Wind power abnormality detection method and device, electronic equipment and storage medium

Abstract

The application relates to a wind power abnormality detection method and device, electronic equipment and a storage medium. The method comprises the following steps: determining a neighbor wind turbine set of each wind turbine, screening SCADA data of each wind turbine to obtain initial screened data, obtaining target wind turbine data and neighbor wind turbine data corresponding to the target wind turbine in a target time window, constructing and iteratively sampling samples, establishing a power curve model, setting an inner point evaluation standard and an iteration termination condition, obtaining a target fitting model, calculating residual values of each data point in the initial screened data, comparing the residual values with a judgment threshold, and obtaining an abnormality detection result. Thus, the problems of high false detection and missed detection rates of abnormal data and low power curve fitting accuracy in the related art are solved, accurate identification of nonlinear power curve abnormalities is achieved through adaptive grouping and a multi-wind turbine cooperative PROSAC algorithm, and the operation and maintenance efficiency and power generation reliability of a wind farm are significantly improved.

Description

Technical Field

[0001] This application relates to the field of wind power generation technology, and in particular to a method, device, electronic equipment and storage medium for detecting abnormal wind power output. Background Technology

[0002] With the large-scale development of the wind power industry, wind turbine power curves have become the core basis for evaluating the operating status of units and guiding operation and maintenance. The data quality directly affects the power generation and operation and maintenance efficiency of wind farms. However, current anomaly detection methods mostly use global models to process SCADA data, which is difficult to adapt to complex wind farm environments and individual differences of wind turbines, resulting in insufficient detection accuracy and frequent false positives and false negatives.

[0003] In related technologies, traditional anomaly detection methods often ignore the power curve deviation of wind turbines caused by aging, terrain or turbulence, misjudging reasonable performance changes as faults. At the same time, linear fitting cannot accurately characterize highly nonlinear power characteristics, and rigid time windows make it difficult to distinguish between sudden failures and gradual degradation, which seriously affects the reliability of operation and maintenance decisions and urgently needs to be solved. Summary of the Invention

[0004] This application provides a wind power anomaly detection method, device, electronic equipment, and storage medium to solve the problems of high false detection and false negative rates and low power curve fitting accuracy in related technologies. It achieves accurate identification of nonlinear power curve anomalies through adaptive grouping and multi-wind turbine collaborative PROSAC algorithm, which significantly improves the operation and maintenance efficiency and power generation reliability of wind farms.

[0005] The first aspect of this application provides a method for detecting abnormal wind power output, including the following steps: Collect SCADA data for each wind turbine and determine the set of neighboring wind turbines for each wind turbine; The SCADA data of each wind turbine is filtered for outliers based on a preset filtering strategy to obtain the data after initial screening. The target time window is dynamically determined based on the power curve degradation rate of the target wind turbine, and the target wind turbine data and the neighboring wind turbine data corresponding to the target wind turbine are acquired within the target time window. Based on the target wind turbine data and the neighboring wind turbine data, the data after the initial screening is used to construct sample priorities and iteratively sample, and a power curve model is established using a preset nonlinear piecewise fitting strategy. The interior point evaluation criteria and iteration termination conditions are set to obtain the target fitting model. The residual value of each data point in the initial screening data is calculated using the target fitting model. The residual value of each data point is then compared with the judgment threshold in the inlier evaluation criterion to obtain the anomaly detection result.

[0006] According to one embodiment of this application, the step of constructing sample priorities and iteratively sampling the initially screened data based on the target wind turbine data and the neighboring wind turbine data includes: Calculate the consistency score between each data point in the initial screening data and the power mode of the neighboring wind turbines based on the target wind turbine data and the neighboring wind turbine data; Based on a preset sorting strategy, the data after initial screening is sorted according to the consistency score. From the sorted data, target consistency points of the current sample set size are selected sequentially as sample subsets to use the sample subsets for model fitting.

[0007] According to one embodiment of this application, the step of establishing a power curve model using a preset nonlinear piecewise fitting strategy, and setting an interior point evaluation criterion and an iteration termination condition to obtain a target fitting model includes: The wind speed is divided into a first wind speed range, a second wind speed range, and a third wind speed range. The first wind speed range is fitted using a quadratic polynomial model to obtain a first fitting result; the second wind speed range is fitted using a cubic polynomial model to obtain a second fitting result; and the third wind speed range is fitted using a rated power constant model to obtain a third fitting result. A power curve model is then established based on the first fitting result, the second fitting result, and the third fitting result. For each sample subset selected in each iteration, the residual value of each sample point relative to the current power curve model is calculated, and the interior point determination threshold is dynamically calculated based on the residual distribution of all sample points in the sample subset. Sample points whose residual value is less than or equal to the inlier determination threshold are determined as inliers, and sample points whose residual value is greater than the inlier determination threshold are determined as outliers. When the current iteration number is greater than or equal to the preset iteration number, or when the number of interior points of the current power curve model reaches the preset threshold, the iteration stops and the current power curve model is used as the target fitting model.

[0008] According to one embodiment of this application, comparing the residual value of each data point with the judgment threshold in the inlier evaluation criterion to obtain an anomaly detection result includes: Determine whether there are target data points with residual values ​​greater than the determination threshold; If there is a target data point with a residual value greater than the judgment threshold, the target data point will be judged as an abnormal data point, and the abnormal detection result will be generated based on the abnormal data point. The anomaly detection results include the timestamp, wind speed, power, residual value, and anomaly label for each anomaly data point.

