step S340: By analogy, the K compliant operation thresholds of the K wind turbines are constructed. Specifically, steps S310 to S330 are repeated to construct the compliant operation threshold for each wind turbine in the wind farm. Step S350: The target time-series data of the target faulty turbine is retrieved from the K runtime sequence data. Specifically, the target time-series data of the target faulty turbine is retrieved from the K runtime sequence data. Step S360: The target compliance operation threshold of the target faulty unit is retrieved from the K compliance operation thresholds, and the target time-series data and the target compliance operation threshold are quantitatively evaluated based on Euclidean distance to obtain a first operation risk coefficient. Specifically, the difference between the target time-series data and the target compliance operation threshold is quantitatively evaluated using Euclidean distance, and the first operation risk coefficient is obtained based on the magnitude of the difference. That is, the first operation risk coefficient is an indicator that quantitatively represents the degree to which the operating state of the target faulty unit deviates from the normal range. Step S370: The first operation risk coefficient is used to traverse a pre-constructed risk investigation time table to obtain the collaborative control window. Specifically, the risk investigation time table is constructed based on historical data and experience, and is used to estimate the time required for fault investigation under different risk levels.The system iterates through a pre-built risk assessment time table using the first operational risk coefficient. Based on the results, it determines the size and duration of the collaborative control window, which is used to schedule and execute collaborative control strategies to ensure the overall stable operation of the wind farm during fault handling. This approach constructs compliance operation thresholds for all wind turbine units and calls upon data from the target faulty unit. It then quantifies the distance between the target time-series data and the compliance operation thresholds, and configures the size and duration of the collaborative control window accordingly, based on the risk assessment time table. This achieves a significant improvement in the accuracy of the collaborative control window configuration.

[0028] In step S400, the grid operator obtains the power demand forecast and allocates power to K-1 wind turbine units based on the power demand forecast, thereby obtaining K-1 output power constraints. Specifically, a connection is established with the grid operator to obtain the grid power demand forecast for a future period. The output power target value of each unit is calculated based on the power demand forecast and the performance parameters (such as rated power and current status) of each non-faulty wind turbine unit.

[0029] In one possible implementation, the grid operator obtains power demand forecasts and allocates power to K-1 wind turbines based on these forecasts, obtaining K-1 output power constraints. Step S400 further includes step S410, calculating the variance of the K compliant operating thresholds to obtain K operating variances. Specifically, the K compliant operating thresholds represent the range of parameter variations that allow wind turbines to operate stably under different operating conditions. For each wind turbine's compliant operating threshold, the variance of its data distribution is calculated. The variance reflects the variable range of the wind turbine's operating parameters, i.e., the turbine's fault tolerance or stability, resulting in K operating variances, each corresponding to one wind turbine. Step S420 extracts the K-1 operating variances of the K-1 wind turbines from the K operating variances and calculates K-1 power allocation weights based on these K-1 operating variances. Specifically, since the target faulty turbine does not participate in power allocation, the corresponding variance value is removed from the K operating variances. Based on the remaining K-1 operating variances, a power allocation weight is calculated for each wind turbine. The larger the variance (higher fault tolerance), the larger the allocated weight, meaning the corresponding wind turbine will undertake more power allocation tasks. Step S430: Power allocation is performed on the K-1 wind turbines according to the K-1 power allocation weights and power demand forecast, obtaining the K-1 output power constraints. Specifically, based on the power allocation weights obtained in step S420 and the power demand forecast from the grid operator, a proportional allocation method is used to distribute power demand to the K-1 wind turbines, generating an output power constraint for each wind turbine, i.e., the range of output power that the wind turbine should maintain for a given period. This implementation allocates more power to wind turbines with higher fault tolerance, fully utilizing their potential while reducing reliance on wind turbines with lower fault tolerance, thus improving the reliability of power allocation.

