Method and device for identifying and processing abnormal points of wind turbine power curve

Abstract

Embodiments of the present application provide a kind of wind turbine power curve anomaly point identification and processing method and device.The present application relates to wind turbine operation monitoring and data analysis field, the method comprises: obtaining and screening the operation data of wind turbine, according to operation data, the actual power curve of wind turbine is obtained;Based on density clustering algorithm, the abnormal scatter point deviating from standard power curve in actual power curve is identified;According to the distribution characteristics of abnormal scatter point and corresponding operation data, match the preset fault type rule base, determine fault root cause;According to fault root cause, output analysis report and improvement countermeasure.The method is used to improve the accuracy of anomaly identification, enhance fault attribution ability, shorten artificial investigation cycle, improve operation monitoring efficiency and operation and maintenance decision effectiveness.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method and apparatus for identifying and processing anomalies in the power curve of a wind turbine generator. Background Technology

[0002] As a crucial component of clean energy, wind power generation's stable operation and power generation efficiency directly impact the economic benefits of wind farms and the stability of the power grid. In actual operation, the power curve of a wind turbine (i.e., the curve relating wind speed to output power) is a core indicator for evaluating the turbine's operating status. However, due to environmental factors (such as extreme weather), equipment failures (such as blade icing and sensor malfunctions), and control strategy deviations, abnormal scattered points often appear in the power curve data. These anomalies may mask true performance issues, leading to distorted power generation estimates, incorrect operation and maintenance decisions, and even equipment damage or safety hazards.

[0003] Currently, the identification of outliers in wind turbine power curves relies heavily on manual analysis. However, with the expansion of wind farms and the surge in the number of turbines, traditional manual analysis methods are struggling to meet the demands of real-time processing of massive amounts of data. Therefore, existing technologies suffer from low accuracy in identifying outliers in wind turbine power curves. Summary of the Invention

[0004] This application provides a method and apparatus for identifying and processing anomalies in the power curve of a wind turbine. By addressing the power curve analysis needs under complex operating conditions of wind turbines, it comprehensively utilizes multi-dimensional operating information of the wind turbine to identify anomalies in the power curve. Based on the identification of anomalies, it further analyzes and judges the causes of the anomalies, thereby achieving effective discovery of power curve anomalies, accurate location of fault root causes, and output of anomaly processing results.

[0005] In a first aspect, embodiments of this application provide a method for identifying and processing outliers in the power curve of a wind turbine generator, including:

[0006] Obtain and filter the operating data of the wind turbine units, and obtain the actual power curve of the wind turbine units based on the operating data; the operating data shall include at least wind speed, power generation, blade angle and ambient temperature;

[0007] Based on density clustering algorithm, identify abnormal scatter points in the actual power curve that deviate from the standard power curve;

[0008] Based on the distribution characteristics of abnormal scattered points and the corresponding operational data, a preset fault type rule base is matched to determine the root cause of the fault.

[0009] Output an analysis report and improvement measures based on the root cause of the failure.

[0010] In one possible embodiment, based on a density clustering algorithm, identifying outlier scatter points in the actual power curve that deviate from the standard power curve includes:

[0011] The wind speed range in the actual power curve is divided according to the set step size.

[0012] Within each wind speed range, joint clustering is performed using two parameters: power generation and blade angle.

[0013] The minimum number of samples is fixed, and the neighborhood radius parameter is adaptively adjusted according to the proportion of noise points.

[0014] The clustering results of each wind speed range are compared with the standard power curve, and the scatter points that exceed the corresponding wind speed threshold range are identified as abnormal scatter points.

[0015] In one possible embodiment, the fault type rule base includes at least the discrimination rules for blade icing, blade high-temperature stall, and anemometer jamming or icing faults; based on the distribution characteristics of abnormal scatter points and the corresponding operational data, a preset fault type rule base is matched to determine the root cause of the fault, including:

[0016] Extract data features from outlier scattered points; these data features should include at least distribution features, power deviation features, and propeller angle state features.

[0017] Cross-validate data features by linking second-level data and event data;

[0018] Sort by rule priority and output the unique matching root cause of the fault from the fault type rule base.

[0019] In one possible embodiment, the analysis report is presented in document format and includes at least:

[0020] The system includes basic information about the entire site and each unit, visualization charts of anomalies, analysis and visualization charts of anomaly types, statistical distribution of anomalies in each unit, and suggestions for improvement measures to address the root causes of failures.

[0021] In one possible embodiment, after obtaining the analysis report, the process includes:

[0022] Based on the analysis report and improvement measures, control commands are output and sent to the wind turbine main control system. The wind turbine main control system receives the control commands and controls the wind turbine based on them.

[0023] In one possible embodiment, after obtaining the actual power curve of the wind turbine based on operating data, the process includes:

[0024] Determine whether the user requires a preset method to identify abnormal scatter points in the actual power curve;

[0025] If so, then based on the operational data and combined with the statistical patterns of historical operational data, calculate the average value and standard deviation of the power for each wind speed segment; determine the upper and lower limits of the theoretical power based on the average value and standard deviation, connect the upper and lower limits of the theoretical power for each wind speed segment and smooth them to obtain the upper and lower limit curves of the standard power curve; and identify abnormal scattered points based on the upper and lower limit curves of the standard power curve and the actual power curve.

[0026] If not, then anomaly points are identified based on density clustering algorithms.

[0027] Secondly, embodiments of this application provide an anomaly identification and processing device for the power curve of a wind turbine generator, comprising:

[0028] The first processing module is used to acquire and filter the operating data of the wind turbine, and obtain the actual power curve of the wind turbine based on the operating data; wherein, the operating data includes at least wind speed, power generation, blade angle and ambient temperature;

[0029] The identification module is used to identify abnormal scatter points in the actual power curve that deviate from the standard power curve based on the density clustering algorithm.

[0030] The second processing module is used to match the distribution characteristics of abnormal scattered points and the corresponding operating data with a preset fault type rule base to determine the root cause of the fault.

[0031] The output module is used to generate analysis reports and improvement strategies based on the root causes of failures.

[0032] In one possible embodiment, the identification module is further used for:

[0033] The wind speed range in the actual power curve is divided according to the set step size.

[0034] Within each wind speed range, joint clustering is performed using two parameters: power generation and blade angle.

[0035] The minimum number of samples is fixed, and the neighborhood radius parameter is adaptively adjusted according to the proportion of noise points.

[0036] The clustering results of each wind speed range are compared with the standard power curve, and the scatter points that exceed the corresponding wind speed threshold range are identified as abnormal scatter points.

[0037] In one possible embodiment, the fault type rule base includes at least the discrimination rules for blade icing, blade high-temperature stall, and anemometer jamming or icing faults; the second processing module is also used for:

[0038] Extract data features from outlier scattered points; these data features should include at least distribution features, power deviation features, and propeller angle state features.