[0009] According to one embodiment of this application, determining the set of neighboring wind turbines for each wind turbine includes: Obtain the latitude and longitude coordinates of all wind turbines within the wind farm; Based on the latitude and longitude coordinates of each wind turbine, a preset parameter density clustering algorithm is used to cluster all wind turbines in the wind farm to obtain at least one geographical cluster. Other wind turbines within the same geographical cluster, excluding the target wind turbine, are identified as the set of neighboring wind turbines of the target wind turbine.

[0010] According to one embodiment of this application, the SCADA data of each wind turbine includes at least one of timestamp, wind speed, power, wind turbine ID, and latitude and longitude coordinates.

[0011] According to the wind power anomaly detection method provided in this application, a set of neighboring wind turbines for each wind turbine is determined. The SCADA data of each wind turbine is screened to obtain initially screened data. Within a target time window, data of the target wind turbine and its corresponding neighboring wind turbines are acquired for sample priority construction and iterative sampling. A power curve model is established, and an interior point evaluation standard and iteration termination condition are set to obtain the target fitting model. The residual value of each data point in the initially screened data is calculated, and the residual value is compared with a judgment threshold to obtain the anomaly detection result. This solves the problems of high false positive and false negative rates and low power curve fitting accuracy in related technologies. Through adaptive grouping and the multi-wind turbine collaborative PROSAC algorithm, accurate identification of nonlinear power curve anomalies is achieved, significantly improving the operation and maintenance efficiency and power generation reliability of wind farms.

[0012] A second aspect of this application provides a wind power anomaly detection device, comprising: The data acquisition module is used to collect SCADA data for each wind turbine and determine the set of neighboring wind turbines for each wind turbine. The screening module is used to screen outliers in the SCADA data of each wind turbine based on a preset screening strategy to obtain the data after initial screening. The processing module is used to dynamically determine the target time window based on the power curve degradation rate of the target wind turbine, and to acquire the target wind turbine data and the neighboring wind turbine data corresponding to the target wind turbine within the target time window; The iteration module is used to construct sample priorities and iteratively sample the data after the initial screening based on the target wind turbine data and the neighboring wind turbine data, and to establish a power curve model using a preset nonlinear piecewise fitting strategy, and to set the interior point evaluation criteria and iteration termination conditions to obtain the target fitting model. The detection module is used to calculate the residual value of each data point in the initial screening data using the target fitting model, and compare the residual value of each data point with the judgment threshold in the in-point evaluation criterion to obtain the anomaly detection result.

[0013] According to one embodiment of this application, the iteration module is configured to: Calculate the consistency score between each data point in the initial screening data and the power mode of the neighboring wind turbines based on the target wind turbine data and the neighboring wind turbine data; Based on a preset sorting strategy, the data after initial screening is sorted according to the consistency score. From the sorted data, target consistency points of the current sample set size are selected sequentially as sample subsets to use the sample subsets for model fitting.

[0014] According to one embodiment of this application, the iteration module is configured to: The wind speed is divided into a first wind speed range, a second wind speed range, and a third wind speed range. The first wind speed range is fitted using a quadratic polynomial model to obtain a first fitting result; the second wind speed range is fitted using a cubic polynomial model to obtain a second fitting result; and the third wind speed range is fitted using a rated power constant model to obtain a third fitting result. A power curve model is then established based on the first fitting result, the second fitting result, and the third fitting result. For each sample subset selected in each iteration, the residual value of each sample point relative to the current power curve model is calculated, and the interior point determination threshold is dynamically calculated based on the residual distribution of all sample points in the sample subset. Sample points whose residual value is less than or equal to the inlier determination threshold are determined as inliers, and sample points whose residual value is greater than the inlier determination threshold are determined as outliers. When the current iteration number is greater than or equal to the preset iteration number, or when the number of interior points of the current power curve model reaches the preset threshold, the iteration stops and the current power curve model is used as the target fitting model.

[0015] According to one embodiment of this application, the detection module is used for: Determine whether there are target data points with residual values ​​greater than the determination threshold; If there is a target data point with a residual value greater than the judgment threshold, the target data point will be judged as an abnormal data point, and the abnormal detection result will be generated based on the abnormal data point. The anomaly detection results include the timestamp, wind speed, power, residual value, and anomaly label for each anomaly data point.

[0016] According to one embodiment of this application, the acquisition module is used for: Obtain the latitude and longitude coordinates of all wind turbines within the wind farm; Based on the latitude and longitude coordinates of each wind turbine, a preset parameter density clustering algorithm is used to cluster all wind turbines in the wind farm to obtain at least one geographical cluster. Other wind turbines within the same geographical cluster, excluding the target wind turbine, are identified as the set of neighboring wind turbines of the target wind turbine.

[0017] According to one embodiment of this application, the SCADA data of each wind turbine includes at least one of timestamp, wind speed, power, wind turbine ID, and latitude and longitude coordinates.