[0030] Step S500: Environmental data is collected from the K-1 wind turbine units, and the obtained K-1 real-time wind power environment data and K-1 output power constraints are synchronized to the pre-built control strategy analysis model to obtain K-1 short-term control strategies. Specifically, environmental data such as wind speed, wind direction, and temperature at the location of each non-faulty unit are collected in real time. The real-time wind power environment data and output power constraints are input into the pre-built control strategy analysis model. Based on the input data, combined with the wind turbine unit performance parameters and current operating status, the model generates the optimal control strategy for each wind turbine unit, i.e., the short-term control strategy.

[0031] In one possible implementation, environmental data is collected from the K-1 wind turbine units, and the obtained K-1 real-time wind power environment data and K-1 output power constraints are synchronized to a pre-constructed control strategy analysis model to obtain K-1 short-term control strategies. Step S500 further includes step S510, interacting with the wind farm to obtain the first unit control log of the first wind turbine unit. Specifically, the control log of the first wind turbine unit is obtained through the wind farm's data management system. This log contains various parameter records during the operation of the first wind turbine unit, such as environmental parameters (wind speed, wind direction, temperature, etc.), control commands, output power, etc. Step S520, invalid data is removed from the first unit control log using the multiple sets of faulty operating nodes to obtain a first optimized control log, wherein the first optimized control log includes multiple sample environment arrays, multiple sample control arrays, and multiple sample power outputs. Specifically, abnormal data in the control log of the first wind turbine is identified using multiple sets of faulty operating nodes. The identified abnormal data is then removed from the control log to obtain an optimized control log, which contains more accurate and effective operating information for the first wind turbine. The first optimized control log includes multiple sample environment arrays (representing different environmental conditions), multiple sample control arrays (representing corresponding control commands), and multiple sample power outputs (representing actual output power). Step S530: The first optimized control log is used as training data to construct a first control strategy analysis branch. Specifically, using the first optimized control log as training data, a control strategy analysis branch is trained using a machine learning algorithm. This branch can generate optimal control commands based on input environmental parameters and output power constraints. Step S540: Similarly, K control strategy analysis branches are constructed for the K wind turbines, and the control strategy analysis model is obtained by connecting the K control strategy analysis branches in parallel. Specifically, steps S510 to S530 are repeated for the remaining K-1 wind turbines to construct an independent control strategy analysis branch for each wind turbine. All K control strategy analysis branches are connected in parallel to form a complete control strategy analysis model. This model can simultaneously process data from multiple wind turbine units and generate corresponding control strategies. In step S550, after setting the control strategy analysis branch of the target faulty turbine unit as an analysis taboo, the K-1 real-time wind power environments and K-1 output power constraints are synchronized to the control strategy analysis model to obtain the K-1 short-term control strategies. Specifically, the control strategy analysis branch corresponding to the target faulty turbine unit is set as an analysis taboo, meaning it is not used for strategy generation in this step. The K-1 real-time wind power environments and output power constraints are synchronized to the control strategy analysis model, and the model generates short-term control strategies for each non-faulty turbine unit based on the input data and constraints.This implementation method uses the first optimized control log to train the first control strategy analysis branch. The first optimized control log contains the operating experience and knowledge of the wind turbine over a period of time. Since the operating conditions of the wind farm (such as wind speed, wind direction, temperature, etc.) are constantly changing, the model trained by historical data can learn the impact of these changes on the operation of the wind turbine, thereby generating a control strategy that is more adapted to the current conditions, achieving the technical effect of improving the accuracy of short-term control strategy generation.

[0032] Step S600 involves isolating the target faulty turbine unit under the constraint of the collaborative control window, and performing collaborative control optimization on the K-1 wind turbine units using the K-1 short-term control strategies. Specifically, within the collaborative control window, the target faulty turbine unit is physically and logically isolated to prevent it from further affecting the wind farm operation. Based on the generated short-term control strategies, the operating states of each non-faulty turbine unit (such as blade angle and generator speed) are adjusted to ensure that the overall output power of the wind farm meets the power demand forecast. This embodiment of the application uses real-time acquisition of operating data from each wind turbine unit in the wind farm to construct a multivariate time-series analysis model. Once the target faulty turbine unit is identified, a risk assessment is immediately performed and a collaborative control window is configured. Based on the power demand forecast of the grid operator, power is allocated to the remaining wind turbine units. Short-term control strategies are generated in conjunction with real-time wind power environment data, and collaborative control optimization of the wind farm is performed. Through these collaborative control strategies, effective coordination between the wind turbine units within the wind farm is achieved, resulting in improved overall operating efficiency and stability of the wind farm.