[0039] Cross-validate data features by linking second-level data and event data;

[0040] Sort by rule priority and output the unique matching root cause of the fault from the fault type rule base.

[0041] In one possible embodiment, the analysis report is presented in document format and includes at least:

[0042] The system includes basic information about the entire site and each unit, visualization charts of anomalies, analysis and visualization charts of anomaly types, statistical distribution of anomalies in each unit, and suggestions for improvement measures to address the root causes of failures.

[0043] In one possible embodiment, the second processing module is further configured to:

[0044] Based on the analysis report and improvement measures, control commands are output and sent to the wind turbine main control system. The wind turbine main control system receives the control commands and controls the wind turbine based on them.

[0045] In one possible embodiment, the first processing module is further configured to:

[0046] Determine whether the user requires a preset method to identify abnormal scatter points in the actual power curve;

[0047] If so, then based on the operational data and combined with the statistical patterns of historical operational data, calculate the average value and standard deviation of the power for each wind speed segment; determine the upper and lower limits of the theoretical power based on the average value and standard deviation, connect the upper and lower limits of the theoretical power for each wind speed segment and smooth them to obtain the upper and lower limit curves of the standard power curve; and identify abnormal scattered points based on the upper and lower limit curves of the standard power curve and the actual power curve.

[0048] If not, then anomaly points are identified based on density clustering algorithms.

[0049] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0050] The memory stores computer-executed instructions;

[0051] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0052] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0053] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0054] The method and apparatus for identifying and processing anomalies in the power curve of a wind turbine provided in this application acquire and filter operating data including wind speed, power generation, blade angle, and ambient temperature to form the actual power curve of the wind turbine. This introduces multi-dimensional operating status information into the power characteristic analysis, improving the pertinence and reliability of anomaly data identification under complex operating conditions, thereby reducing misjudgments and omissions caused by relying solely on single power data or fixed thresholds. By identifying abnormal scattered points in the actual power curve that deviate from the standard power curve based on density clustering algorithm, and combining the distribution characteristics of the abnormal scattered points with the corresponding operating data to match a preset fault type rule base, the method can link anomaly identification with fault attribution, improving the ability to analyze the root causes of anomalies, thereby reducing the burden of manual troubleshooting and providing a basis for wind turbine operation optimization and fault handling. By outputting analysis reports and improvement countermeasures based on the root causes of faults, a complete analysis process from anomaly identification and cause determination to handling suggestions can be formed, thereby improving the efficiency of wind turbine operation and maintenance analysis and the level of refined management. Attached Figure Description

[0055] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0056] Figure 1 Flowchart of the method for identifying and handling outliers in the power curve of wind turbine provided in this application Figure 1 ;

[0057] Figure 2 Flowchart of the method for identifying and handling outliers in the power curve of wind turbine provided in this application Figure 2 ;

[0058] Figure 3 A schematic diagram illustrating the anomaly identification method for the wind turbine power curve anomaly identification and processing method provided in this application;

[0059] Figure 4 A schematic diagram of the structure of the device for identifying and processing abnormal points in the power curve of the wind turbine provided in this application;

[0060] Figure 5 A hardware schematic diagram of the device for identifying and processing abnormal points in the power curve of a wind turbine provided in this application.

[0061] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0062] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of methods and approaches consistent with some aspects of this application.

[0063] Wind turbine operation status assessment and fault diagnosis belong to the field of wind power operation monitoring and data analysis technology. Its typical application scenario is that during the daily production and operation of wind farms, relying on the unit-side sensors, field control units, SCADA monitoring platforms and background analysis modules, the operating parameters such as wind speed, power generation, blade angle and ambient temperature are continuously collected, stored and analyzed, and the power curve of the wind turbine is monitored accordingly.

[0064] In practical applications, wind farms often consist of a large number of wind turbines of varying models, located in dispersed locations. Each turbine operates continuously under different seasons, wind conditions, and terrain environments, generating large-scale, multi-time-series operational data. Maintenance personnel typically need to examine the deviation between the actual power curves and standard power curves of each turbine at the control center or on a remote monitoring platform to determine if there are any issues such as performance degradation, control anomalies, or environmental impacts.

[0065] Especially under complex operating conditions such as low-temperature icing, gusty wind disturbances, and frequent changes in local wind conditions, the unit's output power may fluctuate significantly, and scattered, offset, or even clustered abnormal points may appear on the power curve. In order to ensure the unit's power generation efficiency and equipment safety, the relevant systems must not only complete the basic data collection and curve generation, but also promptly identify abnormal points in massive amounts of operating data, and determine the cause of the anomaly by combining the unit's operating status and environmental conditions at the time.

[0066] Therefore, this type of technology is usually deployed in the wind farm control center or cloud analysis platform, and works in conjunction with the field data acquisition system to support unit status early warning, fault diagnosis, power generation performance evaluation and subsequent operation and maintenance decision making.

[0067] In existing technologies, anomaly detection of wind turbine power curves typically relies on wind speed and power generation data. A reference curve is first established based on historical operating samples or standard power curves. Then, statistical analysis or cluster analysis is used to determine whether certain operating points deviate from the normal distribution range. A common approach is to divide wind speed into sections, calculate the average power and fluctuation range for each wind speed interval, and then set a fixed threshold, considering points exceeding the threshold as anomalies. Another approach is to construct a two-dimensional sample using wind speed and power, and then use clustering algorithms to identify outliers in sparsely distributed areas, thereby completing anomaly detection. This type of solution is relatively straightforward in its data processing logic and is suitable for basic anomaly screening. It has some practicality in scenarios with a small number of turbines, relatively stable operating conditions, or minimal data fluctuations.

[0068] However, the limitations of the aforementioned technical approaches become apparent quickly when wind farms enter complex operating environments. Fixed threshold schemes typically assume that power fluctuation patterns are relatively consistent across different turbine models, wind farms, and seasons. However, in reality, changes in air density, terrain disturbances, differences in control strategies, and component conditions can all cause deviations in the actual power curve. Static thresholds alone cannot accurately distinguish between normal fluctuations and genuine anomalies, easily leading to false alarms and missed alarms. While cluster analysis relying solely on the two-dimensional relationship between wind speed and power can identify some outliers, the lack of interpretability for turbine operating conditions and environmental background often only leads to the conclusion that "anomalies exist," without further explanation as to whether the anomalies are caused by blade icing, changes in control actions, sudden wind changes, or measurement deviations.