[0018] According to the wind power anomaly detection device provided in this application embodiment, the set of neighboring wind turbines for each wind turbine is determined, and the SCADA data of each wind turbine is screened to obtain the initially screened data. Within the target time window, the target wind turbine data and the corresponding neighboring wind turbine data are acquired to construct sample priorities and iteratively sample, and a power curve model is established. Internal point evaluation criteria and iteration termination conditions are set to obtain the target fitting model. The residual value of each data point in the initially screened data is calculated, and the residual value is compared with the judgment threshold to obtain the anomaly detection result. This solves the problems of high false positive and false negative rates and low power curve fitting accuracy in related technologies. Through adaptive grouping and multi-wind turbine collaborative PROSAC algorithm, accurate identification of nonlinear power curve anomalies is achieved, significantly improving the operation and maintenance efficiency and power generation reliability of wind farms.

[0019] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the wind power anomaly detection method as described in the above embodiments.

[0020] A fourth aspect of this application provides a computer-readable storage medium storing computer instructions for causing the computer to perform the wind power anomaly detection method as described in the above embodiments.

[0021] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0022] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a wind power anomaly detection method provided according to an embodiment of this application; Figure 2 This is a flowchart of a wind power anomaly detection method according to an embodiment of this application; Figure 3 The following is a detailed flowchart of a wind power anomaly detection algorithm based on cooperative PROSAC according to an embodiment of this application; Figure 4 The original power curve and abnormal data distribution diagram of wind turbine No. 10 according to an embodiment of this application; Figure 5 The original power curve and abnormal data distribution diagram of wind turbine No. 5 according to an embodiment of this application; Figure 6 This is a block diagram of a wind power anomaly detection device according to an embodiment of this application; Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0023] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0024] As those skilled in the art will understand, current wind farms mostly use SCADA systems to collect data such as wind speed and power for anomaly detection. However, existing technologies still have many key limitations and are difficult to adapt to complex wind farm environments and individual differences in wind turbines.

[0025] In related technologies, current wind power curve outlier detection typically processes the entire wind farm's SCADA data centrally, such as through global RANSAC or statistical methods, ignoring key differences between wind turbines. A single turbine may experience a decrease in power output due to aging or insufficient maintenance, while environmental factors such as terrain undulations, turbulence intensity, and wind shear can also cause power curve shifts. This global model's averaging process smooths out individual characteristics, misjudging normal degradation as anomalies, or treating reasonable location-related shifts as faults.

[0026] In addition, traditional models often use linear fitting to describe nonlinear wind power curves, which have insufficient fitting accuracy and rigid time window settings, making it difficult to distinguish between short-term anomalies such as sensor failures and gradual degradation. These problems lead to a high false detection rate and false negative rate for anomaly detection, which not only increases ineffective operation and maintenance costs, but may also cause equipment damage and power generation loss due to missed fault detection.

[0027] To address the aforementioned issues, this invention proposes a method for detecting anomalous wind power curve data by integrating an adaptive mechanism with multi-turbine collaboration, based on existing artificial intelligence technologies. This invention detects anomalous values ​​through multi-turbine collaboration, preserving the characteristics of individual turbine power curve degradation while also drawing on the power curve patterns of surrounding healthy turbines. It integrates a collaborative mechanism between turbines into PROSAC to resolve the discrepancies in power anomaly data among turbines. By using adaptive DBSCAN clustering to fuse turbine grouping logic based on geographical distance and wind resource correlation, and optimizing the sample selection strategy of the PROSAC algorithm, it accurately captures the nonlinear characteristics of wind power curves and individual turbine differences, improving adaptability and anomaly detection accuracy across different wind farm scenarios, ultimately providing reliable decision support for wind farm operation and maintenance.

[0028] The following description, with reference to the accompanying drawings, describes a wind power anomaly detection method, apparatus, electronic device, and storage medium according to embodiments of this application.

[0029] Specifically, Figure 1 This is a flowchart illustrating a wind power anomaly detection method provided in an embodiment of this application.

[0030] like Figure 1 As shown, the wind power anomaly detection method includes the following steps: In step S101, SCADA data for each wind turbine is collected, and the set of neighboring wind turbines for each wind turbine is determined.

[0031] In some embodiments, the SCADA data for each wind turbine includes at least one of the following: timestamp, wind speed, power, turbine ID, and latitude and longitude coordinates.

[0032] Specifically, this application embodiment can collect multi-dimensional time-series data generated during the operation of each wind turbine through the existing Supervisory Control and Data Acquisition (SCADA) system of the wind farm, with a sampling frequency of 10 minutes. In addition, depending on the actual deployment, auxiliary parameters such as wind direction, speed, pitch angle, temperature, and vibration can be collected to enhance the contextual awareness capability of anomaly detection.

[0033] Furthermore, static attribute data of the wind field are collected simultaneously, including wind field range, number of wind turbines, and historical average turbulence intensity, for subsequent adaptive parameter calculation.

[0034] Furthermore, in some embodiments, determining the set of neighboring wind turbines for each wind turbine includes: obtaining the latitude and longitude coordinates of all wind turbines in the wind farm; clustering all wind turbines in the wind farm based on the latitude and longitude coordinates of each wind turbine using a preset parameter density clustering algorithm to obtain at least one geographical cluster; and determining the other wind turbines in the same geographical cluster, excluding the target wind turbine, as the set of neighboring wind turbines of the target wind turbine.