[0033] In the preceding text, a wind turbine cooperative control optimization method based on time-series analysis according to an embodiment of the present invention was described in detail with reference to FIG1. ​​Next, a wind turbine cooperative control optimization apparatus based on time-series analysis according to an embodiment of the present invention will be described with reference to FIG2.

[0034] The wind turbine collaborative control optimization device based on time-series analysis according to embodiments of the present invention is used to solve the technical problem of poor overall operating efficiency and stability of wind farms due to the lack of effective coordination between wind turbines in existing wind turbine control systems, thereby improving the overall operating efficiency and stability of wind farms. The wind turbine collaborative control optimization device based on time-series analysis includes: a multivariate time-series analysis model construction module 10, a target fault turbine location module 20, a collaborative control window configuration module 30, a power allocation module 40, a short-term control strategy acquisition module 50, and a collaborative control optimization module 60.

[0035] The multivariate time-series analysis model construction module 10 is used to collect operational data from the wind farm and construct a multivariate time-series analysis model based on the data collection results. The wind farm includes K wind turbine units. The target faulty unit location module 20 is used by the multivariate time-series analysis model to receive and analyze K operational sequence data from the K wind turbine units in real time to locate the target faulty unit. The collaborative control window configuration module 30 is used to perform risk assessment on the target faulty unit and configure a collaborative control window based on the assessment results. The power allocation module 40 is used to interact with the grid operator to obtain power demand forecasts and, based on... The power demand prediction allocates power to K-1 wind turbine units to obtain K-1 output power constraints; the short-term control strategy acquisition module 50 collects environmental data from the K-1 wind turbine units and synchronizes the obtained K-1 real-time wind power environment data and K-1 output power constraints to the pre-built control strategy analysis model to obtain K-1 short-term control strategies; the collaborative control optimization module 60 uses the collaborative control window as a constraint to isolate the target faulty unit and performs collaborative control optimization on the K-1 wind turbine units using the K-1 short-term control strategies.

[0036] The specific configuration of the multivariate time-series analysis model construction module 10 will be described in detail below. As mentioned above, the module collects operational data from the wind farm and constructs a multivariate time-series analysis model based on the data collection results. The multivariate time-series analysis model construction module 10 may further include: a K-historical operational data stream retrieval unit for retrieving the operational logs of the wind farm and obtaining K-historical operational data streams of the K wind turbine units based on the operational logs; an operational feature extraction unit for extracting operational features from the K-historical operational data streams to obtain K-device operational feature data; a K-multivariate time-series analysis branch acquisition unit for constructing K-multivariate time-series analysis branches using the K-device operational feature data; and a branch paralleling unit for identifying the K-multivariate time-series analysis branches using the K wind turbine units and then paralleling the K-multivariate time-series analysis branches to complete the construction of the multivariate time-series analysis model.

[0037] The process involves extracting operational features from the K historical operational data streams to obtain K equipment operational feature data. The operational feature extraction unit may further include: a first historical operational data stream invocation subunit for invoking the first historical operational data stream of a first wind turbine from the K historical operational data streams, wherein the first wind turbine is any one of the K wind turbines; a first historical operational data stream segmentation subunit for segmenting the first historical operational data stream based on environmental change cycles to obtain multiple periodic operational data streams; a first maintenance sequence acquisition subunit for interacting with the wind farm's maintenance logs to obtain a first maintenance sequence; a fault correlation index analysis subunit for aligning the multiple periodic operational data streams and the first maintenance sequence, and performing fault correlation index analysis based on the alignment results to obtain a first multi-dimensional operational feature index; a first equipment operational feature data separation subunit for separating the first equipment operational feature data from the first historical operational data stream based on the first multi-dimensional operational feature index; and a K equipment operational feature data acquisition subunit for extracting operational features from the K historical operational data streams to obtain the K equipment operational feature data.

