Data mining identification method and system for high-temperature power limitation of wind turbine generator

Abstract

The invention provides a data mining identification method and system for high-temperature power limitation of a wind turbine generator, and relates to the technical field of fault monitoring of the wind turbine generator, and the method comprises the steps: obtaining related characteristic variables of the wind turbine generator, and carrying out the aggregation processing; preprocessing the available wind turbine generator feature data set; power is used as working condition interval division, a box plot method is used for conducting blade angle normal interval calculation on the standard wind turbine generator feature data set, and unit high-temperature power limiting recognition is conducted; and calculating normal distribution conditions of gear box oil temperature and generator temperature for the standard wind turbine generator characteristic data set by using a statistical analysis method, and performing unit high-temperature power-limiting identification based on the first unit high-temperature power-limiting identification result and the multi-working-condition temperature abnormal distribution threshold. According to the method and the device, the technical problem of high false alarm rate of high-temperature power limitation caused by adopting a fixed threshold to judge the state of the component and lacking working condition correlation analysis in the prior art is solved, and the accuracy of high-temperature power limitation identification is improved.

Description

Technical Field

[0001] This application relates to the field of wind turbine fault monitoring technology, specifically to a data mining and identification method and system for wind turbines facing high-temperature power limitation. Background Technology

[0002] Due to the randomness of wind and the harshness of the environment, the operating status of components such as wind turbine gearboxes and generators is highly uncertain. Especially as the unit's operating time increases, component aging and poor lubrication become more frequent, sometimes leading to persistently high gearbox oil and generator bearing temperatures, thus limiting the unit's output—a phenomenon known as reduced power operation. Currently, wind farms primarily rely on pre-set fixed temperature thresholds in monitoring systems for alarms, combined with manual analysis of historical data by maintenance personnel to determine if power curtailment is caused by high temperatures. However, fixed thresholds cannot accurately reflect the normal range of component temperatures dynamically changing with ambient temperature, unit load, and other operating conditions. This results in frequent false alarms under high load and high ambient temperature conditions, while early anomalies may be missed under low temperature conditions, leading to insufficient accuracy in monitoring results. Furthermore, power curtailment events and high-temperature alarms are isolated at the data level, lacking automated causal correlation and diagnostic logic. Relying on manual experience for post-event analysis is not only inefficient and slow to respond, but also results in highly subjective diagnostic conclusions, making it difficult to achieve accurate root cause location and early warning.

[0003] In summary, the existing technology has a technical problem: the use of fixed thresholds to determine component status and the lack of operational condition correlation analysis lead to a high false alarm rate for high-temperature power rationing, which further affects the operating efficiency of wind turbine units. Summary of the Invention

[0004] The purpose of this application is to provide a data mining identification method and system for high-temperature power curtailment of wind turbines, in order to solve the technical problem in the prior art that the use of fixed thresholds to judge the status of components and the lack of working condition correlation analysis leads to a high false alarm rate for high-temperature power curtailment, which further affects the operating efficiency of wind turbines.

[0005] To achieve the above objectives, this application provides a data mining and identification method and system for high-temperature power limitation of wind turbine units.

[0006] Firstly, this application provides a data mining and identification method for wind turbine power curtailment. This method is implemented through a data mining and identification system for wind turbine power curtailment. The method includes: obtaining wind turbine-related feature variables from the Kairosdb time-series database of a centralized SCADA system; aggregating these feature variables to obtain a usable wind turbine feature dataset; and preprocessing the usable wind turbine feature dataset using a combination of wind turbine operating mechanisms and the DBSCAN algorithm to obtain a standard wind turbine feature dataset. Using power as the operating condition division, the standard wind turbine characteristic dataset is analyzed using box plots to calculate the normal range of blade angles, resulting in multi-condition blade angle anomaly thresholds. Based on these thresholds, high-temperature power curtailment identification is performed to obtain the first unit's high-temperature power curtailment identification result. Using power and ambient temperature as operating condition divisions, statistical analysis is used to calculate the normal distribution of gearbox oil temperature and generator temperature in the standard wind turbine characteristic dataset, resulting in multi-condition temperature anomaly distribution thresholds. Based on the first unit's high-temperature power curtailment identification result and the multi-condition temperature anomaly distribution thresholds, high-temperature power curtailment identification is performed to determine the second unit's high-temperature power curtailment identification result.

[0007] Optionally, the relevant characteristic variables of the wind turbine unit specifically include: unit load, ambient temperature, wind speed, first blade angle, second blade angle, third blade angle, generator bearing temperature, and gearbox oil temperature.

[0008] Optionally, the relevant feature variables of the wind turbine are aggregated and time-aligned to obtain standardized wind turbine feature variable data; the standardized wind turbine feature variable data are then structured and transformed at preset time intervals to obtain a usable wind turbine feature dataset.

[0009] Optionally, based on the unit's standard power curve, abnormal data points in the available wind turbine characteristic dataset are deleted: , , ,in, For wind speed, Unit cut-in wind speed, Cut off the wind speed for the unit. For the unit load; based on the variation law of the unit blade angle, retain the centralized and standardized data points of the available wind turbine characteristic dataset: and , and , and , ,in, The rated power of the unit, The blade angle of the wind turbine unit; the available wind turbine feature dataset is binned, with each wind speed range defined as follows: Using intervals, we obtain the dataset within each small interval. for: Then, the DBSCAN algorithm is used for cleaning to obtain a standard wind turbine feature dataset; among which, Represents the entire dataset. Indicates the first A dataset of intervals, For the first The elements of interval n, where n is the dataset. Medium sample size.

[0010] Optionally, the standard wind turbine feature dataset is divided into buckets based on power as the operating condition interval, with each power interval defined as follows: Using intervals, we obtain the dataset within each small interval. for: , For each power range dataset, the box plot method is used to calculate the quartiles of the blade angle data within each range. According to the box plot anomaly detection rules, outliers in the blade angle data within each range are those greater than the 3 / 4 quartile plus 1.5 times the interquartile range. The multi-condition blade angle anomaly threshold is thus obtained, defined as follows: ,in, For the first Blade angle thresholds within a power range , For the first The 1 / 4 and 3 / 4 quantiles of the blade angle data for each power range are used to store the power range data and the multi-condition blade angle anomaly thresholds in the corresponding database.

[0011] Optionally, real-time data of the generator set is collected, and the target power range corresponding to the generator set power in the database is searched based on the generator set power in the real-time data. It is then determined whether the current blade angle is greater than the multi-condition blade angle abnormality threshold corresponding to the target power range, and the first generator set high-temperature power limitation identification result is obtained.