[0035] Specifically, in this application embodiment, based on wind turbine coordinates and static attributes, adaptive parameter DBSCAN clustering is used to divide the wind field into multiple geographical clusters, and the key DBSCAN parameters, neighborhood radius eps and minimum number of samples min_samples, are dynamically calculated.

[0036] The adaptive calculation domain radius eps is obtained by calculating the geographical span of the wind field and the wind turbine density. (1) (2) in, The geographical span of the wind field is determined by the maximum value of either the east-west or north-south span. For fan density, This represents the total number of wind turbines in the wind farm. The wind field area.

[0037] Adaptive calculation of min_samples requires satisfying the principles of sufficient neighbor reference sample size and avoiding misclustering of outliers to calculate the minimum effective number of fans within a cluster. (3) in, This is the floor symbol.

[0038] The wind turbines are clustered using DBSCAN with adaptive parameter optimization. For each wind turbine... Obtain the neighbor set This step defines meaningful neighbors for each wind turbine, ensuring that turbines within a geographical cluster are affected by similar wind resources and have comparable power curve patterns; at the same time, inter-cluster differences in terrain are isolated, avoiding erroneous referencing.

[0039] In step S102, outlier values ​​are filtered out from the SCADA data of each wind turbine based on a preset filtering strategy to obtain the data after initial screening.

[0040] Specifically, this application embodiment requires initial screening of outliers to remove invalid data, such as outliers with wind speed <0, power <0, or power > 110% of rated power. Wind speed is divided into bins with a width of 0.5 m / s for cross-wind turbine comparison. Simultaneously, a set of physical rules related to wind turbine operation is used to eliminate outliers that clearly violate operating principles.

[0041] For example, data points that meet the following rules are initially identified as outliers: (4) (5) (6) (7) in, For wind speed, For power, This refers to the cutoff wind speed. In actual wind turbine operation, negative wind speed or negative power output violates basic physical principles, therefore such abnormal data needs to be eliminated.

[0042] In step S103, the target time window is dynamically determined based on the power curve degradation rate of the target wind turbine, and the target wind turbine data and the neighboring wind turbine data corresponding to the target wind turbine are acquired within the target time window.

[0043] Specifically, PROSAC accelerates robust model fitting by prioritizing highly consistent samples, but random sample selection is susceptible to being misled by individual outliers. Integrating neighbor information into the PROSAC framework allows sample selection to prioritize points "consistent with their surroundings" while preserving individual characteristics. Therefore, a collaborative PROSAC improvement is implemented for multiple wind turbines.

[0044] Furthermore, the size of the dynamic time window is calculated. This can be based on the power curve degradation rate of the target wind turbine over the past 90 days. Determine the time window. The power curve degradation rate is the percentage difference between the average monthly power and the initial power at rated wind speed; determine the time window accordingly. The calculation method is as follows: (8) If the power curve degradation rate For the initial calculation, the average degradation rate of the same type of wind turbine in the wind farm was used as the initial value. The power curve degradation rate was then updated every 7 days thereafter. Adjust the time window synchronously .

[0045] For each wind turbine, within the current time window Obtain target data ,in for wind speed, For the corresponding power; at the same time, obtain synchronous data of neighboring wind turbines. , Let m be the ID of the neighboring wind turbine, and m be the number of neighbors.

[0046] In step S104, the data after initial screening is used to construct sample priorities and iteratively sample based on the target wind turbine data and neighboring wind turbine data. A power curve model is established using a preset nonlinear piecewise fitting strategy, and the interior point evaluation criteria and iteration termination conditions are set to obtain the target fitting model.

[0047] Furthermore, in some embodiments, sample priority construction and iterative sampling are performed on the initially screened data based on the target wind turbine data and neighboring wind turbine data, including: calculating the consistency score between each data point in the initially screened data and the power mode of the neighboring wind turbines based on the target wind turbine data and neighboring wind turbine data; sorting the initially screened data according to the consistency score based on a preset sorting strategy, and sequentially selecting target consistency points of the current sample set size from the sorted data as sample subsets, so as to use the sample subsets for model fitting.

[0048] Specifically, this application embodiment constructs PROSAC sample priorities.

[0049] First, calculate the neighbor reference power. For each point in the target turbine data... Find the wind speed bin to which it belongs, calculate the average power of all neighboring wind turbines within that bin during the same period, and use this average power as the neighboring reference power. : (9) in, This represents the number of data points for a single neighboring wind turbine within that wind speed bin.

[0050] Secondly, based on the average turbulence intensity of the wind field over the past 30 days Dynamically calculate and adjust consistency scores : (10) (11) (12) in, The standard deviation of wind speed over 10 minutes. The average wind speed over 10 minutes. Consistency score. The higher the value, the more consistent the point is with its neighboring patterns; the lower the value, the more likely it is to be an anomaly. This wind turbine data point is then categorized according to... Arrange in descending order to prioritize points with high consistency.

[0051] For degraded fans, if the overall power is low, Evergrande The values ​​are generally low, but PROSAC still fits its own curve; for healthy fans, if a certain point... A sudden decrease is easily identified as abnormal.