[0069] Especially in cases of low power output, different root causes may exhibit highly similar patterns on the two-dimensional curves. Without incorporating operational data such as blade angle and ambient temperature for joint analysis, effective differentiation becomes difficult. Consequently, maintenance personnel still need to manually review trend charts, compare logs, and rely on experience to troubleshoot item by item. This results in lengthy analysis cycles, inconsistent standards, and an inability to meet the timeliness and accuracy requirements of large-scale wind farms. Furthermore, existing solutions often remain at the identification level for anomaly handling, lacking a complete connection from scattered anomalies to root causes, and then to analysis reports and improvement suggestions. This makes it difficult to directly translate monitoring results into actionable maintenance measures.

[0070] Therefore, improving the accuracy of anomaly identification in wind turbine power curves and enhancing the ability to analyze root causes of anomalies under complex operating conditions is a pressing technical problem in the field of wind power operation monitoring. To address this issue, this application proposes a method for identifying and processing anomalies in wind turbine power curves. After receiving turbine operating data from the wind farm control center or cloud analysis platform, the method first acquires and filters the turbine's operating data, obtaining the actual power curve based on this data. The operating data includes at least wind speed, power generation, blade angle, and ambient temperature. Then, based on a density clustering algorithm, it identifies anomalous points in the actual power curve that deviate from the standard power curve. After identifying these anomalous points, it matches them to a pre-defined fault type rule base based on their distribution characteristics and corresponding operating data to determine the root cause of the fault. Finally, it outputs an analysis report and improvement strategies based on the determined root cause. This technical approach goes beyond a simple screening of anomalies; it further integrates anomaly identification with cause determination, allowing anomalies to be interpreted more specifically in conjunction with the turbine's operating status and ambient temperature, thereby reducing ambiguity caused by relying solely on a single power curve. Through the above processing flow, in the operation scenarios of wind farms with multiple units, long cycles, and complex operating conditions, it is possible to more efficiently complete the identification of power curve anomalies, fault attribution, and result output, providing operation and maintenance personnel with directly referable analytical conclusions and improvement directions, thereby improving the accuracy of unit operation monitoring and the efficiency of operation and maintenance decision-making.

[0071] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0072] Figure 1 Flowchart of the method for identifying and handling outliers in the power curve of wind turbine provided in this application Figure 1 ,like Figure 1 As shown, the method includes:

[0073] S101. Obtain and filter the operating data of the wind turbine unit, and obtain the actual power curve of the wind turbine unit based on the operating data.

[0074] In this embodiment, this step can be performed by a data processing module deployed in the wind farm control center, the main server of the wind farm, the edge computing gateway, or the cloud analysis platform. The operating data of the wind turbine can come from the historical database or real-time data bus of the nacelle wind measurement device, the generator power acquisition unit, the pitch control system, the environmental monitoring device, and the SCADA (Superviso (Control and Data Acquisition) monitoring platform.

[0075] Operating data includes at least wind speed, power generation, blade angle, and ambient temperature. Wind speed characterizes the level of incoming wind energy received by the wind turbine at a given time; power generation characterizes the unit's electrical output at that time; blade angle characterizes the control angle of the blades relative to the incoming flow direction; and ambient temperature characterizes the external environmental conditions surrounding the unit and serves as an important auxiliary variable for identifying low-temperature icing, changes in air density, and differences in operating conditions. The actual power curve can be understood as a curve showing the correspondence between wind speed and power generation based on continuous operating data. It can be expressed either as a scatter plot or as a representative value sequence after segmented statistical analysis.

[0076] In one possible embodiment, the actual power curve can be generated by directly plotting a two-dimensional scatter plot of wind speed and power generation, thereby preserving the discrete distribution characteristics under complex operating conditions and facilitating the identification of abnormal scatter points by subsequent density clustering algorithms.

[0077] In another possible embodiment, the mean, median, upper and lower quantiles or confidence intervals of power can be calculated first based on each wind speed segment, and then an actual power curve with statistical characteristics can be formed to express the output characteristics under different wind conditions in a structured way.

[0078] To enhance data comparability, the rated power of the unit can be introduced to normalize the power generation, or an air density correction model can be used to convert the wind speed and power data to operating conditions. Based on these processing steps, the resulting actual power curve can more realistically reflect the output behavior of the target wind turbine during the analysis period, providing a stable data foundation for subsequent comparison with standard power curves and identification of deviation points.

[0079] Based on the above analysis, it can be seen that by completing time synchronization, status filtering, invalid data removal, and curve construction at the source data stage, interference from non-target factors such as downtime, power outages, maintenance, and abnormal data acquisition can be avoided in subsequent analysis, thereby improving the reliability of anomaly identification results. It should be understood that the above examples are for illustrative purposes only and are not limiting.

[0080] S102. Based on the density clustering algorithm, identify abnormal scattered points in the actual power curve that deviate from the standard power curve.

[0081] In this embodiment of the application, the density clustering algorithm can be DBSCAN (Density-Based Spatial Clustering of Applications with Noise), HDBSCAN (Hierar (Augmented Reality) Chemical Density-Based Spatial Clustering of Applications with Noise), or other clustering methods based on the density distribution of sample neighborhoods.

[0082] The core idea of ​​density clustering is to determine whether a sample point belongs to a high-density region based on the number of neighboring samples in the feature space, and to identify data points in low-density regions that cannot be classified into effective clusters as noise points or outliers. Abnormal scattered points deviating from the standard power curve refer to data points whose power generation differs significantly from the corresponding power on the standard power curve under similar wind speed conditions, and whose differences exhibit low-density, isolated distribution, or anomalous clustering characteristics in the multidimensional operating feature space.

[0083] This step is crucial for solving the problem of accurate anomaly identification under complex operating conditions. Traditional fixed threshold methods typically rely solely on the two-dimensional wind speed-power relationship for simple judgments, making it difficult to distinguish between scattered diffusion caused by normal environmental fluctuations and abnormal deviations caused by genuine faults. However, the embodiments of this application incorporate wind speed, power generation, blade angle, ambient temperature, and standard curve deviation into the density analysis space, transforming anomaly identification from a simple geometric deviation judgment into a distribution identification process that incorporates the background of the operating state.

[0084] S103. Based on the distribution characteristics of abnormal scattered points and the corresponding operational data, match the preset fault type rule base to determine the root cause of the fault.

[0085] In this embodiment, the fault type rule base can be pre-established and stored in a local database, rule engine service, or cloud knowledge base. Each rule in the rule base describes the correspondence between an anomaly pattern and the corresponding root cause of the fault. Distribution characteristics may include the occurrence range of the abnormal scatter points in the wind speed dimension, the deviation direction and magnitude in the power dimension, the persistence in the time dimension, the degree of clustering in the sample space, and the overall offset trend from the standard power curve. The corresponding operational data includes wind speed, power generation, blade angle, and ambient temperature when the abnormal scatter points occur, and may further include auxiliary data such as unit speed, yaw status, power limiting command, vibration value, and fault logs. The root cause of the fault refers to the category of causes that lead to the formation of abnormal scatter points on the power curve, such as blade icing, abnormal blade pitch control, wind speed measurement deviation, sudden changes in local wind conditions, power limiting operation, and equipment performance degradation.

