A wind power operation data cleaning method for wind direction sector and related device

Abstract

The application discloses a wind power operation data cleaning method for wind direction sectors and related devices, belonging to the technical field of wind power generation. The method comprises the following steps: obtaining a historical operation data set of a wind turbine, the data set comprising wind speed, power, wind direction and time stamp fields; preprocessing the historical operation data set, eliminating invalid data with empty key fields or exceeding the physical range, and performing time alignment to obtain preprocessed data; dividing a full wind direction range of 0-360 degrees into several sectors, classifying the preprocessed data into corresponding sectors, and counting the data volume and data volume proportion of each sector; based on the relationship between the data volume proportion and the preset classification threshold, the several sectors are divided into dominant sectors, secondary dominant sectors and low-frequency sectors; assigning corresponding cleaning strategies to the dominant sectors, secondary dominant sectors and low-frequency sectors to obtain cleaned data of each sector; and integrating the cleaned data of each sector to merge into complete cleaned data.

Description

Technical Field

[0001] This invention belongs to the field of wind power generation technology, and relates to a method and related apparatus for cleaning wind power operation data of wind direction sectors. Background Technology

[0002] With the large-scale development of wind power, the analysis and application of wind farm operation data has become crucial, serving as the foundation for power prediction, performance evaluation, fault diagnosis, and optimized control. However, due to various reasons such as on-site communication equipment failures, unit operation status switching, sensor errors, and human intervention, the raw operation data (such as wind speed, power, and wind direction) collected by wind farms usually contains a large number of outliers, noise, and missing data, resulting in generally low data quality. Therefore, effective data cleaning and management are necessary before conducting in-depth analysis and application of these data. Currently, various methods have been developed in the field of wind power operation data cleaning. Common existing technologies include: (1) Statistical methods: such as the 3σ principle, which directly removes data points that exceed three times the standard deviation of the mean. (2) Data distribution-based methods: such as cluster analysis (e.g., K-means, DBSCAN), kernel density estimation (KDE), etc. These methods mainly analyze the scatter distribution pattern of data pairs such as wind speed and power, and identify and remove outliers that are far from the main data clusters. (3) Physical model-based method: Based on the theoretical power curve of the wind turbine, reasonable upper and lower limits are set, and data outside the limits are removed. Among these methods, the data distribution-based cleaning method (such as the method combined with clustering or kernel density estimation) is currently a more advanced and commonly used technique. This technical solution usually processes all wind speed-power data of a wind farm or a single unit over a period of time as a whole dataset. Its specific steps generally include: 1) obtaining the original wind speed-power data pairs; 2) using clustering algorithms or kernel density estimation algorithms to model the entire dataset to describe the distribution area of ​​normal data; 3) setting thresholds according to the model, judging data points falling outside the normal distribution area as outliers and removing or correcting them.

[0003] However, after in-depth research and analysis, it was found that the closest existing technical solution has the following significant drawbacks: (1) Ignoring the physical influence of wind direction limits cleaning accuracy: The operating characteristics of wind turbines are closely related to the incoming wind direction. Affected by factors such as wake effect, terrain obstruction, and interference between units, the theoretical output power corresponding to different incoming wind directions may differ significantly under the same wind speed. Existing methods mix data under different wind directions, and the "overall" data distribution model they construct is actually a fuzzy result of superimposing multiple sub-distributions with different wind directions. This results in a broad and inaccurate model boundary, making it impossible to accurately identify real abnormal data under specific wind directions, which can easily lead to "false cleaning" or "missed cleaning".

[0004] (2) The cleaning strategy is too simplistic and cannot meet the requirements for the integrity of differentiated data: Existing methods use a uniform and rigid cleaning standard for all data. The frequency of occurrence of different wind directions in wind farms varies greatly. For the dominant wind direction sector with abundant data, strict cleaning can obtain high-quality data; however, for the low-frequency wind direction sector with scarce data, if strict cleaning is also used, it may lead to the removal of a large number of the already limited effective samples, which will seriously damage the data integrity under that wind direction and make subsequent analysis impossible due to insufficient samples.

[0005] (3) Lack of in-depth analysis of the anomaly generation mechanism: Existing methods mostly focus on the "appearance" of data (scattered distribution) and do not fully consider the causes and frequency characteristics of the abnormal data. For example, some anomalies (such as momentary communication interruptions) may be randomly distributed in various wind directions; while other anomalies (such as persistent low power caused by terrain disturbance in a specific direction) have a clear wind direction correlation. A unified overall cleaning method is difficult to effectively address the latter. Summary of the Invention

[0006] The purpose of this invention is to solve the technical problems in the prior art of wind power operation data cleaning that do not consider the influence of wind direction and implement cleaning strategies according to the differences in data abundance, and to provide a wind power operation data cleaning method and related apparatus for wind direction sector.

[0007] To achieve the above objectives, the present invention employs the following technical solution: In a first aspect, the present invention discloses a method for cleaning wind power operation data by wind direction sector, comprising: Obtain the historical operation dataset of the wind turbine, which includes wind speed, power, wind direction, and timestamp fields; The historical operation dataset is preprocessed to remove invalid data with empty key fields or data exceeding the physical range, and time alignment is performed to obtain preprocessed data. Based on the 0°-360° all-wind-direction range, several sectors are divided, the preprocessed data is classified into the corresponding sectors, and the data volume and data volume ratio of each sector are counted. Based on the relationship between the data volume ratio and the preset classification threshold, several sectors are divided into dominant sectors, secondary dominant sectors and low-frequency sectors. The dominant sector, the secondary dominant sector and the low-frequency sector are assigned corresponding cleaning strategies to obtain the cleaned data of each sector. The cleaned data from each sector is integrated and merged into complete cleaned data based on the timestamp field.