[0052] Furthermore, in some embodiments, a power curve model is established using a preset nonlinear piecewise fitting strategy, and an interior point evaluation criterion and an iteration termination condition are set to obtain a target fitting model. This includes: dividing the wind speed into a first wind speed interval, a second wind speed interval, and a third wind speed interval; fitting the first wind speed interval using a quadratic polynomial model to obtain a first fitting result; fitting the second wind speed interval using a cubic polynomial model to obtain a second fitting result; fitting the third wind speed interval using a rated power constant model to obtain a third fitting result; establishing a power curve model based on the first fitting result, the second fitting result, and the third fitting result; for each sample subset selected in each iteration, calculating the residual value of each sample point relative to the current power curve model, and dynamically calculating the interior point determination threshold based on the residual distribution of all sample points in the sample subset; determining sample points with residual values ​​less than or equal to the interior point determination threshold as interior points, and determining sample points with residual values ​​greater than the interior point determination threshold as exterior points; stopping the iteration and using the current power curve model as the target fitting model when the current iteration number is greater than or equal to the preset iteration number, or when the number of interior points in the current power curve model reaches a preset number threshold.

[0053] Specifically, the embodiments of this application are based on the forced priority iteration of PROSAC in cooperation with neighboring wind turbines. This step changes the random sampling of the original PROSAC standard, forcing PROSAC to prioritize high-priority turbines. The points are composed of initialization, incremental sampling, and iterative updates.

[0054] Initialization, initialize sample size Iterator counter ; Progressive sampling, selecting the top from the sorted queue Tall Points, as the current sample set ; Iterative updates; if the current number of points in the model does not meet the termination condition, , Repeat sampling.

[0055] Furthermore, a nonlinear piecewise fitting power curve model is employed. Considering the nonlinear characteristics of the wind power curve, a piecewise fitting strategy is used to cover the entire wind speed range. In this embodiment, the wind speed can be divided into three ranges: (13) For the low wind speed section, a quadratic polynomial model is used; for the rated upstream section, a cubic polynomial model is used; and for the rated downstream section, a rated power constant model is used.

[0056] Furthermore, establish internal point evaluation criteria. For each data point... Based on the wind speed range, select the corresponding piecewise model and calculate the fitted power. residual Based on the sample set residual standard deviation Combined with turbulence intensity Adaptive threshold adjustment : (14) Wherein, initial value , This represents the mean of the residuals in the sample set. It is updated every 3 iterations. To ensure that the threshold is adapted to the characteristics of wind field fluctuations Interior point determination is: when the residual If the condition is met, then the point is determined to be an interior point; otherwise, the point is determined to be an exterior point.

[0057] Furthermore, termination conditions are set. For example, iteration stops when any of the following conditions are met, and the current model is the final model, i.e., the target fitted model.

[0058] (15) In step S105, the residual value of each data point in the initial screening data is calculated using the target fitting model. The residual value of each data point is compared with the judgment threshold in the in-point evaluation standard to obtain the anomaly detection result.

[0059] Furthermore, in some embodiments, the residual value of each data point is compared with the judgment threshold in the internal point evaluation criterion to obtain an anomaly detection result, including: determining whether there is a target data point with a residual value greater than the judgment threshold; if there is a target data point with a residual value greater than the judgment threshold, the target data point is judged as an anomaly data point, and an anomaly detection result is generated based on the anomaly data point; wherein, the anomaly detection result includes the timestamp, wind speed, power, residual value, and anomaly label corresponding to each anomaly data point.

[0060] Specifically, the residuals of all non-initial screening outliers are calculated. Combined with dynamic threshold Perform outlier detection. Normal value; otherwise, abnormal value. Final output fan. The list of outliers includes timestamps, wind speed, power, residual values, and outlier annotations.

[0061] To facilitate a clearer and more intuitive understanding of the wind power anomaly detection method of this application by those skilled in the art, the following is combined with... Figures 2-5 Please provide a detailed explanation.

[0062] like Figure 2As shown, the overall implementation steps of this PROSAC-based multi-wind turbine collaborative wind power anomaly data detection method include four main stages: The first stage is data preprocessing and wind turbine grouping. By collecting SCADA data, and based on the wind turbine coordinates and static attributes, the adaptive parameter DBSCAN clustering algorithm is used to spatially group all wind turbines in the wind farm.

[0063] The second stage is the initial screening of outliers, which aims to remove invalid data that exceeds the normal data limit, while also using physical rules to remove outliers that clearly violate the operating principles.

[0064] The third stage is the core processing step, namely the PROSAC sample selection and model fitting based on collaboration. Specifically, it includes first determining the time window, calculating the consistency score of the neighbor reference power, and constructing the PROSAC sample priority according to the consistency score. Then, iterating the priority based on the neighbor wind turbine collaborative PROSAC, then fitting the power curve model in a nonlinear piecewise manner, and finally setting the internal point evaluation criteria and termination conditions.

[0065] The fourth stage involves outlier detection based on the model fitting results. The final established power curve model is used to perform residual analysis on the data, thereby identifying and marking outlier data points that deviate from the model.

[0066] Furthermore, such as Figure 3 As shown in the embodiment of this application, the specific execution logic of the wind power anomaly detection algorithm based on collaborative PROSAC is as follows: First, SCADA data of each wind turbine is collected. Then, adaptive parameter DBSCAN clustering is used to divide the wind field into multiple geographical clusters to obtain the neighbor set of a single wind turbine. Anomalies are initially screened through boundary value checks and physical rules.