[0012] Optionally, the standard wind turbine characteristic dataset is divided into buckets based on power and ambient temperature as operating conditions, and the ambient temperature data within each power range is further divided into buckets. Using intervals, the dataset within each grid interval is obtained. for: , , ,in, For ambient temperature, This refers to the gearbox oil temperature or the generator bearing temperature. The number of divisions for the power range. This indicates the number of temperature intervals within each power range; calculate the mean and variance of the gearbox oil temperature or generator bearing temperature within each ambient temperature interval under each power range of the target unit, based on... In principle, the normal range of each temperature distribution should be within three standard deviations above and below its mean. The threshold values ​​for the first gearbox oil temperature or generator bearing temperature distribution are defined as follows: ,in, Indicates the first Each power range Indicates the first The power range under the first The first gearbox oil temperature or generator bearing temperature distribution threshold is determined by performing quartile statistics on the first gearbox oil temperature or generator bearing temperature distribution threshold. Following the box plot anomaly detection method, outliers in the temperature data within the grid interval are located in areas greater than the 3 / 4 quartile plus 1.5 times the interquartile range. This yields the second gearbox oil temperature or generator bearing temperature distribution threshold, defined as follows: ,in, For the first The power range under the first Gearbox oil temperature or generator bearing temperature within a specific ambient temperature range. , For the first The power range under the first The 1 / 4 and 3 / 4 quantiles of gearbox oil temperature or generator bearing temperature within a given ambient temperature range; based on the first gearbox oil temperature or generator bearing temperature distribution threshold and the second gearbox oil temperature or generator bearing temperature distribution threshold, determine the multi-condition temperature anomaly distribution threshold: The corresponding power range, ambient temperature range, and the multi-condition temperature anomaly distribution threshold are used to determine the appropriate power range, ambient temperature range, and the threshold values. Store it in the corresponding database.

[0013] Optionally, if the first unit's high-temperature power rationing identification result indicates that the unit is in a power rationing state, the target grid interval corresponding to the unit's load and ambient temperature is searched from the database based on the unit's load and ambient temperature; it is then determined whether the current gearbox oil temperature or generator bearing temperature is greater than the multi-condition temperature anomaly distribution threshold corresponding to the target grid interval, and the second unit's high-temperature power rationing identification result is determined.

[0014] Secondly, this application also provides a data mining and identification system for wind turbine power curtailment at high temperatures, used to execute the data mining and identification method for wind turbine power curtailment at high temperatures as described in the first aspect. The data mining and identification system for wind turbine power curtailment at high temperatures includes: an aggregation processing module for obtaining wind turbine-related feature variables from the Kairosdb time-series database of a centralized SCADA system, and performing aggregation processing on the wind turbine-related feature variables to obtain a usable wind turbine feature dataset; a data preprocessing module for preprocessing the usable wind turbine feature dataset based on the wind turbine operating mechanism and the DBSCAN algorithm to obtain a standard wind turbine feature dataset; and a blade angle anomaly threshold calculation module. The first unit uses power as the operating condition division, and uses box plot method to calculate the normal range of blade angles in the standard wind turbine characteristic dataset to obtain multi-condition blade angle anomaly thresholds. Based on the multi-condition blade angle anomaly thresholds, it identifies the unit's high-temperature power limitation and obtains the first unit's high-temperature power limitation identification result. The second unit's high-temperature power limitation identification result is determined based on the first unit's high-temperature power limitation identification result and the first unit's high-temperature power limitation identification result. The third unit uses power and ambient temperature as the operating condition division, and uses statistical analysis method to calculate the normal distribution of gearbox oil temperature and generator temperature in the standard wind turbine characteristic dataset to obtain multi-condition temperature anomaly distribution thresholds. Based on the first unit's high-temperature power limitation identification result and the multi-condition temperature anomaly distribution thresholds, it identifies the unit's high-temperature power limitation and determines the second unit's high-temperature power limitation identification result.

[0015] One or more technical solutions provided in this application have at least the following technical effects or advantages: By obtaining relevant characteristic variables of wind turbines from the Kairosdb time-series database of the centralized SCADA system, and aggregating these variables, a usable wind turbine characteristic dataset is obtained. Based on the wind turbine operating mechanism and the DBSCAN algorithm, this dataset is preprocessed to obtain a standard wind turbine characteristic dataset. Using power as the operating condition interval, the standard wind turbine characteristic dataset is analyzed using a box plot method to calculate the normal range of blade angles, resulting in multi-condition blade angle anomaly thresholds. Based on these thresholds, high-temperature power limitation identification is performed to obtain the first unit's high-temperature power limitation identification result. Using power and ambient temperature as operating condition intervals, statistical analysis is used to calculate the normal distribution of gearbox oil temperature and generator temperature in the standard wind turbine characteristic dataset, resulting in multi-condition temperature anomaly distribution thresholds. Based on the first unit's high-temperature power limitation identification result and the multi-condition temperature anomaly distribution thresholds, high-temperature power limitation identification is performed to determine the second unit's high-temperature power limitation identification result. In other words, by extracting wind turbine-related feature variables from the Kairosdb time-series database of the centralized SCADA system, and combining the wind turbine's operating mechanism and the DBSCAN clustering algorithm, the aggregated feature dataset is preprocessed. Using power as the basis for dividing operating conditions, box plots are used to calculate the normal range of blade angles in the standardized wind turbine feature dataset. By comparing the blade angle distribution under different power ranges, the abnormal blade angle thresholds under multiple operating conditions are determined. Using power and ambient temperature as the basis for dividing operating conditions, statistical analysis methods are used to calculate the normal distribution of gearbox oil temperature and generator temperature in the standardized wind turbine feature dataset. Combined with the high-temperature power-limiting identification results of the first unit, the abnormal temperature distribution thresholds under multiple operating conditions are determined. This effectively identifies power-limiting situations caused by high temperatures in wind turbine oil components, thereby improving the operating efficiency of the wind turbine.

[0016] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the data mining and identification method for high-temperature power limitation of wind turbines proposed in this application.

[0019] Figure 2 The power curves before and after data cleaning are shown in this application.

[0020] Figure 3 This is a box plot showing the power versus blade angle of the unit under normal operating conditions.

[0021] Figure 4 This is a diagram showing the diagnostic results of the unit's power outage caused by high gearbox oil temperature in this application.

[0022] Figure 5 This is a diagram showing the gearbox oil temperature distribution in this application.

[0023] Figure 6 This diagram shows the diagnostic results of the generator unit's power outage caused by high temperature of the generator drive end bearing in this application.

[0024] Figure 7 This is a temperature distribution diagram of the generator drive end bearing in this application.

[0025] Figure 8 This is a schematic diagram of the data mining and identification system for high-temperature power limitation of wind turbine units, as described in this application.