[0086] In practice, statistical analysis of the anomalous scatter points can be performed first to extract their group characteristics. For example, the proportion of anomalous scatter points in each wind speed range, average power deviation, ratio of low to high scatter points, duration of continuous occurrence, cluster center location, and average distance from the standard power curve can be calculated. If the anomalous scatter points are mainly distributed in the medium-to-high wind speed range and generally exhibit low power generation, accompanied by a large blade angle, it can be determined that the unit may have abnormal control actions or decreased blade aerodynamic performance. If the anomalous scatter points are concentrated under low ambient temperature conditions and form a downward cluster in multiple adjacent wind speed ranges, while the blade angle and power response show an abnormal correspondence, they can be matched with icing feature rules in the rule base. If the anomalous points are distributed in patches of high wind speed within a specific wind speed range, but the ambient temperature, blade angle, and other data do not show significant abnormalities, the wind speed sensor calibration status can be further combined to determine whether there is a wind speed measurement deviation.

[0087] Based on the above analysis, this step achieves a technical connection from anomaly identification to anomaly interpretation, which is crucial for enhancing the ability to analyze the root causes of anomalies. By combining the geometric distribution characteristics of anomaly points with relevant parameters of unit operating mechanisms, the problem of cause confusion caused by relying solely on the two-dimensional relationship between wind speed and power can be avoided. The rule base transforms historical experience, equipment mechanisms, and statistical characteristics into executable judgment logic, enabling the system to automatically output more targeted fault conclusions, reducing the workload of manually reviewing trend charts and logs, and improving consistency when analyzing multiple units in batches. It should be understood that the above example is for demonstration purposes only and is not limiting.

[0088] S104. Output an analysis report and improvement measures based on the root cause of the fault.

[0089] In this embodiment, the analysis report can be automatically generated by the report generation module and output to the operation and maintenance platform, mobile terminal, email system, or work order system in the form of text, charts, web interface, alarm message, or structured data file. Improvement strategies refer to processing suggestions generated for specific fault root causes, aiming to ensure that anomaly identification results directly serve operation and maintenance decisions and subsequent handling processes.

[0090] In practice, after determining the root cause of the fault, the system can summarize the number of corresponding abnormal scatter points, the distribution wind speed range, the time period of occurrence, the estimated average power loss, and the confidence level of rule matching to form a structured diagnostic result. Then, it calls the report template engine to populate the structured result into a preset report template. The visualization content in the report can include a scatter plot of the actual power curve, a reference line for the standard power curve, a highlighted map of abnormal scatter points, a distribution map of abnormal points along the ambient temperature dimension, and a description of the root cause matching. For improvement measures, corresponding handling suggestions can be extracted from the strategy library based on the root cause of the fault. For example, when the root cause of the fault is blade icing, improvement measures may include arranging on-site inspections, checking the operating status of heating or de-icing devices, conducting key monitoring in conjunction with low-temperature weather warnings, and assessing whether the shutdown protection strategy needs to be adjusted; when the root cause of the fault is abnormal pitch control, improvement measures may include checking the pitch actuator, verifying controller parameters, and comparing the deviation between the feedback angle and the command angle; when the root cause of the fault is wind speed measurement deviation, improvement measures may include calibrating the wind speed sensor, cleaning the wind measurement device, comparing the wind speed of neighboring units, and performing measurement correction; when the root cause of the fault is power-limited operation or external scheduling constraints, improvement measures may include marking this period as a non-performance degradation problem and adjusting the subsequent evaluation criteria.

[0091] Based on the above analysis, it is evident that by unifying the output of anomaly identification results, root cause determination results, and improvement strategies, a complete closed loop can be formed, encompassing data collection, anomaly detection, root cause analysis, and operational recommendations. Compared to existing solutions that only focus on anomaly identification, this application's embodiment can directly transform analytical conclusions into actionable information, thereby shortening the manual troubleshooting cycle, reducing reliance on experience, and improving the efficiency of state assessment under complex wind conditions. The structured and visualized output of the analysis report also facilitates sharing across work teams and wind farms, supporting unified monitoring and standardized operation and maintenance of large-scale wind farms. It should be understood that the above examples are merely illustrative and not limiting.

[0092] This application provides a method for identifying and processing anomalies in the power curve of a wind turbine. By integrating multi-dimensional operational information such as wind speed, power generation, blade angle, and ambient temperature into the entire process of data filtering, anomaly identification, and root cause determination, and utilizing a processing mechanism combining density clustering and rule-based matching, it can effectively distinguish between normal fluctuations and true anomalies under complex operating conditions, further improving the accuracy and stability of power curve anomaly identification. Simultaneously, through automatic attribution of the causes of anomaly points and the linked output of analysis reports and improvement strategies, monitoring results can be directly transformed into operation and maintenance references, thereby enhancing the efficiency of wind turbine performance evaluation, fault early warning, and maintenance decision-making. It should be understood that the above examples are merely illustrative and not limiting.

[0093] Figure 2Flowchart of the method for identifying and handling outliers in the power curve of wind turbine provided in this application Figure 2 ,like Figure 2 As shown, in this embodiment... Figure 1 Based on the embodiments, a detailed description is provided of the method for identifying and handling outliers in the power curve of a wind turbine. Step S201 involves obtaining the actual power curve of the wind turbine and determining the method for identifying outliers; step S202 involves identifying outliers; step S203 involves obtaining the root cause of the fault; step S204 involves obtaining an analysis report and improvement measures; and step S205 involves outputting control commands. This method includes:

[0094] S201. Obtain and filter the operating data of the wind turbine units, and obtain the actual power curve of the wind turbine units based on the operating data; determine whether the user requires the use of a preset method to identify abnormal scattered points in the actual power curve; if so, calculate the average value and standard deviation of the power for each wind speed segment based on the operating data and the statistical patterns of historical operating data; determine the upper and lower limits of the theoretical power based on the average value and standard deviation, connect the upper and lower limits of the theoretical power for each wind speed segment and perform smoothing to obtain the upper and lower limit curves of the standard power curve; identify abnormal scattered points based on the upper and lower limit curves of the standard power curve and the actual power curve; if not, identify abnormal scattered points based on the density clustering algorithm.