[0038] The process involves aligning the multiple periodic operational data streams and the first maintenance sequence, and performing fault correlation index analysis based on the alignment results to obtain a first multivariate operational characteristic index. The fault correlation index analysis subunit may further include: a data alignment microunit for aligning the multiple periodic operational data streams and the first maintenance sequence to obtain multiple sets of faulty operational nodes; a fault correlation window and fault index frequency threshold setting microunit for presetting the fault correlation window and fault index frequency threshold; a data acquisition microunit for acquiring data from the multiple sets of faulty operational nodes based on the fault correlation window to obtain multiple sets of faulty data streams; a fault index identification and acquisition microunit for identifying and acquiring fault indicators based on the multiple sets of faulty data streams to obtain multiple sets of fault indicators; and an aggregation result filtering microunit for filtering the aggregation results based on the fault index frequency threshold after aggregating the multiple sets of fault indicators to obtain the first multivariate operational characteristic index.

[0039] The specific configuration of the collaborative control window configuration module 30 will be described in detail below. As mentioned above, a risk assessment is performed on the target faulty unit, and a collaborative control window is configured based on the assessment results. The collaborative control window configuration module 30 may further include: a first compliant operation feature data acquisition unit for deleting the multiple sets of fault data streams from the first equipment operation feature data to obtain first compliant operation feature data; an M compliant operation feature array acquisition unit for decomposing the first compliant operation feature data using the first multi-dimensional operation feature index to obtain M compliant operation feature arrays, wherein the first multi-dimensional operation feature index includes M first related feature indicators; and a first compliant operation threshold acquisition unit for performing extreme value calls based on the M compliant operation feature arrays to obtain... M compliant operation intervals are defined, and these M compliant operation intervals are used as the first compliant operation thresholds. A K compliant operation threshold construction unit is used to construct the K compliant operation thresholds for the K wind turbine units. A target time series data retrieval unit is used to retrieve the target time series data of the target faulty unit from the K operating time series data. A first operation risk coefficient acquisition unit is used to retrieve the target compliant operation threshold of the target faulty unit from the K compliant operation thresholds, and quantitatively evaluate the target time series data and the target compliant operation thresholds based on Euclidean distance to obtain the first operation risk coefficient. A collaborative control window acquisition unit is used to traverse the pre-constructed risk investigation time table using the first operation risk coefficient to obtain the collaborative control window.

[0040] The specific configuration of the power allocation module 40 will be described in detail below. As mentioned above, the grid operator obtains power demand forecasts and allocates power to K-1 wind turbine units based on the power demand forecasts to obtain K-1 output power constraints. The power allocation module 40 may further include: a variance calculation unit for calculating the variance of K compliant operating thresholds to obtain K operating variances; a K-1 power allocation weight calculation unit for extracting the K-1 operating variances of the K-1 wind turbine units from the K operating variances and calculating K-1 power allocation weights based on the K-1 operating variances; and a K-1 output power constraint acquisition unit for allocating power to the K-1 wind turbine units according to the K-1 power allocation weights and the power demand forecasts to obtain the K-1 output power constraints.

[0041] The specific configuration of the short-term control strategy acquisition module 50 will be described in detail below. As mentioned above, environmental data is collected from the K-1 wind turbines, and the obtained K-1 real-time wind power environments and K-1 output power constraints are synchronized to the pre-built control strategy analysis model to obtain K-1 short-term control strategies. The short-term control strategy acquisition module 50 may further include: a first unit control log acquisition unit for interacting with the wind farm to obtain the first unit control log of the first wind turbine; and an invalid data removal unit for removing invalid data from the first unit control log using the multiple sets of faulty operating nodes to obtain a first optimized control log, wherein the first optimized control log includes multiple sample environment arrays and multiple sample control... The system includes a control array and multiple sample power outputs; a first control strategy analysis branch construction unit is used to construct a first control strategy analysis branch using the first optimized control log as training data; a control strategy analysis model acquisition unit is used to construct K control strategy analysis branches for the K wind turbines and obtain the control strategy analysis model by connecting the K control strategy analysis branches in parallel; and a K-1 short-term control strategy acquisition unit is used to set the control strategy analysis branch of the target faulty unit as an analysis taboo, and then synchronize the K-1 real-time wind power environment and K-1 output power constraints to the control strategy analysis model to obtain the K-1 short-term control strategies.