[0026] Figure labeling: Aggregation processing module 11, data preprocessing module 12, blade angle anomaly threshold calculation module 13, temperature anomaly distribution threshold calculation module 14. Detailed Implementation

[0027] This application provides a data mining and identification method and system for high-temperature power curtailment in wind turbines, solving the technical problem in existing technologies where high false alarm rates due to fixed thresholds for component status judgment and lack of operational condition correlation analysis lead to a further impact on wind turbine operating efficiency. By extracting wind turbine-related feature variables from the Kairosdb time-series database of the centralized SCADA system, and combining the wind turbine operating mechanism and DBSCAN clustering algorithm, the aggregated feature dataset is preprocessed. Using power as the basis for dividing operating condition intervals, box plots are used to calculate the normal range of blade angles in the standardized wind turbine feature dataset. By comparing the blade angle distribution under different power intervals, abnormal blade angle thresholds under multiple operating conditions are determined. Using power and ambient temperature as the basis for dividing operating conditions, statistical analysis methods are used to calculate the normal distribution of gearbox oil temperature and generator temperature in the standardized wind turbine feature dataset. Combined with the high-temperature power curtailment identification results of the first unit, abnormal temperature distribution thresholds under multiple operating conditions are determined. This effectively identifies power curtailment caused by high temperatures in wind turbine oil components, thereby improving the operating efficiency of wind turbines.

[0028] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0029] Example 1, please refer to the appendix. Figure 1 This application provides a data mining and identification method for wind turbine power curtailment, wherein the data mining and identification method for wind turbine power curtailment is applied to a data mining and identification system for wind turbine power curtailment, and the data mining and identification method for wind turbine power curtailment specifically includes the following steps: S100: Obtain wind turbine-related feature variables from the Kairosdb time-series database of the centralized SCADA system, aggregate the wind turbine-related feature variables, and obtain an available wind turbine feature dataset.

[0030] Furthermore, S100 of this application includes: the relevant characteristic variables of the wind turbine specifically include: turbine load, ambient temperature, wind speed, first blade angle, second blade angle, third blade angle, generator bearing temperature, and gearbox oil temperature.

[0031] Specifically, the system retrieves relevant characteristic variables of the wind turbines from the Kairosdb time-series database of the centralized SCADA system. These are key operating parameters related to identifying high-temperature power curtailment, including turbine load, ambient temperature, wind speed, first blade angle, second blade angle, third blade angle, generator bearing temperature, and gearbox oil temperature. The centralized SCADA system is the data acquisition and monitoring system of the wind farm, acting as its "nerve center." It is responsible for real-time acquisition of data from thousands of sensors on every wind turbine throughout the farm, such as temperature, speed, and angle sensors, and for centralized monitoring and control. The Kairosdb time-series database is specifically designed for storing and querying time-series data, recording data points in chronological order, such as temperature and power per minute, thus storing the massive amounts of raw data collected by the SCADA system in an orderly manner.

[0032] Each variable acquired from the centralized Kairosdb time-series database has a different sampling frequency, and the timestamps may not be completely synchronized. Therefore, data alignment and aggregation are required. After aggregation and time alignment, the raw data is converted into standard structured data with 5-minute time intervals. The median of the three variables—first blade angle, second blade angle, and third blade angle—is taken as the blade angle. In other words, with a fixed time window of 5 minutes, calculations are performed on the raw data within each time window. For example, the average of continuous temperature readings is taken as the representative value for that 5-minute time point, and the median of the three blade angle readings is taken as a comprehensive blade angle value. This integrates all variables into a usable wind turbine characteristic dataset with strictly aligned timestamps and uniformly spaced data points.

[0033] Furthermore, this application also includes the following steps: aggregating and aligning the relevant feature variables of the wind turbine generator to obtain standard wind turbine generator feature variable data; and performing structured transformation on the standard wind turbine generator feature variable data at preset time intervals to obtain a usable wind turbine generator feature dataset.

[0034] Specifically, the relevant characteristic variables of wind turbine units are aggregated. Aggregation rules are set for each variable according to its type and physical meaning. Using a reference clock, all data points of all variables are aligned and merged to the nearest standard time frame, generating a set of standardized data with completely synchronized timestamps for all variables. Aggregation refers to the process of merging multiple data points that are close in time into a more representative data point, including taking the average or median within a specified time window. Time alignment refers to the process of unifying variable data from different sources and at different collection times onto the same set of standard timestamps. The standardized wind turbine characteristic variable data is a set of data that has undergone preliminary cleaning and has been aggregated and aligned to ensure a one-to-one correspondence between all variable data in the time series, but its timestamps may still be unevenly spaced or highly frequent.

[0035] Using a preset time interval as a fixed sampling interval, data points are systematically extracted from the aligned data stream, such as one point every 5 minutes, or the standardized data within a 5-minute window are aggregated again. These data points are then arranged in chronological order and filled into a two-dimensional data table. The preset time interval is a fixed analysis time unit, such as 5 minutes. Structured transformation refers to organizing the data into a regular tabular form. Typically, each row represents a unique time point, each column represents a feature variable, and each cell represents the value of that variable at that time point.

[0036] By aggregating and aligning, the temporal disorder and frequency differences of multi-source sensor data are eliminated, ensuring the consistency of data in the time dimension; then, through structured transformation at fixed intervals, a static dataset with uniform time scale and standardized format is generated.

[0037] S200: Combining the wind turbine operating mechanism and the DBSCAN algorithm, the available wind turbine feature dataset is preprocessed to obtain a standard wind turbine feature dataset.

[0038] Furthermore, S200 of this application includes: deleting abnormal data points in the available wind turbine characteristic dataset according to the standard power curve of the turbine unit. , , ,in, For wind speed, Unit cut-in wind speed, Cut off the wind speed for the unit. For the unit load; based on the variation law of the unit blade angle, retain the centralized and standardized data points of the available wind turbine characteristic dataset: and , and , and , ,in, The rated power of the unit, The blade angle of the wind turbine unit; the available wind turbine feature dataset is binned, with each wind speed range defined as follows: Using intervals, we obtain the dataset within each small interval. for: Then, the DBSCAN algorithm is used for cleaning to obtain a standard wind turbine feature dataset; among which, Represents the entire dataset. Indicates the first A dataset of intervals, For the first The elements of interval n, where n is the dataset. Medium sample size.

[0039] Specifically, based on the unit's standard power curve, data points that clearly violate physical laws are removed, such as the following data points: , , .in, Wind speed, in m / s. The unit cut-in wind speed is expressed in m / s. The cutoff wind speed for the unit, in m / s. This represents the unit load, expressed in kW.

[0040] Based on the variation pattern of the turbine blade angle, the following data points are retained, including: and , and , and , ,in, The rated power of the unit, This refers to the turbine blade angle. In other words, based on the coupling relationship between blade angle and power, data combinations that conform to the control logic are retained. For example, the blade angle should be small at low power, while large angles are allowed at high power or during power outages. At the same time, data points where the power and blade angle combinations are clearly contradictory are removed, such as high power combined with a very large blade pitch angle, which itself may indicate a power outage and is not part of the normal baseline. After data removal and retention, most invalid, shutdown, and power outage data have been eliminated.