[0095] In this embodiment, multi-source operational data of the target unit can be extracted from a preset time range, which can be the past 24 hours, the past 7 days, the past 30 days, or a complete analysis cycle. During the extraction process, data from different acquisition sources are aligned according to a unified timestamp. For data with different sampling frequencies, resampling, interpolation, or time window aggregation can be used to complete synchronization processing. For example, 1-second sampling data can be aggregated into 10-second, 1-minute, or 10-minute windows to obtain analysis samples containing average wind speed, average power, average blade angle, and average ambient temperature. Subsequently, the operational data is filtered to exclude data points that do not meet the power curve modeling conditions. Filtering conditions may include the unit being in grid-connected power generation status, no shutdown command from the main control system, valid wind speed sensor data, no missing power signals, blade pitch control not in maintenance mode, and ambient temperature data within the sensor's measurement range. For obviously abnormal raw records, noise reduction and cleaning processing can also be performed, such as removing data segments with negative power values, wind speed mutations exceeding the physically reasonable range, duplicate records at the same timestamp, or continuous missing measurement times exceeding a threshold. To ensure the stability of subsequent anomaly identification, the data can be standardized and organized after screening. The wind speed can be segmented according to a preset step size, which can be set to 0.5 m / s or 1 m / s. Within each wind speed segment, the corresponding power generation, blade angle and ambient temperature samples are retained to form a basic data set for constructing the actual power curve.

[0096] In one possible implementation, the user terminal triggers the selection of anomaly point identification methods. This user terminal can be an interactive terminal in the wind farm control center, a configuration interface of a cloud-based analysis platform, or a mobile maintenance terminal. Its function is to switch between statistical boundary-based and cluster-based identification methods according to current analysis needs. Historical operational data statistical patterns reflect the normal fluctuation range of the unit within different wind speed ranges. These patterns are formed from long-term collected and cleaned wind speed and power samples, reducing the impact of single sampling deviations on boundary estimation. The theoretical power upper and lower limits characterize the normal power range for each wind speed range. They can be obtained by superimposing the average value with a multiple of the standard deviation and are used for subsequent boundary violation detection. The upper and lower limit curves of the standard power curve are continuous boundary curves formed by connecting the discrete boundaries of each wind speed range and then smoothing them, allowing for point-by-point comparison with the actual power curve.

[0097] In practical implementation, after obtaining the actual power curve, the system first reads the identification method command issued by the user terminal. When the user terminal selects a preset method, the operating data is collected according to wind speed segments. For each wind speed segment, the mean and dispersion of the corresponding power generation are statistically analyzed, and the statistical results are corrected by combining historical operating samples to form a stable power boundary. Subsequently, upper and lower limit boundaries are generated based on the mean and standard deviation of each wind speed segment. Then, the boundaries of adjacent wind speed segments are connected, and spline smoothing or sliding fitting is used to eliminate inflection points, so that the upper and lower limit boundaries change continuously on the wind speed axis. The system compares the scattered points in the actual power curve with this continuous boundary point by point. When a scattered point exceeds the upper or lower limit of the corresponding wind speed position, it is marked as an abnormal scattered point. When the user terminal does not select a preset method, the sample composed of wind speed and power is directly input into the density clustering model, and points with low density and separated from the main cluster are identified as abnormal scattered points.

[0098] Figure 3 A schematic diagram of the anomaly identification method for the wind turbine power curve anomaly identification and processing method provided in this application is shown below. Figure 3 As shown, anomaly identification is achieved based on the standard power curve or the overall average power curve, combined with the statistical patterns of historical operating data. The average power P and standard deviation σ are calculated for each wind speed segment, and the theoretical power upper and lower limits are P±σ. Connecting the theoretical power upper and lower limits for each wind speed segment and smoothing the curves yields the upper and lower limit curves. Scattered points outside the upper and lower limit intervals are identified as anomalies, further categorized into low-to-high anomalies and high-to-low anomalies.

[0099] If the user end does not require the use of a preset method to identify abnormal scatter points in the actual power curve, the DBSCAN density clustering algorithm is used to adaptively identify abnormal data based on the distribution characteristics of the power curve scatter points. Wind speed intervals are divided according to a set step size, and joint clustering is performed within each interval using two parameters: power and blade angle. EPS is adaptively adjusted based on the proportion of noise points. The clustering results for each wind speed interval, including both clustered points and noise points, are compared with the standard power curve. Scatter points exceeding the corresponding wind speed threshold range are identified as abnormal points.

[0100] This processing method can switch between statistical boundary identification and density clustering identification according to user needs. It can obtain a more stable power boundary by utilizing historical statistical patterns, and can also identify outliers in complex scattering scenarios through clustering algorithms, thereby improving the adaptability and accuracy of abnormal scattering identification and providing reliable input for subsequent root cause analysis of faults.

[0101] S202. Divide the wind speed range in the actual power curve according to the set step size; perform joint clustering using two parameters, power generation and blade angle, within each wind speed range; fix the minimum sample number parameter and adaptively adjust the neighborhood radius parameter according to the proportion of noise points; compare the clustering results of each wind speed range with the standard power curve, and identify the scattered points that exceed the corresponding wind speed threshold range as abnormal scattered points.

[0102] In this embodiment, the standard power curve is used to characterize the normal power generation pattern of the wind turbine under different wind speeds. Power generation and blade angle are used to reflect the unit's output capacity and control status, respectively. Together, they enhance the adaptability of outlier identification to changes in operating conditions. The minimum sample size parameter is a parameter used in density clustering to constrain the minimum number of samples required to form a cluster. It remains fixed throughout the identification process to avoid cluster structure instability caused by parameter fluctuations. The neighborhood radius parameter is used to limit the neighborhood range between sample points. Its size is adjusted according to the proportion of noise points within the current wind speed range. When the proportion of noise points increases, the neighborhood radius expands to absorb locally discrete data points that may still belong to the normal distribution; when the proportion of noise points decreases, the neighborhood radius shrinks to improve the ability to distinguish off-target samples.

[0103] In practical implementation, the wind speed intervals can be pre-set at fixed intervals based on the rated wind speed range of the unit, so that adjacent intervals cover continuous wind speed segments, and the number of data points within each wind speed interval can meet the statistical requirements of joint clustering. For each wind speed interval, the power generation and pitch angle of the corresponding samples form a two-dimensional joint sample set, and density clustering is performed within this sample set so that data that are similar in both power response characteristics and pitch control characteristics form the same cluster.

[0104] After clustering, the ratio of the number of data points identified as noise points within the wind speed interval to the total number of samples is calculated to form the noise point proportion. The neighborhood radius parameter is then updated based on this ratio, ensuring that the cluster boundary changes synchronously with the local data density. Subsequently, the cluster centers and cluster boundaries formed in each wind speed interval are compared with the power threshold range corresponding to the wind speed on the standard power curve. When the power generation corresponding to certain scattered points exceeds this threshold range, they are identified as anomalous scattered points. These anomalous scattered points include both low and high deviations.