[0042] The wind turbine collaborative control optimization device based on time-series analysis provided in this embodiment of the invention can execute the wind turbine collaborative control optimization method based on time-series analysis provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0043] Although this application makes various references to certain modules in the apparatus 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 intended to limit the scope of protection of this invention.

[0044] 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 wind turbine collaborative control optimization method based on time series analysis, characterized in that, The method includes: collecting operational data from a wind farm and constructing a multivariate time-series analysis model based on the data collection results, wherein the wind farm includes K wind turbines; the multivariate time-series analysis model receives and analyzes K operational sequence data from the K wind turbines in real time to locate the target faulty turbine; performing a risk assessment on the target faulty turbine and configuring a collaborative control window based on the assessment results; obtaining power demand forecasts from the grid operator and allocating power to K-1 wind turbines based on the power demand forecasts to obtain K-1 output power constraints; collecting environmental data from the K-1 wind turbines and synchronizing the obtained K-1 real-time wind power environment data and K-1 output power constraints to a pre-constructed control strategy analysis model to obtain K-1 short-term control strategies; using the collaborative control window as a constraint, isolating the target faulty turbine and performing collaborative control optimization on the K-1 wind turbines using the K-1 short-term control strategies.

2. The wind turbine collaborative control optimization method based on time series analysis as described in claim 1, characterized in that, The method involves collecting operational data from a wind farm and constructing a multivariate time-series analysis model based on the collected data. The method further includes: accessing the wind farm's operational logs and obtaining K historical operational data streams for K wind turbine generators based on the logs; extracting operational features from the K historical operational data streams to obtain K equipment operational feature data; constructing K multivariate time-series analysis branches using the K equipment operational feature data; identifying the K multivariate time-series analysis branches using the K wind turbine generators, and then connecting the K multivariate time-series analysis branches in parallel to complete the construction of the multivariate time-series analysis model.

3. The wind turbine collaborative control optimization method based on time series analysis as described in claim 2, characterized in that, The method further includes: extracting operational features from the K historical operational data streams to obtain K equipment operational feature data; retrieving a first historical operational data stream of a first wind turbine from the K historical operational data streams, wherein the first wind turbine is any one of the K wind turbines; segmenting the first historical operational data streams based on environmental change cycles to obtain multiple periodic operational data streams; interacting with the wind farm's operation and maintenance logs to obtain a first operation and maintenance sequence; aligning the multiple periodic operational data streams and the first operation and maintenance sequence, and performing fault correlation index analysis based on the alignment results to obtain a first multi-dimensional operational feature index; separating first equipment operational feature data from the first historical operational data streams based on the first multi-dimensional operational feature index; and so on, extracting operational features from the K historical operational data streams to obtain the K equipment operational feature data.

4. The wind turbine collaborative control optimization method based on time series analysis as described in claim 3, characterized in that, The method further includes: aligning the multiple periodic operational data streams and the first maintenance sequence to obtain multiple sets of faulty operational nodes; presetting a fault association window and a fault indicator frequency threshold; collecting data from the multiple periodic operational data streams at the multiple sets of faulty operational nodes based on the fault association window to obtain multiple sets of faulty data streams; identifying and collecting fault indicators based on the multiple sets of faulty data streams to obtain multiple sets of fault indicators; and after aggregating the multiple sets of fault indicators, filtering the aggregation results based on the fault indicator frequency threshold to obtain the first multi-dimensional operational characteristic indicator.