[0041] The available wind turbine feature dataset is binned; that is, the remaining available wind turbine feature dataset after mechanism screening is further refined into bins based on wind speed. Each wind speed range is defined as follows: Using intervals, we obtain the dataset within each small interval. for: Then, the DBSCAN algorithm is used for cleaning to obtain a standard wind turbine feature dataset; among which, Represents the entire dataset. Indicates the first A dataset of intervals, For the first The elements of interval n, where n is the dataset. Medium sample size. Binning, also known as binning, is a data discretization technique. It divides the range of a continuous variable into a series of continuous intervals, such as intervals of 0.5 m / s, and assigns data points to the corresponding intervals based on their values ​​for analysis.

[0042] Within each bucket, the DBSCAN clustering algorithm is applied to identify data points clustered in the core density region as normal clusters, while discretely distributed data points far from these clusters are identified as noise and discarded. These noise points may be caused by transient turbulence, transient sensor interference, or subtle anomalies not captured by the first-stage rules.

[0043] After preprocessing based on the wind turbine operating mechanism and the DBSCAN algorithm, a standard wind turbine feature dataset was obtained. Mechanism-based hard rules efficiently removed macroscopically abnormal state data, ensuring the basic physical authenticity of the baseline data. DBSCAN-based soft clustering meticulously eliminated microscopically random noise and statistical anomalies, ensuring the statistical purity of the baseline data.

[0044] S300: Using power as the operating condition range, the standard wind turbine characteristic dataset is used to calculate the normal range of blade angles, and the multi-condition blade angle abnormality threshold is obtained. Based on the multi-condition blade angle abnormality threshold, the unit high-temperature power limitation identification is performed to obtain the first unit high-temperature power limitation identification result.

[0045] Furthermore, S300 of this application includes: dividing the standard wind turbine feature dataset into buckets based on power as the operating condition interval, with each power interval being... Using intervals, we obtain the dataset within each small interval. for: , For each power range dataset, the box plot method is used to calculate the quartiles of the blade angle data within each range. According to the box plot anomaly detection rules, outliers in the blade angle data within each range are those greater than the 3 / 4 quartile plus 1.5 times the interquartile range. The multi-condition blade angle anomaly threshold is thus obtained, defined as follows: ,in, For the first Blade angle thresholds within a power range , For the first The 1 / 4 and 3 / 4 quantiles of the blade angle data for each power range are used to store the power range data and the multi-condition blade angle anomaly thresholds in the corresponding database.

[0046] Specifically, the standard wind turbine feature dataset is binned by power, with each power range defined as follows: Using intervals, we obtain the dataset within each small interval. for: , Where p is the unit load and a is the unit blade angle. For example, the power is divided into multiple continuous intervals from 0 to rated power, with 100 kilowatt intervals.

[0047] For each power range dataset, box plots are used to statistically analyze the quartiles of the blade angle data within each range. The blade angle values ​​of all data points are extracted to form a set of angle values. The 1 / 4 quartile and 3 / 4 quartile of this set are calculated, thus obtaining the interquartile range. Based on the outlier criteria of box plots, a commonly used rule is that a value greater than the 3 / 4 quartile plus 1.5 times the interquartile range is considered a potential outlier. Therefore, the result of the calculation (3 / 4 quartile + 1.5 × interquartile range) is defined as the blade angle threshold for the current power range. This means that under historical normal operating conditions, there is a 99.3% probability that the blade angle corresponding to this power level is lower than this value. The multi-condition blade angle outlier threshold is defined as follows: ,in, For the first Blade angle thresholds within a power range , For the first The 1 / 4 and 3 / 4 quantiles of the blade angle data for each power range.

[0048] After traversing all power ranges, a mapping table between power and blade angle thresholds was established and stored in the database. This threshold statistically represents the reasonable upper limit of blade angle values ​​in historical normal operation data at that power level. Exceeding this limit strongly suggests that the operating state has deviated from the normal power generation mode.

[0049] Each power range is paired with its calculated unique blade angle threshold to form operating condition-threshold rules. The power range data for each range and the multi-operating condition blade angle anomaly thresholds are stored in the corresponding database. The box plot method judges based on the data's own distribution and is insensitive to extreme values ​​in the data; therefore, the constructed threshold model has strong anti-interference capabilities and high stability. Power curtailment is only determined when the real-time blade angle exceeds the multi-operating condition blade angle anomaly threshold, optimizing the allocation of computing resources, avoiding unnecessary continuous analysis, and improving the overall system efficiency.

[0050] Furthermore, this application also includes the following steps: collecting real-time data of the unit, searching the target power range corresponding to the unit power in the database based on the unit power of the real-time data of the unit; determining whether the current blade angle is greater than the multi-condition blade angle abnormal threshold corresponding to the target power range, and obtaining the first unit high-temperature power limitation identification result.

[0051] Specifically, this involves collecting real-time data from the wind turbine, which includes the operating parameters at the current moment collected and uploaded to the centralized control system through the wind turbine's field controller and sensors. These parameters typically include power, wind speed, blade angle, and temperature of various components. The system then captures the turbine's current power and blade angle from this real-time data.

[0052] The system retrieves the target power range corresponding to the unit's power value from the database based on the unit's real-time power data. In other words, it uses the unit's power value from the real-time data as the query key to quickly search the database, locate the unique target power range containing that power value, and read the blade angle anomaly threshold bound to that range. This threshold represents the maximum blade angle allowed during normal operation at that power level in history.

[0053] The real-time blade angle is compared with the multi-condition blade angle anomaly threshold corresponding to the target power range. If the real-time blade angle is greater than the multi-condition blade angle anomaly threshold corresponding to the target power range, the unit is in a power curtailment state; conversely, if the real-time blade angle is less than or equal to the multi-condition blade angle anomaly threshold corresponding to the target power range, it indicates that the unit is operating in normal power generation tracking mode, i.e., the unit is in normal power generation state. The high-temperature power curtailment identification result of the first unit is a binary judgment; it does not directly determine whether the power curtailment is due to high temperature, but first determines whether the unit is in a power curtailment operating state. The result is yes or no.

[0054] The complex identification of power rationing conditions is transformed into rapid database queries and numerical comparisons, resulting in extremely low computational overhead and meeting the millisecond-level response requirements of SCADA systems for real-time monitoring. It completely eliminates the subjective approach of relying on operators' observation of curves or experience-based judgments, using data-driven thresholds for automatic identification, ensuring consistent and reliable results. Only when a power rationing is determined in this step will the subsequent high-temperature root cause analysis module, which consumes more computational resources, be activated. This avoids unnecessary continuous in-depth analysis during normal unit operation, optimizes overall computational resource allocation, and improves processing efficiency.

[0055] S400: Using power and ambient temperature as operating conditions, the normal distribution of gearbox oil temperature and generator temperature is calculated using statistical analysis methods on the characteristic dataset of the standard wind turbine, and the abnormal temperature distribution threshold of multiple operating conditions is obtained. Based on the high temperature power limitation identification result of the first unit and the abnormal temperature distribution threshold of multiple operating conditions, the high temperature power limitation identification result of the second unit is determined.