[0067] Furthermore, the collaborative PROSAC sample selection and model fitting process sequentially includes calculating the dynamic time window size, constructing PROSAC sample priorities, iterating based on neighboring wind turbine collaborative PROSAC priority, fitting the power curve model nonlinearly, setting the interior point evaluation criteria, and setting termination conditions. After completing the model fitting, the process proceeds to the judgment stage, based on the formula... The system determines the attributes of the current data point. If the conditions are met, it is determined to be a normal value; otherwise, it is determined to be an abnormal value. Finally, the detection result is output.

[0068] For example, such as Figure 4 and Figure 5 As shown, the scatter plots illustrate the power-wind speed relationship between wind turbines No. 10 and No. 5 during operation. Figure 4As shown in the figure, the solid black dots represent data points identified as normal values, which are closely clustered near the theoretical power curve, reflecting the typical output characteristics of the wind turbine in a healthy state. The asterisks represent the detected outlier data points, which deviate significantly from the main curve and are distributed in the low-power region, the high-wind-speed zero-power region, or the region with severe power fluctuations. This indicates that there may be problems such as sensor failure, control failure, or environmental interference. The figure intuitively presents the spatial distribution characteristics of outliers in the original data, verifying the effectiveness of the detection method in modeling the nonlinear power curve of a single wind turbine and identifying outliers.

[0069] like Figure 5 As shown, the normal values ​​of wind turbine No. 5 are also densely distributed along the S-shaped power curve, while the abnormal values ​​are widely scattered above and below the curve in the form of asterisks. In particular, there are a large number of discrete points with power below the rated power in the medium and high wind speed range, and some are even close to zero power, indicating that the wind turbine may have potential problems such as blade contamination, yaw error or reduced transmission system efficiency.

[0070] By comparison Figure 4 and Figure 5 It can be seen that different wind turbines have different abnormal modes due to differences in geographical location, equipment aging or maintenance conditions. This also highlights the necessity of the present invention to use multi-wind turbine collaboration and adaptive grouping strategy for personalized modeling, so as to avoid the risk of misjudgment caused by a globally unified threshold.

[0071] Therefore, based on the wind power curve anomaly data detection method proposed in this application, which integrates adaptive mechanisms and multi-turbine collaboration, the method achieves accurate capture of the nonlinear characteristics of wind power curves and individual differences of wind turbines by adjusting DBSCAN clustering parameters adaptively, incorporating wind turbine grouping logic that integrates geographical distance and wind resource correlation, using a dynamic time window nonlinear piecewise fitting model, and optimizing the sample selection strategy of the PROSAC algorithm. This effectively distinguishes between short-term power surges and long-term gradual degradation, providing reliable decision support for wind farm operation and maintenance. Its beneficial effects include at least the following: (1) Each wind turbine independently fits the power curve to avoid global averaging that obliterates individual characteristics; the new DBSCAN parameter adaptive and wind resource correlation verification ensures the reliability of neighbor reference and further improves the accuracy of individual characteristic identification.

[0072] (2) Utilizing the power patterns of geographically nearby wind turbines as prior knowledge, combined with adaptive Adjustments are made to reduce the interference of environmental factors such as turbulence on the consistency score and improve robustness to outliers.

[0073] (3) Introduce dynamic time windows and piecewise nonlinear fitting. Dynamic windows adapt to the wind turbine degradation state and distinguish between short-term anomalies and long-term degradation; piecewise fitting accurately describes the power relationship across the entire wind speed range and reduces residual calculation bias.

[0074] (4) Based on the progressive characteristics and parameter adaptation of PROSAC, only local data and dynamic window data are needed, reducing computational overhead. At the same time, it adapts to different terrain wind fields and wind turbine states, improving versatility.

[0075] According to the wind power anomaly detection method proposed in this application, a set of neighboring wind turbines is determined for each wind turbine. The SCADA data of each wind turbine is screened to obtain initially screened data. Within a target time window, data of the target wind turbine and its corresponding neighboring wind turbines are acquired for sample priority construction and iterative sampling. A power curve model is established, and an interior point evaluation standard and iteration termination condition are set to obtain the target fitting model. The residual value of each data point in the initially screened data is calculated, and the residual value is compared with a judgment threshold to obtain the anomaly detection result. This solves the problem in traditional wind power anomaly data cleaning, which fails to consider nonlinear characteristics and individual wind turbine differences, leading to high false positive and false negative rates and low power curve fitting accuracy. Through adaptive grouping and the multi-wind turbine collaborative PROSAC algorithm, accurate identification of nonlinear power curve anomalies is achieved, significantly improving wind farm operation and maintenance efficiency and power generation reliability.

[0076] Next, the wind power anomaly detection device according to the embodiments of this application is described with reference to the accompanying drawings.

[0077] Figure 6 This is a block diagram of a wind power anomaly detection device according to an embodiment of this application.

[0078] like Figure 6 As shown, the wind power anomaly detection device 10 includes: a data acquisition module 100, a screening module 200, a processing module 300, an iteration module 400, and a detection module 500.