5. The wind turbine collaborative control optimization method based on time series analysis as described in claim 4, characterized in that, The method further includes: performing a risk assessment on the target faulty unit and configuring a collaborative control window based on the assessment results; deleting the multiple sets of fault data streams from the first equipment operating characteristic data to obtain first compliant operating characteristic data; decomposing the first compliant operating characteristic data using the first multi-dimensional operating characteristic index to obtain M compliant operating characteristic arrays, wherein the first multi-dimensional operating characteristic index includes M first related characteristic indicators; performing extreme value calls based on the M compliant operating characteristic arrays to obtain M compliant operating intervals, and using the M compliant operating intervals as first compliant operating thresholds; and so on, constructing K compliant operating thresholds for the K wind turbine units; calling the target time-series data of the target faulty unit from the K operating time-series data; calling the target compliant operating threshold of the target faulty unit from the K compliant operating thresholds, and quantifying and evaluating the target time-series data and the target compliant operating threshold based on Euclidean distance to obtain a first operating risk coefficient; and using the first operating risk coefficient to traverse a pre-constructed risk investigation time table to obtain the collaborative control window.

6. The wind turbine collaborative control optimization method based on time series analysis as described in claim 5, characterized in that, The interactive grid operator obtains power demand forecasts and allocates power to K-1 wind turbine units based on these forecasts to obtain K-1 output power constraints. The method further includes: calculating the variance of K compliant operating thresholds to obtain K operating variances; extracting K-1 operating variances from the K operating variances and calculating K-1 power allocation weights based on these K-1 operating variances; and allocating power to the K-1 wind turbine units according to the K-1 power allocation weights and the power demand forecasts to obtain the K-1 output power constraints.

7. The wind turbine collaborative control optimization method based on time series analysis as described in claim 4, characterized in that, The method involves collecting environmental data from the K-1 wind turbine units and synchronizing the obtained K-1 real-time wind power environment data and K-1 output power constraints to a pre-constructed control strategy analysis model to obtain K-1 short-term control strategies. The method further includes: interacting with the wind farm to obtain the first unit control log of the first wind turbine unit; using the multiple sets of faulty operating nodes to remove invalid data from the first unit control log to obtain a first optimized control log, wherein the first optimized control log includes multiple sample environment arrays, multiple sample control arrays, and multiple sample power outputs; using the first optimized control log as training data to construct a first control strategy analysis branch; and so on, constructing K control strategy analysis branches for the K wind turbine units, and obtaining the control strategy analysis model by connecting the K control strategy analysis branches in parallel; setting the control strategy analysis branch of the target faulty unit as an analysis taboo, and then synchronizing the K-1 real-time wind power environment data and K-1 output power constraints to the control strategy analysis model to obtain the K-1 short-term control strategies.

8. A wind turbine collaborative control optimization device based on time series analysis, characterized in that, The apparatus is used to implement the wind turbine collaborative control optimization method based on time-series analysis as described in any one of claims 1-7. The apparatus includes: a multivariate time-series analysis model construction module, which collects operational data from the wind farm and constructs a multivariate time-series analysis model based on the data collection results, wherein the wind farm includes K wind turbines; a target faulty turbine location module, which receives and analyzes K operational sequence data from the K wind turbines in real time to locate the target faulty turbine; and a collaborative control window configuration module, which performs a risk assessment on the target faulty turbine and configures a collaborative control window based on the assessment results. The system comprises: a power allocation module, which obtains power demand forecasts from the grid operator and allocates power to K-1 wind turbines based on these forecasts to obtain K-1 output power constraints; a short-term control strategy acquisition module, which collects environmental data from the K-1 wind turbines and synchronizes the obtained K-1 real-time wind power environment data and K-1 output power constraints to a pre-built control strategy analysis model to obtain K-1 short-term control strategies; and a collaborative control optimization module, which uses the collaborative control window as a constraint to isolate the target faulty turbine and performs collaborative control optimization on the K-1 wind turbines using the K-1 short-term control strategies.

Images