[0056] Furthermore, S400 of this application includes: dividing the standard wind turbine characteristic dataset into buckets based on power and ambient temperature as operating conditions, and dividing the ambient temperature data within each power range into buckets. Using intervals, the dataset within each grid interval is obtained. for: , , ,in, For ambient temperature, This refers to the gearbox oil temperature or the generator bearing temperature. The number of divisions for the power range. This indicates the number of temperature intervals within each power range; calculate the mean and variance of the gearbox oil temperature or generator bearing temperature within each ambient temperature interval under each power range of the target unit, based on... In principle, the normal range of each temperature distribution should be within three standard deviations above and below its mean. The threshold values ​​for the first gearbox oil temperature or generator bearing temperature distribution are defined as follows: ,in, Indicates the first Each power range Indicates the first The power range under the first The first gearbox oil temperature or generator bearing temperature distribution threshold is determined by performing quartile statistics on the first gearbox oil temperature or generator bearing temperature distribution threshold. Following the box plot anomaly detection method, outliers in the temperature data within the grid interval are located in areas greater than the 3 / 4 quartile plus 1.5 times the interquartile range. This yields the second gearbox oil temperature or generator bearing temperature distribution threshold, defined as follows: ,in, For the first The power range under the first Gearbox oil temperature or generator bearing temperature within a specific ambient temperature range. , For the first The power range under the first The 1 / 4 and 3 / 4 quantiles of gearbox oil temperature or generator bearing temperature within a given ambient temperature range; based on the first gearbox oil temperature or generator bearing temperature distribution threshold and the second gearbox oil temperature or generator bearing temperature distribution threshold, determine the multi-condition temperature anomaly distribution threshold: The corresponding power range, ambient temperature range, and the multi-condition temperature anomaly distribution threshold are used to determine the appropriate power range, ambient temperature range, and the threshold values. Store it in the corresponding database.

[0057] Specifically, the standard wind turbine characteristic dataset is divided into buckets based on power and ambient temperature as operating conditions. That is, based on the already divided power ranges, the ambient temperature data within each power range is further divided, which is similar to the process of dividing the operating conditions based on power.

[0058] Ambient temperature data within each power range Using intervals, the dataset within each grid interval is obtained. for: , , ,in, For ambient temperature, This refers to the gearbox oil temperature or the generator bearing temperature. The number of divisions for the power range. This indicates the number of temperature ranges within each power range. The entire operating state space is divided into small grids, and the data within each grid exhibits highly consistent operating conditions.

[0059] Calculate the mean and variance of the gearbox oil temperature or generator bearing temperature within each ambient temperature range for each power range of the target unit, based on... In principle, the normal range of each temperature distribution should be within three standard deviations above and below its mean. The threshold values ​​for the first gearbox oil temperature or generator bearing temperature distribution are defined as follows: ,in, Indicates the first Each power range Indicates the first The power range under the first An ambient temperature range is defined. The historical normal dataset is traversed, and data points are assigned to corresponding grids based on their power and ambient temperature values. For each grid, the component temperature of all data points is extracted, and the mean and standard deviation of this temperature set are calculated. Subsequently, the parameterized threshold for that grid is calculated based on the 3σ rule of normal distribution. The 3σ rule is an empirical rule in statistics. For data that follows or approximately follows a normal distribution, about 99.7% of the data points will fall within the interval [μ-3σ, μ+3σ] of the mean μ plus or minus 3 times the standard deviation σ. Based on this, μ+3σ is often used as a threshold to judge whether the data is abnormally high.

[0060] The first gearbox oil temperature or generator bearing temperature distribution threshold is statistically analyzed using quartiles. Following the box plot anomaly detection method, outliers in the temperature data within the grid interval are located in areas greater than the 3 / 4 quartile plus 1.5 times the interquartile range. This yields the second gearbox oil temperature or generator bearing temperature distribution threshold, defined as follows: ,in, For the first The power range under the first Gearbox oil temperature or generator bearing temperature within a specific ambient temperature range. , For the first The power range under the first The 1 / 4 and 3 / 4 quantiles of gearbox oil temperature or generator bearing temperature within a given ambient temperature range. In other words, for the same set of component temperature data within the same grid, the 1 / 4 quantile, 3 / 4 quantile, and interquartile range are calculated. Based on the outlier detection rules of the box plot, a nonparametric threshold for that grid is calculated. This threshold does not rely on any distribution assumptions but is directly determined by the data's sorting position, making it more robust for non-normally distributed datasets or datasets with potential outliers.

[0061] Based on the first gearbox oil temperature or generator bearing temperature distribution threshold and the second gearbox oil temperature or generator bearing temperature distribution threshold, determine the multi-condition temperature anomaly distribution threshold: ; The corresponding power range, ambient temperature range, and multi-condition temperature anomaly distribution thresholds are set. Store the data in the corresponding database. The 3σ principle is highly efficient when the data is well-normalized; the box plot method is more robust in dealing with non-normal data and resisting the influence of extreme values.

[0062] By employing both parametric 3σ and nonparametric box plot methods in parallel, the problem that actual industrial data may not strictly follow a normal distribution is overcome. Regardless of the distribution pattern of temperature data, at least one method can provide reasonable anomaly boundaries.

[0063] Furthermore, this application also includes the following steps: if the first unit's high-temperature power rationing identification result is that the unit is in a power rationing state, based on the unit load and ambient temperature, the target grid interval where the corresponding unit load and ambient temperature are located is searched from the database; it is determined whether the current gearbox oil temperature or generator bearing temperature is greater than the multi-condition temperature anomaly distribution threshold corresponding to the target grid interval, and the second unit's high-temperature power rationing identification result is determined.

[0064] Specifically, if the high-temperature power rationing identification result of the first unit indicates that the unit is in a power rationing state, the target grid interval corresponding to the unit load and ambient temperature is searched from the database based on the unit load and ambient temperature. Using the unit load and ambient temperature as a composite query key, a fast search is performed in the database to accurately locate the unique target grid interval to which these two values ​​belong.

[0065] The real-time monitored component temperature, i.e., the current gearbox oil temperature or generator bearing temperature, is compared with the multi-condition temperature anomaly distribution threshold corresponding to the target grid interval. If the current gearbox oil temperature or generator bearing temperature is greater than the multi-condition temperature anomaly distribution threshold corresponding to the target grid interval, it indicates that under the current specific power and ambient temperature conditions, the component temperature has significantly exceeded the upper limit of historical normal fluctuations, and the unit power curtailment is caused by high gearbox oil temperature or generator bearing temperature. Conversely, if the current gearbox oil temperature or generator bearing temperature is less than or equal to the multi-condition temperature anomaly distribution threshold corresponding to the target grid interval, it indicates that although the unit is experiencing power curtailment, the temperature of the critical component is still within the normal range allowed under the current operating conditions, thus ruling out high temperature as the direct cause, and the curtailment is due to other reasons.