[0105] During operation, this scheme first segments the wind speed dimension, then introduces the joint distribution information of power generation and blade angle within the local wind speed range. Stable clustering is achieved using a fixed minimum sample size and an adaptive neighborhood radius. Finally, outlier points are screened by comparing with a standard power curve. Because the clustering process considers both power output and control attitude, it reduces misjudgments caused by relying solely on a single power dimension, making outlier points more closely correspond to actual deviations from the operating conditions.

[0106] This approach improves the accuracy and robustness of identifying abnormal scattered points under complex wind conditions, reduces the probability of false alarms and missed alarms caused by fixed thresholds, and makes the identification results closer to the actual operating status of the unit, thereby providing a more reliable data foundation for subsequent root cause analysis of faults and operation and maintenance decisions.

[0107] S203. Extract data features from abnormal scattered points; the data features include at least distribution features, power deviation features, and propeller angle status features; cross-validate the data features by linking second-level data and event data; sort according to rule priority and output the unique matching root cause of the fault from the fault type rule base.

[0108] In this embodiment, the fault type rule base includes at least the discrimination rules for blade icing, blade high-temperature stall, and anemometer jamming or icing faults. The fault type rule base stores the discrimination conditions for different fault types. Its rule content can be established based on historical operating samples, alarm records, and maintenance conclusions of the wind turbine, and can be expanded when new fault types are added later. The data features of abnormal scatter points are used to characterize the spatial distribution of the scatter points in the actual power curve, the degree of deviation from the standard power curve, and the corresponding blade angle state, thereby providing a joint judgment basis for fault attribution. Second-level data is used to describe the instantaneous changes in wind speed, power, and blade angle within a short period before and after the anomaly occurs. Event data is used to record fault alarms, protection actions, or manual maintenance records; both are used together to verify the consistency between the feature extraction results and the actual operating state. Rule priority ranking is used to determine the unique root cause of the final output based on the matching confidence and feature completeness when multiple discrimination rules are simultaneously satisfied.

[0109] In practical processing, the aggregation patterns of abnormal scattered points within each wind speed range are first statistically analyzed to form distribution characteristics. Then, the magnitude of their deviation from the standard power curve is calculated to form power deviation characteristics. Simultaneously, the variation range, fluctuation amplitude, and stability of the blade angle at the corresponding moment are extracted to form blade angle state characteristics. Subsequently, the second-level data corresponding to the abnormal scattered points are correlated and compared with the event data. When the abnormal scattered points are accompanied by persistently low power, abnormal blade angle response, and icing alarms in the event data under low temperature conditions, they can be matched with blade icing discrimination rules. When the abnormal scattered points show a decrease in power under high temperature conditions and are accompanied by limited blade angle adjustment, they can be matched with blade high-temperature stall discrimination rules. When the abnormal scattered points are inconsistent with wind speed changes and the wind speed signal in the second-level data remains unchanged or fluctuates abnormally for a long time, they can be matched with anemometer jamming or icing fault discrimination rules.

[0110] When multiple rules simultaneously meet some of the conditions, the system sorts the rules according to a preset priority and prioritizes the rule with the highest consistency with the distribution characteristics, power deviation characteristics, and propeller angle state characteristics, thereby avoiding multiple root causes of the same abnormal point. This method can combine the abnormal point identification results with the unit operating status, instantaneous data, and historical events, making fault attribution more specific and stable.

[0111] By adopting the above method, the accuracy of fault identification of abnormal scattered points can be improved, the misjudgment caused by relying solely on power curves can be reduced, and the ability to distinguish faults such as blade icing, blade high-temperature stall, and anemometer jamming or icing can be improved. This makes the output fault root cause unique and interpretable, and provides a more reliable basis for subsequent operation and maintenance decisions.

[0112] S204. Output an analysis report and improvement measures based on the root cause of the fault.

[0113] In one possible implementation, the analysis report is presented in document format and includes at least: basic information of the entire site and each unit, visualization charts of anomalies, anomaly type analysis and visualization charts, distribution statistics of anomalies in each unit, and improvement countermeasure suggestions for the root causes of failures.

[0114] The analysis report presents the results of anomaly identification and fault attribution, providing a basis for subsequent operation and maintenance decisions. Basic information for the entire wind farm and each unit can include wind farm name, site location, unit number, unit capacity, operating status, data collection period, and grid connection status of the corresponding units, allowing for unified identification of the analysis objects. Anomaly point visualization charts can be in the form of power curve scatter plots, density heat maps, or curve overlay plots, visually displaying the location, deviation trend, and clustering of anomaly points on the actual power curve. Anomaly type analysis and visualization charts categorize the matched root causes of faults by type, presenting the proportion, frequency, and corresponding unit distribution of each type of anomaly using bar charts, pie charts, or classification statistics. Anomaly point distribution statistics across units are used to statistically analyze the number, proportion, concentration, and recurrence frequency of anomalies across different units, thereby identifying whether anomalies are concentrated in a single unit or multiple units. The improvement strategy suggestions for the root cause of the fault generate corresponding handling suggestions, review suggestions or operation adjustment suggestions based on the identified root cause. For example, for icing-related root causes, the suggestion is to carry out de-icing inspections, and for anemometer malfunctions, the suggestion is to check the sensor status, so as to assist maintenance personnel in handling the situation quickly.

[0115] In practical implementation, the analysis report can be automatically generated by the backend analysis module after anomaly identification and root cause matching, and output in PDF (Portable Document Format), Word, or other archiveable document formats for easy viewing, printing, and archiving at the control center. The chart data in the report comes from the matching results of the wind farm operation database, SCADA database, and fault type rule base, while the text descriptions are generated by combining templated statements with real-time statistical results. By integrating basic information, chart statistics, and improvement suggestions into a single document, maintenance personnel can avoid switching between multiple interfaces for queries, creating a closed-loop presentation of anomaly information, anomaly types, and handling suggestions. The document format facilitates transfer and long-term storage across different platforms; however, other formats can be selected in practical applications, and this embodiment does not limit the choice of format.

[0116] The analysis report uses a standardized format to simultaneously display the overall site overview, unit status, anomaly location, and anomaly attribution, enabling maintenance personnel to quickly locate abnormal units and determine the anomaly type. This, combined with suggestions for improvement measures, allows for review or action. Because the anomaly distribution statistics and anomaly type analysis are interconnected, the report can indicate whether anomalies are clustered, periodic, or concentrated on a single unit, thereby improving the depth of understanding of the root causes of failures.