[0079] The system comprises the following modules: Acquisition module 100, which acquires SCADA data for each wind turbine and determines the set of neighboring wind turbines for each turbine; Screening module 200, which performs outlier screening on the SCADA data of each wind turbine based on a preset screening strategy to obtain the initially screened data; Processing module 300, which dynamically determines the target time window based on the power curve degradation rate of the target wind turbine and acquires the target wind turbine data and the corresponding neighboring wind turbine data within the target time window; Iteration module 400, which constructs sample priorities and iteratively samples the initially screened data based on the target wind turbine data and the neighboring wind turbine data, and establishes a power curve model using a preset nonlinear piecewise fitting strategy, and sets the interior point evaluation criteria and iteration termination conditions to obtain the target fitting model; and Detection module 500, which calculates the residual value of each data point in the initially screened data using the target fitting model, compares the residual value of each data point with the judgment threshold in the interior point evaluation criteria, and obtains the anomaly detection result.

[0080] Furthermore, in some embodiments, the iteration module 400 is used to: calculate the consistency score between each data point in the initially screened data and the power mode of the neighboring wind turbines based on the target wind turbine data and the neighboring wind turbine data; sort the initially screened data according to the consistency score based on a preset sorting strategy, and sequentially select the target consistency points of the current sample set size from the sorted data as a sample subset, so as to use the sample subset for model fitting.

[0081] Further, in some embodiments, the iteration module 400 is used to: divide the wind speed into a first wind speed interval, a second wind speed interval, and a third wind speed interval; fit the first wind speed interval using a quadratic polynomial model to obtain a first fitting result; fit the second wind speed interval using a cubic polynomial model to obtain a second fitting result; fit the third wind speed interval using a rated power constant model to obtain a third fitting result; establish a power curve model based on the first fitting result, the second fitting result, and the third fitting result; for each sample subset selected in each iteration, calculate the residual value of each sample point relative to the current power curve model, and dynamically calculate the interior point determination threshold based on the residual distribution of all sample points in the sample subset; determine sample points with residual values ​​less than or equal to the interior point determination threshold as interior points, and determine sample points with residual values ​​greater than the interior point determination threshold as exterior points; stop the iteration and use the current power curve model as the target fitting model when the current iteration number is greater than or equal to the preset iteration number, or when the number of interior points of the current power curve model reaches a preset number threshold.

[0082] Furthermore, in some embodiments, the detection module 500 is used to: determine whether there is a target data point with a residual value greater than the determination threshold; if there is a target data point with a residual value greater than the determination threshold, then the target data point is determined to be an abnormal data point, and an abnormal detection result is generated based on the abnormal data point; wherein, the abnormal detection result includes the timestamp, wind speed, power, residual value and abnormal label corresponding to each abnormal data point.

[0083] Furthermore, in some embodiments, the acquisition module 100 is used to: acquire the latitude and longitude coordinates of all wind turbines in the wind farm; based on the latitude and longitude coordinates of each wind turbine, cluster all wind turbines in the wind farm using a preset parameter density clustering algorithm to obtain at least one geographical cluster; and determine the other wind turbines in the same geographical cluster, excluding the target wind turbine, as the set of neighboring wind turbines of the target wind turbine.

[0084] According to one embodiment of this application, the SCADA data for each wind turbine includes at least one of timestamp, wind speed, power, turbine ID, and latitude and longitude coordinates.

[0085] It should be noted that the foregoing explanation of the wind power anomaly detection method embodiment also applies to the wind power anomaly detection device of this embodiment, and will not be repeated here.

[0086] According to the wind power anomaly detection device proposed in this application, the set of neighboring wind turbines for each wind turbine is determined, and the SCADA data of each wind turbine is screened to obtain the initially screened data. Within the target time window, the target wind turbine data and the corresponding neighboring wind turbine data are acquired to construct sample priorities and iteratively sample, and a power curve model is established. Internal point evaluation criteria and iteration termination conditions are set to obtain the target fitting model. The residual value of each data point in the initially screened data is calculated, and the residual value is compared with the judgment threshold to obtain the anomaly detection result. This solves the problems of high false positive and false negative rates and low power curve fitting accuracy in related technologies. Through adaptive grouping and multi-wind turbine collaborative PROSAC algorithm, accurate identification of nonlinear power curve anomalies is achieved, significantly improving the operation and maintenance efficiency and power generation reliability of wind farms.

[0087] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 701, the processor 702, and the computer program stored on the memory 701 and executable on the processor 702.

[0088] When the processor 702 executes the program, it implements the wind power anomaly detection method provided in the above embodiments.

[0089] Furthermore, electronic devices also include: Communication interface 703 is used for communication between memory 701 and processor 702.

[0090] The memory 701 is used to store computer programs that can run on the processor 702.

[0091] The memory 701 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0092] If the memory 701, processor 702, and communication interface 703 are implemented independently, then the communication interface 703, memory 701, and processor 702 can be interconnected via a bus to complete communication between them. 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 into address buses, data buses, control buses, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0093] Optionally, in a specific implementation, if the memory 701, processor 702, and communication interface 703 are integrated on a single chip, then the memory 701, processor 702, and communication interface 703 can communicate with each other through an internal interface.

[0094] The processor 702 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0095] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described wind power anomaly detection method.