[0066] In a specific example, the data sampling period was from November 2023 to December 2024, a total of 12 months. After downsampling, the data was converted into data with a time interval of 5 minutes. During the period from November to December 2024, the unit experienced power outages due to high gearbox oil temperature and high generator temperature. Modeling was performed using the period from November 2023 to October 2024, and the data cleaning method described in this invention was used to remove power outage, noise, and abnormal data. The cleaning results are as follows. Figure 2 As shown. Then, using power as the dividing point for operating conditions, the relationship between power and blade angle under each operating condition was calculated, and the results are shown in the figure. Figure 3 As shown in the figure. Secondly, based on the power-based operating conditions, the ambient temperature within each power range is further divided into ranges. The threshold temperatures for the gearbox oil temperature and generator temperature under this operating condition are calculated for each grid range. The results for some grid ranges are shown in Table 1. Table 1. Upper Threshold Threshold Threshold of Gearbox Oil Temperature under Some Operating Conditions (Load, ambient temperature) Gearbox oil temperature upper limit threshold (1500.0,-15.0) 59.38 (1500.0,-10.0) 59.85 (1500.0,-5.0) 59.86 (1500.0,0.0) 59.94 (1500.0,5.0) 59.12 (1500.0,10.0) 61.71 (1500.0,15.0) 60.85 (1500.0,20.0) 61.36 (1600.0,-15.0) 59.25 (1600.0,-10.0) 61.15 (1600.0,-5.0) 58.87 (1600.0,-0.0) 58.6 (1600.0,5.0) 60.3 (1600.0,10.0) 61.53 (1600.0,15.0) 61.18 (1600.0,20.0) 67.07 The method described in this patent was verified during the period from November to December 2024, and the verification results are as follows: Figure 4 , Figure 5 , Figure 6 , Figure 7 As shown, power curtailment was detected between 12:30 and 13:00 on November 11, 2024, and between 14:00 and 14:40 on November 11, 2024. Based on the current power output and corresponding operating conditions, the current blade angle exceeded the upper limit threshold for that operating condition. Then, by matching the current power output and ambient temperature to the corresponding operating conditions, it was found that during these two time periods, the gearbox oil temperature and generator drive end temperature were higher than the upper limit of the normal temperature distribution under those operating conditions. Therefore, it can be determined that power curtailment due to high gearbox oil temperature and high generator drive end bearing temperature occurred during these two time periods, verifying the effectiveness of the method.

[0067] Without human intervention, it accurately correlates power curtailment events with high temperatures in specific components, directly outputting the conclusions urgently needed for operation and maintenance, significantly improving fault location efficiency. Through blade angle anomaly threshold calculation and temperature anomaly distribution threshold calculation, it comprehensively identifies blade angle anomalies under different operating conditions to determine whether wind turbines are subject to limitations, comprehensively identifies anomalies in wind turbines under high-temperature conditions, effectively identifies high-temperature power curtailment situations of wind turbines, and improves the operational efficiency and reliability of wind farms.

[0068] In summary, the data mining and identification method for high-temperature power curtailment of wind turbines provided in this application has the following technical effects: By obtaining relevant characteristic variables of wind turbines from the Kairosdb time-series database of the centralized SCADA system, and aggregating these variables, a usable wind turbine characteristic dataset is obtained. Based on the wind turbine operating mechanism and the DBSCAN algorithm, this dataset is preprocessed to obtain a standard wind turbine characteristic dataset. Using power as the operating condition interval, the standard wind turbine characteristic dataset is analyzed using a box plot method to calculate the normal range of blade angles, resulting in multi-condition blade angle anomaly thresholds. Based on these thresholds, high-temperature power limitation identification is performed to obtain the first unit's high-temperature power limitation identification result. Using power and ambient temperature as operating condition intervals, statistical analysis is used to calculate the normal distribution of gearbox oil temperature and generator temperature in the standard wind turbine characteristic dataset, resulting in multi-condition temperature anomaly distribution thresholds. Based on the first unit's high-temperature power limitation identification result and the multi-condition temperature anomaly distribution thresholds, high-temperature power limitation identification is performed to determine the second unit's high-temperature power limitation identification result. In other words, by extracting wind turbine-related feature variables from the Kairosdb time-series database of the centralized SCADA system, and combining the wind turbine's operating mechanism and the DBSCAN clustering algorithm, the aggregated feature dataset is preprocessed. Using power as the basis for dividing operating conditions, box plots are used to calculate the normal range of blade angles in the standardized wind turbine feature dataset. By comparing the blade angle distribution under different power ranges, the abnormal blade angle thresholds under multiple operating conditions are determined. Using power and ambient temperature as the basis for dividing operating conditions, statistical analysis methods are used to calculate the normal distribution of gearbox oil temperature and generator temperature in the standardized wind turbine feature dataset. Combined with the high-temperature power-limiting identification results of the first unit, the abnormal temperature distribution thresholds under multiple operating conditions are determined. This effectively identifies power-limiting situations caused by high temperatures in wind turbine oil components, thereby improving the operating efficiency of the wind turbine.

[0069] Example 2: Based on the same inventive concept as the data mining and identification method for high-temperature power curtailment of wind turbines in Example 1, this application also provides a data mining and identification system for high-temperature power curtailment of wind turbines. Please refer to the appendix. Figure 8 The data mining and identification system for high-temperature power limitation of wind turbine units includes: Aggregation processing module 11 is used to obtain wind turbine-related feature variables from the Kairosdb time-series database of the centralized control SCADA system, and aggregate the wind turbine-related feature variables to obtain a usable wind turbine feature dataset; data preprocessing module 12 is used to preprocess the usable wind turbine feature dataset based on the wind turbine operating mechanism and the DBSCAN algorithm to obtain a standard wind turbine feature dataset; blade angle anomaly threshold calculation module 13 is used to divide the operating condition range by power and use the box plot method to calculate the blade angle normality threshold of the standard wind turbine feature dataset. Interval calculations are performed to obtain multi-condition blade angle anomaly thresholds, and based on these thresholds, high-temperature power rationing identification of the unit is performed to obtain the first unit's high-temperature power rationing identification result. The temperature anomaly distribution threshold calculation module 14 is used to divide the operating conditions by power and ambient temperature, and to calculate the normal distribution of gearbox oil temperature and generator temperature on the standard wind turbine characteristic dataset using statistical analysis methods to obtain multi-condition temperature anomaly distribution thresholds. Based on the first unit's high-temperature power rationing identification result and the multi-condition temperature anomaly distribution thresholds, high-temperature power rationing identification of the unit is performed to determine the second unit's high-temperature power rationing identification result.