[0117] After adopting this analysis report, anomaly identification results are no longer output as isolated data points, but are presented in a complete, readable, and archiveable document format, which improves the efficiency of operation and maintenance diagnosis and the consistency of reports. Through the combined display of anomaly point visualization charts, anomaly type analysis, and distribution statistics, the cost of manual review can be reduced, and the accuracy of identifying the root cause of the fault can be improved. Furthermore, the improvement countermeasure suggestions can directly correspond to the fault category and output handling recommendations, thereby shortening the time between anomaly discovery and operation and maintenance response, and improving the overall efficiency of wind farm operation monitoring and fault handling.

[0118] To improve operational feasibility, the analysis report can also include anomaly level assessments and handling priorities. Anomaly levels are calculated based on a combination of factors, including power loss severity, duration, affected wind speed range, and fault reliability. Handling priorities guide decisions on whether to immediately dispatch a work order, include it in planned maintenance, or continue monitoring. In one possible implementation, the system can also correlate the current analysis results with historical reports to track the recurrence of similar faults on the same turbine unit and provide assessments of potential worsening or improvement. In another possible implementation, the analysis report can be pushed to the wind farm's centralized monitoring platform via an interface, allowing maintenance personnel to simultaneously obtain anomaly summaries and recommended actions while viewing the turbine list.

[0119] S205. Based on the analysis report and improvement measures, output control commands and send the control commands to the wind turbine main control system.

[0120] In this embodiment, the analysis report can be a documented result summarizing anomalies, root causes of failures, and improvement suggestions. The improvement measures are operational adjustment suggestions or protection control suggestions corresponding to the root causes of the failures. Control commands are used to transmit operational adjustment intentions to the wind turbine main control system. Upon receiving the commands, the main control system can adjust the wind turbine's power limits, yaw control, pitch control, de-icing control, derating operation, or protective shutdown strategies accordingly. The wind turbine main control system receives control commands and controls the wind turbine based on these commands.

[0121] In practical implementation, the root causes of faults in the analysis report are mapped to a preset control strategy library, and corresponding control command parameter sets are generated by combining improvement measures. These parameter sets include at least the control type, triggering condition, execution duration, and recovery condition. Upon receiving the control command, the main control system invokes its internal control logic according to the command content to perform corresponding adjustments to the wind turbine, thereby correcting its operating state under abnormal conditions. This control command can be sent to the main control system via a communication interface, which can employ Ethernet, industrial bus, or wireless communication links to ensure real-time transmission of the control command.

[0122] By directly translating analysis reports and improvement measures into control commands and issuing them to the wind turbine's main control system, anomaly identification results can be linked with on-site control actions, reducing the lag caused by manual intervention and making fault handling more timely and accurate. Simultaneously, after the main control system controls the wind turbine based on these commands, it can mitigate the adverse effects of the persistent root causes of anomalies on turbine performance and equipment safety, thereby improving the wind turbine's operational stability, fault handling efficiency, and overall maintenance level.

[0123] Figure 4 A schematic diagram of the structure of the device for identifying and processing anomalies in the power curve of a wind turbine provided in this application is shown below. Figure 4 As shown, the anomaly identification and processing device 40 for the power curve of a wind turbine provided in this embodiment includes:

[0124] The first processing module 401 is used to acquire and filter the operating data of the wind turbine unit, and obtain the actual power curve of the wind turbine unit based on the operating data; wherein, the operating data includes at least wind speed, power generation, blade angle and ambient temperature;

[0125] The identification module 402 is used to identify abnormal scattered points in the actual power curve that deviate from the standard power curve based on the density clustering algorithm.

[0126] The second processing module 403 is used to determine the root cause of the fault by matching the distribution characteristics of the abnormal scattered points and the corresponding operating data with a preset fault type rule base.

[0127] Output module 404 is used to output an analysis report and improvement measures based on the root cause of the fault.

[0128] In one possible embodiment, the identification module 402 is further configured to:

[0129] The wind speed range in the actual power curve is divided according to the set step size.

[0130] Within each wind speed range, joint clustering is performed using two parameters: power generation and blade angle.

[0131] The minimum number of samples is fixed, and the neighborhood radius parameter is adaptively adjusted according to the proportion of noise points.

[0132] The clustering results of each wind speed range are compared with the standard power curve, and the scatter points that exceed the corresponding wind speed threshold range are identified as abnormal scatter points.

[0133] In one possible embodiment, the fault type rule base includes at least the discrimination rules for blade icing, blade high-temperature stall, anemometer jamming, or icing faults; the second processing module 403 is also used for:

[0134] Extract data features from outlier scattered points; these data features should include at least distribution features, power deviation features, and propeller angle state features.

[0135] Cross-validate data features by linking second-level data and event data;

[0136] Sort by rule priority and output the unique matching root cause of the fault from the fault type rule base.

[0137] In one possible embodiment, the analysis report is presented in document format and includes at least:

[0138] The system includes basic information about the entire site and each unit, visualization charts of anomalies, analysis and visualization charts of anomaly types, statistical distribution of anomalies in each unit, and suggestions for improvement measures to address the root causes of failures.

[0139] In one possible embodiment, the second processing module 403 is further configured to:

[0140] Based on the analysis report and improvement measures, control commands are output and sent to the wind turbine main control system. The wind turbine main control system receives the control commands and controls the wind turbine based on them.

[0141] In one possible embodiment, the first processing module 401 is further configured to:

[0142] Determine whether the user requires a preset method to identify abnormal scatter points in the actual power curve;

[0143] If so, then based on the operational data and combined with the statistical patterns of historical operational data, calculate the average value and standard deviation of the power for each wind speed segment; determine the upper and lower limits of the theoretical power based on the average value and standard deviation, connect the upper and lower limits of the theoretical power for each wind speed segment and smooth them to obtain the upper and lower limit curves of the standard power curve; and identify abnormal scattered points based on the upper and lower limit curves of the standard power curve and the actual power curve.

[0144] If not, then anomaly points are identified based on density clustering algorithms.

[0145] The first processing module 401 jointly acquires and filters wind speed, power generation, blade angle, and ambient temperature to generate an actual power curve that more closely reflects the unit's true state, thereby reducing the interference of invalid data on the analysis results. The identification module 402 uses a density clustering algorithm to identify abnormal scattered points deviating from the standard power curve, effectively detecting discrete points, offset points, and clustered anomalies under complex operating conditions, thus improving the accuracy of anomaly identification. The second processing module 403 combines the distribution characteristics of abnormal scattered points with corresponding operating data and a fault type rule base for matching, thus further transforming mere anomalies into specific fault root cause judgments. The output module 404 generates an analysis report and improvement strategies based on the fault root causes, achieving a complete connection from data filtering, anomaly identification, fault attribution to maintenance recommendations, enabling maintenance personnel to quickly locate problems and develop targeted solutions.