[0096] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0097] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0098] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0099] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0100] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0101] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0102] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0103] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for detecting abnormal wind power output, characterized in that, Includes the following steps: Collect SCADA data for each wind turbine and determine the set of neighboring wind turbines for each wind turbine; The SCADA data of each wind turbine is filtered for outliers based on a preset filtering strategy to obtain the data after initial screening. The target time window is dynamically determined based on the power curve degradation rate of the target wind turbine, and the target wind turbine data and the neighboring wind turbine data corresponding to the target wind turbine are acquired within the target time window. Based on the target wind turbine data and the neighboring wind turbine data, the data after the initial screening is used to construct sample priorities and iteratively sample, and a power curve model is established using a preset nonlinear piecewise fitting strategy. The interior point evaluation criteria and iteration termination conditions are set to obtain the target fitting model. The residual value of each data point in the initial screening data is calculated using the target fitting model. The residual value of each data point is then compared with the judgment threshold in the inlier evaluation criterion to obtain the anomaly detection result.

2. The method according to claim 1, characterized in that, The step of constructing sample priorities and iteratively sampling the initially screened data based on the target wind turbine data and the neighboring wind turbine data includes: Calculate the consistency score between each data point in the initial screening data and the power mode of the neighboring wind turbines based on the target wind turbine data and the neighboring wind turbine data; Based on a preset sorting strategy, the data after initial screening is sorted according to the consistency score. From the sorted data, target consistency points of the current sample set size are selected sequentially as sample subsets to use the sample subsets for model fitting.

3. The method according to claim 2, characterized in that, The process of establishing a power curve model using a preset nonlinear piecewise fitting strategy, and setting interior point evaluation criteria and iteration termination conditions to obtain the target fitting model includes: The wind speed is divided into a first wind speed range, a second wind speed range, and a third wind speed range. The first wind speed range is fitted using a quadratic polynomial model to obtain a first fitting result; the second wind speed range is fitted using a cubic polynomial model to obtain a second fitting result; and the third wind speed range is fitted using a rated power constant model to obtain a third fitting result. A power curve model is then established based on the first fitting result, the second fitting result, and the third fitting result. For each sample subset selected in each iteration, the residual value of each sample point relative to the current power curve model is calculated, and the interior point determination threshold is dynamically calculated based on the residual distribution of all sample points in the sample subset. Sample points whose residual value is less than or equal to the inlier determination threshold are determined as inliers, and sample points whose residual value is greater than the inlier determination threshold are determined as outliers. When the current iteration number is greater than or equal to the preset iteration number, or when the number of interior points of the current power curve model reaches the preset threshold, the iteration stops and the current power curve model is used as the target fitting model.

4. The method according to claim 1, characterized in that, The step of comparing the residual value of each data point with the judgment threshold in the internal point evaluation criterion to obtain the anomaly detection result includes: Determine whether there are target data points with residual values ​​greater than the determination threshold; If there is a target data point with a residual value greater than the judgment threshold, the target data point will be judged as an abnormal data point, and the abnormal detection result will be generated based on the abnormal data point. The anomaly detection results include the timestamp, wind speed, power, residual value, and anomaly label for each anomaly data point.

5. The method according to claim 1, characterized in that, Determining the set of neighboring wind turbines for each wind turbine includes: Obtain the latitude and longitude coordinates of all wind turbines within the wind farm; Based on the latitude and longitude coordinates of each wind turbine, a preset parameter density clustering algorithm is used to cluster all wind turbines in the wind farm to obtain at least one geographical cluster. Other wind turbines within the same geographical cluster, excluding the target wind turbine, are identified as the set of neighboring wind turbines of the target wind turbine.

6. The method according to claim 1, characterized in that, The SCADA data for each wind turbine includes at least one of the following: timestamp, wind speed, power, turbine ID, and latitude and longitude coordinates.

7. A wind power anomaly detection device, characterized in that, include: The data acquisition module is used to collect SCADA data for each wind turbine and determine the set of neighboring wind turbines for each wind turbine. The screening module is used to screen outliers in the SCADA data of each wind turbine based on a preset screening strategy to obtain the data after initial screening. The processing module is used to dynamically determine the target time window based on the power curve degradation rate of the target wind turbine, and to acquire the target wind turbine data and the neighboring wind turbine data corresponding to the target wind turbine within the target time window; The iteration module is used to construct sample priorities and iteratively sample the data after the initial screening based on the target wind turbine data and the neighboring wind turbine data, and to establish a power curve model using a preset nonlinear piecewise fitting strategy, and to set the interior point evaluation criteria and iteration termination conditions to obtain the target fitting model. The detection module is used to calculate the residual value of each data point in the initial screening data using the target fitting model, and compare the residual value of each data point with the judgment threshold in the in-point evaluation criterion to obtain the anomaly detection result.

8. The apparatus according to claim 7, characterized in that, The iterative module is used for: Calculate the consistency score between each data point in the initial screening data and the power mode of the neighboring wind turbines based on the target wind turbine data and the neighboring wind turbine data; Based on a preset sorting strategy, the data after initial screening is sorted according to the consistency score. From the sorted data, target consistency points of the current sample set size are selected sequentially as sample subsets to use the sample subsets for model fitting.

9. An electronic device, characterized in that, include: The method includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to implement the wind power anomaly detection method as described in any one of claims 1-6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is executed by a processor to implement the wind power anomaly detection method as described in any one of claims 1-6.

Images