[0070] Furthermore, the aggregation processing module 11 in the data mining and identification system for high-temperature power limitation of wind turbine units is also used for: the relevant characteristic variables of the wind turbine unit specifically include: unit load, ambient temperature, wind speed, first blade angle, second blade angle, third blade angle, generator bearing temperature, and gearbox oil temperature.

[0071] Furthermore, the aggregation processing module 11 in the data mining and identification system for high-temperature power limitation of wind turbine units is also used to: perform aggregation processing and time alignment on the relevant feature variables of the wind turbine units to obtain standardized wind turbine unit feature variable data; and perform structured transformation on the standardized wind turbine unit feature variable data according to a preset time interval to obtain a usable wind turbine unit feature dataset.

[0072] Furthermore, the data preprocessing module 12 in the data mining and identification system for high-temperature power curtailment of wind turbine units is also used to: delete abnormal data points in the available wind turbine unit feature data according to the unit's standard power curve. , , ,in, For wind speed, Unit cut-in wind speed, Cut off the wind speed for the unit. For the unit load; based on the variation law of the unit blade angle, retain the centralized and standardized data points of the available wind turbine characteristic dataset: and , and , and , ,in, The rated power of the unit, The blade angle of the wind turbine unit; the available wind turbine feature dataset is binned, with each wind speed range defined as follows: Using intervals, we obtain the dataset within each small interval. for: Then, the DBSCAN algorithm is used for cleaning to obtain a standard wind turbine feature dataset; among which, Represents the entire dataset. Indicates the first A dataset of intervals, For the first The elements of interval n, where n is the dataset. Medium sample size.

[0073] Furthermore, the blade angle anomaly threshold calculation module 13 in the data mining and identification system for high-temperature power curtailment of wind turbines is also used to: divide the standard wind turbine feature dataset into buckets based on power as the operating condition interval, with each power interval being... Using intervals, we obtain the dataset within each small interval. for: , For each power range dataset, the box plot method is used to calculate the quartiles of the blade angle data within each range. According to the box plot anomaly detection rules, outliers in the blade angle data within each range are those greater than the 3 / 4 quartile plus 1.5 times the interquartile range. The multi-condition blade angle anomaly threshold is thus obtained, defined as follows: ,in, For the first Blade angle thresholds within a power range , For the first The 1 / 4 and 3 / 4 quantiles of the blade angle data for each power range are used to store the power range data and the multi-condition blade angle anomaly thresholds in the corresponding database.

[0074] Furthermore, the blade angle anomaly threshold calculation module 13 in the data mining and identification system for high-temperature power limitation of wind turbine units is also used for: collecting real-time data of the unit, searching for the target power range corresponding to the power in the database based on the unit power of the real-time data of the unit; determining whether the current blade angle is greater than the multi-condition blade angle anomaly threshold corresponding to the target power range, and obtaining the first unit high-temperature power limitation identification result.

[0075] Furthermore, the temperature anomaly distribution threshold calculation module 14 in the data mining and identification system for high-temperature power curtailment of wind turbines is also used to: divide the standard wind turbine feature dataset into buckets based on power and ambient temperature as operating conditions, and calculate the ambient temperature data within each power range using... Using intervals, the dataset within each grid interval is obtained. for: , , ,in, For ambient temperature, This refers to the gearbox oil temperature or the generator bearing temperature. The number of divisions for the power range. This indicates the number of temperature intervals within each power range; calculate the mean and variance of the gearbox oil temperature or generator bearing temperature within each ambient temperature interval under each power range of the target unit, based on... In principle, the normal range of each temperature distribution should be within three standard deviations above and below its mean. The threshold values ​​for the first gearbox oil temperature or generator bearing temperature distribution are defined as follows: ,in, Indicates the first Each power range Indicates the first The power range under the first The first gearbox oil temperature or generator bearing temperature distribution threshold is determined by performing quartile statistics on the first gearbox oil temperature or generator bearing temperature distribution threshold. Following the box plot anomaly detection method, outliers in the temperature data within the grid interval are located in areas greater than the 3 / 4 quartile plus 1.5 times the interquartile range. This yields the second gearbox oil temperature or generator bearing temperature distribution threshold, defined as follows: ,in, For the first The power range under the first Gearbox oil temperature or generator bearing temperature within a specific ambient temperature range. , For the first The power range under the first The 1 / 4 and 3 / 4 quantiles of gearbox oil temperature or generator bearing temperature within a given ambient temperature range; based on the first gearbox oil temperature or generator bearing temperature distribution threshold and the second gearbox oil temperature or generator bearing temperature distribution threshold, determine the multi-condition temperature anomaly distribution threshold: The corresponding power range, ambient temperature range, and the multi-condition temperature anomaly distribution threshold are used to determine the appropriate power range, ambient temperature range, and the threshold values. Store it in the corresponding database.

[0076] Furthermore, the temperature anomaly distribution threshold calculation module 14 in the data mining and identification system for high-temperature power curtailment of wind turbine units is also used to: if the high-temperature power curtailment identification result of the first unit is that the unit is in a power curtailment state, search for the target grid interval where the corresponding unit load and ambient temperature are located in the database according to the unit load and ambient temperature; determine whether the current gearbox oil temperature or generator bearing temperature is greater than the multi-condition temperature anomaly distribution threshold corresponding to the target grid interval, and determine the high-temperature power curtailment identification result of the second unit.

[0077] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The data mining and identification method and specific examples for high-temperature power curtailment of wind turbines in the aforementioned embodiment 1 are also applicable to the data mining and identification system for high-temperature power curtailment of wind turbines in this embodiment. Through the foregoing detailed description of the data mining and identification method for high-temperature power curtailment of wind turbines, those skilled in the art can clearly understand the data mining and identification system for high-temperature power curtailment of wind turbines in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0078] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0079] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A data mining and identification method for high-temperature power curtailment in wind turbine units, characterized in that, include: Relevant feature variables of wind turbines are obtained from the Kairosdb time-series database of the centralized SCADA system. The relevant feature variables of wind turbines are aggregated to obtain a usable wind turbine feature dataset. By combining the wind turbine operating mechanism and the DBSCAN algorithm, the available wind turbine feature dataset is preprocessed to obtain a standard wind turbine feature dataset. Using power as the operating condition range, the standard wind turbine characteristic dataset is used to calculate the normal range of blade angles, and multi-condition blade angle abnormality thresholds are obtained. Based on the multi-condition blade angle abnormality thresholds, the unit's high-temperature power limitation is identified, and the first unit's high-temperature power limitation identification result is obtained. Using power and ambient temperature as operating conditions, statistical analysis methods are used to calculate the normal distribution of gearbox oil temperature and generator temperature in the standard wind turbine characteristic dataset, obtaining the multi-condition temperature abnormal distribution threshold. Based on the first unit's high-temperature power limitation identification result and the multi-condition temperature abnormal distribution threshold, the unit's high-temperature power limitation identification result is determined, and the second unit's high-temperature power limitation identification result is determined.