[0146] Figure 5 A hardware schematic diagram of the device for identifying and processing outliers in the power curve of a wind turbine provided in this application. Figure 5 As shown, the electronic device 50 provided in this embodiment includes at least one processor 501 and a memory 502. Optionally, the device 50 further includes a communication component 503. The processor 501, memory 502, and communication component 503 are connected via a bus 504.

[0147] In a specific implementation, at least one processor 501 executes computer execution instructions stored in memory 502, causing at least one processor 501 to perform the above-described method.

[0148] This electronic device can be deployed in the wind farm control center or cloud analysis platform. Its memory stores data processing programs and rule bases for wind speed, power generation, blade angle, and ambient temperature. The processor calls and executes these programs to acquire operational data, generate actual power curves, identify anomalies, match root causes of faults, and output analysis reports. Because the processor directly executes the above methods based on computer instructions stored in memory, it can link anomaly detection with cause analysis, thereby improving the accuracy of power curve anomaly identification under complex operating conditions. This allows for more targeted differentiation of problems such as blade icing, control anomalies, sudden wind changes, and measurement deviations, thus reducing the time required for manual troubleshooting by maintenance personnel. Therefore, it helps improve the efficiency of unit condition assessment, fault diagnosis capabilities, and the timeliness of maintenance decisions.

[0149] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0150] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0151] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0152] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0153] This computer program product embeds or stores the process for identifying and handling anomalies in wind turbine power curves in the form of program instructions. This allows the processor to call upon functional modules such as turbine operation data acquisition, filtering, curve generation, anomaly point identification, root cause matching, and analysis report output. This integrates the previously manual, decentralized data processing and fault assessment into an automated analysis process. Because the program can combine multi-dimensional data such as wind speed, power generation, blade angle, and ambient temperature to complete anomaly identification and root cause determination, it makes anomaly point analysis under complex operating conditions more targeted, thereby improving the accuracy of anomaly identification and fault interpretation capabilities. Therefore, it can more efficiently support wind farm control centers or cloud platforms in conducting turbine status early warning, performance evaluation, and operation and maintenance decisions.

[0154] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0155] This computer-readable storage medium can be deployed in wind farm control center servers, edge computing terminals, or cloud analysis platforms. The processor invokes stored computer execution instructions to acquire, filter, analyze, and process operational data such as wind speed, power generation, blade angle, and ambient temperature, generating actual power curves. Subsequently, a density clustering algorithm is used to identify abnormal scatter points deviating from the standard power curve. Based on the distribution characteristics of these abnormal scatter points and the corresponding operational data, a preset fault type rule base is matched to determine the root cause of the fault. This allows for integrated software instruction-based anomaly identification, root cause determination, and result output, reducing the burden of manual troubleshooting. Furthermore, this storage medium is easily copied, installed, and invoked on different hardware platforms, enabling the method to adapt to the data analysis needs of multiple units, long cycles, and complex operating conditions. This improves the accuracy of wind turbine power curve anomaly identification and fault diagnosis efficiency, thus facilitating the generation of more targeted analysis reports and improvement strategies for maintenance personnel.

[0156] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0157] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0158] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, methods, or units, and may be electrical, mechanical, or other forms.

[0159] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0160] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0161] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0162] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0163] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for identifying and processing outliers in the power curve of a wind turbine generator, characterized in that, include: The wind turbine's operating data is acquired and filtered, and the actual power curve of the wind turbine is obtained based on the operating data; wherein, the operating data includes at least wind speed, power generation, blade angle and ambient temperature; Based on the density clustering algorithm, abnormal scatter points in the actual power curve that deviate from the standard power curve are identified; Based on the distribution characteristics of the abnormal scattered points and the corresponding operational data, a preset fault type rule base is matched to determine the root cause of the fault. Based on the root causes of the failures, an analysis report and improvement measures will be generated.

2. The method according to claim 1, characterized in that, The method of identifying outlier points in the actual power curve that deviate from the standard power curve based on the density clustering algorithm includes: The wind speed range in the actual power curve is divided according to the set step size; Within each wind speed range, joint clustering is performed using the power generation capacity and the blade angle as two parameters; The minimum number of samples is fixed, and the neighborhood radius parameter is adaptively adjusted according to the proportion of noise points. The clustering results of each wind speed range are compared with the standard power curve, and the scatter points that exceed the corresponding wind speed threshold range are identified as the abnormal scatter points.

3. The method according to claim 1, characterized in that, The fault type rule base includes at least the discrimination rules for blade icing, blade high-temperature stall, and anemometer jamming or icing faults; the step of determining the root cause of the fault by matching the preset fault type rule base with the distribution characteristics of the abnormal scatter points and the corresponding operating data includes: Extract the data features of the abnormal scatter points; wherein the data features include at least distribution features, power deviation features, and propeller angle state features; The data features are cross-validated by linking second-level data and event data; Sort by rule priority and output the unique matching root cause of the fault from the fault type rule base.

4. The method according to claim 1, characterized in that, The analysis report is presented in document format and includes at least the following: The system includes basic information about the entire site and each unit, visualization charts of anomalies, analysis and visualization charts of anomaly types, statistical distribution of anomalies in each unit, and suggestions for improvement measures to address the root causes of the faults.

5. The method according to claim 4, characterized in that, After receiving the analysis report, including: Based on the analysis report and improvement measures, control commands are output and sent to the wind turbine main control system; wherein, the wind turbine main control system is used to receive the control commands and control the wind turbine based on the control commands.

6. The method according to any one of claims 1-5, characterized in that, After obtaining the actual power curve of the wind turbine based on the operating data, the process includes: Determine whether the user terminal requires the use of a preset method to identify abnormal scattered points in the actual power curve; If so, based on the operational data and combined with the statistical patterns of historical operational data, calculate the average value and standard deviation of power for each wind speed segment; determine the upper and lower limits of theoretical power according to the average value and the standard deviation, connect the upper and lower limits of theoretical power for each wind speed segment and perform smoothing to obtain the upper and lower limit curves of the standard power curve; identify the abnormal scattered points according to the upper and lower limit curves of the standard power curve and the actual power curve. If not, the abnormal scatter points are identified based on the density clustering algorithm.

7. A device for identifying and processing outliers in the power curve of a wind turbine generator set, characterized in that, include: The first processing module is used to acquire and filter the operating data of the wind turbine unit, and obtain the actual power curve of the wind turbine unit based on the operating data; wherein, the operating data includes at least wind speed, power generation, blade angle and ambient temperature; The identification module is used to identify abnormal scatter points in the actual power curve that deviate from the standard power curve based on a density clustering algorithm. The second processing module is used to match the distribution characteristics of the abnormal scattered points and the corresponding operating data with a preset fault type rule base to determine the root cause of the fault. The output module is used to output an analysis report and improvement measures based on the root causes of the fault.

8. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method described in any one of claims 1-6.

Images