2. The data mining and identification method for high-temperature power curtailment of wind turbines as described in claim 1, characterized in that, The specific characteristic variables of the wind turbine include: turbine load, ambient temperature, wind speed, first blade angle, second blade angle, third blade angle, generator bearing temperature, and gearbox oil temperature.

3. The data mining and identification method for high-temperature power curtailment of wind turbines as described in claim 1, characterized in that, The available wind turbine feature dataset is obtained, including: The relevant feature variables of the wind turbine are aggregated and time-aligned to obtain standardized wind turbine feature variable data; The characteristic variable data of the standard wind turbine are structurally transformed according to a preset time interval to obtain a usable wind turbine characteristic dataset.

4. The data mining and identification method for high-temperature power curtailment of wind turbines as described in claim 1, characterized in that, Obtain a standard wind turbine feature dataset, including: Based on the standard power curve of the wind turbine unit, delete the abnormal data points in the available wind turbine unit feature data set: , , ,in, For wind speed, Unit cut-in wind speed, Cut off the wind speed for the unit. For unit load; Based on the variation pattern of the turbine blade angle, retain the centralized and standardized data points of the available wind turbine characteristic dataset: and , and , and , ,in, The rated power of the unit, The blade angle of the generator set; The available wind turbine feature dataset is binned, with each wind speed range defined as follows: Using intervals, we obtain the dataset within each small interval. for: Then, the DBSCAN algorithm is used to clean the data to obtain a standard wind turbine feature dataset. in, Represents the entire dataset. Indicates the first A dataset of intervals, For the first The elements of interval n, where n is the dataset. Medium sample size.

5. The data mining and identification method for high-temperature power curtailment of wind turbine units as described in claim 4, characterized in that, The threshold values ​​for abnormal blade angles under multiple operating conditions were obtained, including: The standard wind turbine feature dataset is divided into buckets based on power as the operating condition range. Using intervals, we obtain the dataset within each small interval. for: , ; For each power range dataset, the box plot method is used to calculate the quartiles of the blade angle data within each range. According to the box plot anomaly detection rules, outliers in the blade angle data within each range are those greater than the 3 / 4 quartile plus 1.5 times the interquartile range. The multi-condition blade angle anomaly threshold is obtained, and its threshold is defined as follows: ,in, For the first Blade angle thresholds within a power range , For the first The 1 / 4 and 3 / 4 quantiles of blade angle data for each power range; The power segment data for each interval and the abnormal blade angle threshold under the multi-condition conditions are stored in the corresponding database.

6. The data mining and identification method for high-temperature power curtailment of wind turbine units as described in claim 5, characterized in that, The results of the high-temperature power rationing identification for the first generating unit were obtained, including: Collect real-time data of the generating unit, and search the database for the target power range corresponding to the unit power based on the unit power of the real-time data. Determine whether the current blade angle is greater than the multi-condition blade angle abnormality threshold corresponding to the target power range, and obtain the high temperature power limitation identification result of the first unit.

7. The data mining and identification method for high-temperature power curtailment of wind turbine units as described in claim 6, characterized in that, The threshold values ​​for abnormal temperature distribution under multiple operating conditions were obtained, including: Using power and ambient temperature as operating conditions, the standard wind turbine feature dataset is binned, and the ambient temperature data within each power range is further divided... Using intervals, the dataset within each grid interval is obtained. for: , , ,in, For ambient temperature, This refers to the gearbox oil temperature or the generator bearing temperature. The number of divisions for the power range. This indicates the number of temperature ranges within each power range; Calculate the mean and variance of the gearbox oil temperature or generator bearing temperature within each ambient temperature range for each power range of the target unit, based on... In principle, the normal range of each temperature distribution should be within three standard deviations above and below its mean. The threshold values ​​for the first gearbox oil temperature or generator bearing temperature distribution are defined as follows: ,in, Indicates the first Each power range Indicates the first The power range under the first A range of ambient temperatures; The first gearbox oil temperature or generator bearing temperature distribution threshold is statistically analyzed using quartiles. Following the box plot anomaly detection method, outliers in the temperature data within the grid interval are located in areas greater than the 3 / 4 quartile plus 1.5 times the interquartile range. This yields the second gearbox oil temperature or generator bearing temperature distribution threshold, defined as follows: ,in, For the first The power range under the first Gearbox oil temperature or generator bearing temperature within a specific ambient temperature range. , For the first The power range under the first The 1 / 4 and 3 / 4 quantiles of gearbox oil temperature or generator bearing temperature within a certain ambient temperature range; Based on the first gearbox oil temperature or generator bearing temperature distribution threshold and the second gearbox oil temperature or generator bearing temperature distribution threshold, determine the multi-condition temperature anomaly distribution threshold: ; The corresponding power range, ambient temperature range, and the multi-condition temperature anomaly distribution threshold are used. Store it in the corresponding database.

8. The data mining and identification method for high-temperature power curtailment of wind turbine units as described in claim 7, characterized in that, The results of the high-temperature power rationing identification for the second unit were determined, including: If the first unit's high-temperature power restriction identification result indicates that the unit is in a power restriction state, the target grid interval where the corresponding unit load and ambient temperature are located is searched from the database based on the unit load and ambient temperature. Determine whether the current gearbox oil temperature or generator bearing temperature is greater than the multi-condition temperature anomaly distribution threshold corresponding to the target grid interval, and determine the high temperature power limitation identification result of the second unit.

9. A data mining and identification system for high-temperature power limitation of wind turbine units, characterized in that, The step of implementing the data mining and identification method for high-temperature power curtailment of wind turbines according to any one of claims 1 to 8, wherein the data mining and identification system for high-temperature power curtailment of wind turbines comprises: The aggregation processing module is used to obtain wind turbine-related feature variables from the Kairosdb time-series database of the centralized control SCADA system, and to perform aggregation processing on the wind turbine-related feature variables to obtain an available wind turbine feature dataset. The data preprocessing module is used to preprocess the available wind turbine feature dataset by combining the wind turbine operating mechanism and the DBSCAN algorithm to obtain the standard wind turbine feature dataset. The blade angle anomaly threshold calculation module is used to divide the operating conditions into intervals based on power, and to calculate the normal interval of blade angle in the standard wind turbine feature dataset using the box plot method to obtain the multi-condition blade angle anomaly threshold. Based on the multi-condition blade angle anomaly threshold, the module identifies the high-temperature power limitation of the unit and obtains the first unit high-temperature power limitation identification result. The temperature anomaly distribution threshold calculation module is used to divide the operating conditions by power and ambient temperature, and to calculate the normal distribution of gearbox oil temperature and generator temperature on the standard wind turbine characteristic dataset using statistical analysis methods. This yields the multi-condition temperature anomaly distribution threshold, and the module identifies the high-temperature power restriction of the unit based on the first unit's high-temperature power restriction identification result and the multi-condition temperature anomaly distribution threshold, thus determining the high-temperature power restriction identification result of the second unit.

Images