[0008] Further improvements are made in the following aspects: Based on the 0°-360° all-directional wind range, several sectors are divided as follows: The wind direction range of 0°-360° is divided into 12 sectors, with each sector being 30°; or the wind direction range of 0°-360° is divided into 16 sectors, with each sector being 22.5°; or the historical wind rose diagram is used to divide the dominant wind direction range into sectors every 10° and the non-dominant wind direction range into sectors every 45°.

[0009] Based on the relationship between the data volume ratio and the preset classification threshold, several sectors are specifically divided into dominant sectors, secondary dominant sectors, and low-frequency sectors as follows: A preset classification threshold is set based on the data volume ratio, and the preset classification threshold includes a first threshold and a second threshold; the first threshold > the second threshold. When the data volume ratio is greater than the first threshold, the corresponding sector will be assigned to the dominant sector; When the first threshold > data volume ratio > second threshold, the corresponding sector will be assigned to the secondary dominant sector; When the second threshold exceeds the data volume ratio, the corresponding sector will be classified as a low-frequency sector.

[0010] The first threshold is set to 10%~15%, and the second threshold is set to 5%~8%.

[0011] The specific cleaning strategies assigned to the dominant sector, secondary dominant sector, and low-frequency sector are as follows: The dominant sector is subjected to a first cleaning strategy that employs a strict cleaning method based on kernel density estimation; the secondary dominant sector is subjected to a second cleaning strategy that combines physical boundaries and statistics; and the low-frequency sector is subjected to a third cleaning strategy that employs a relaxed cleaning method.

[0012] The first cleaning strategy is as follows: A two-dimensional kernel density estimation model is constructed using wind speed-power data pairs within the corresponding sector; a probability density threshold is set, and data points whose probability density values ​​estimated by the two-dimensional kernel density estimation model are lower than the probability density threshold are identified as outliers and removed. The second cleaning strategy is as follows: The theoretical power curve of the wind turbine is used to set upper and lower limit envelopes for preliminary screening, and data that obviously exceeds the physical limits is removed; the remaining data are then cleaned a second time using a relaxed statistical method or the KDE threshold T2. The third cleaning strategy is specifically as follows: Data that violates physical laws is removed, and all physically existing data points are retained. In specific application scenarios where the integrity of low-frequency wind direction data is required, the data of that sector is not cleaned, and the original data is retained for analysis.

[0013] The time range for obtaining the historical operating dataset of the wind turbine covers at least one complete annual cycle.

[0014] Secondly, this invention discloses a wind power operation data cleaning system for wind direction sectors, comprising: The data acquisition module is used to acquire the historical operating dataset of the wind turbine, which includes wind speed, power, wind direction and timestamp fields; The preprocessing module is used to preprocess the historical running dataset, remove invalid data with empty key fields or data that exceeds the physical range, and perform time alignment to obtain preprocessed data. The sector data statistics module is used to divide the entire wind direction range of 0°-360° into several sectors, classify the preprocessed data into the corresponding sectors, and count the amount of data and the proportion of data in each sector. The sector partitioning unit is used to divide several sectors into dominant sectors, secondary dominant sectors, and low-frequency sectors based on the relationship between the data volume ratio and the preset classification threshold. The partition cleaning module is used to assign corresponding cleaning strategies to the dominant sector, the secondary dominant sector and the low-frequency sector, and obtain the cleaned data of each sector. The post-cleaning data integration module is used to integrate the post-cleaning data of each sector and merge it into complete post-cleaning data based on the timestamp field.

[0015] Thirdly, the present invention discloses 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 computer program to implement the above-mentioned wind power operation data cleaning method for wind direction sectors.

[0016] Fourthly, the present invention discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned wind power operation data cleaning method for wind direction sectors.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention discloses a wind power operation data cleaning method for wind direction sectors. Firstly, it significantly improves cleaning accuracy. Traditional indiscriminate cleaning methods mix all wind direction data together for modeling, leading to interference between the operating characteristics of different wind directions. This causes the cleaning boundary to deviate from the actual operating conditions, easily resulting in "false cleaning" and "missed cleaning." This invention, by independently modeling by wind direction sector, can accurately capture the wind speed-power operating patterns of each sector, ensuring that the cleaning boundary closely matches the actual normal operating mode of the units under that wind direction. This significantly reduces the false positive and false negative rates of abnormal data, significantly improving the accuracy and reliability of data cleaning. Secondly, it achieves an intelligent balance between data quality and integrity. The distribution of incoming wind direction in wind farms is often extremely uneven, with abundant dominant wind direction data but scarce low-frequency wind direction data. This invention classifies sectors based on data abundance and configures differentiated cleaning strategies accordingly: for data-rich dominant sectors, a high-precision "fine cleaning" strategy is adopted to strictly remove outliers and produce high-quality data, providing a solid data foundation for core analyses such as unit performance evaluation and power curve optimization; for data-scarce low-frequency sectors, a more lenient "wide cleaning" or even "no cleaning" strategy is adopted to retain valuable original samples to the greatest extent, thereby ensuring data integrity across the entire wind direction range. This makes it possible to comprehensively evaluate the operating characteristics of units under extreme wind directions and complex wind conditions, avoiding the shortcomings of traditional methods that lose key information in low-frequency scenarios due to over-cleaning. Furthermore, this method has high flexibility and wide applicability. Users can flexibly adjust the sector classification threshold and the combination of cleaning strategies for each sector according to the geographical environment of the wind farm, wind resource characteristics, and specific needs of downstream application scenarios. For example, in mountainous wind farms with complex wind conditions, the accuracy of cleaning can be enhanced by increasing the granularity of sector division in the dominant wind direction and raising the anomaly detection threshold. In scenarios where low-frequency wind direction data is a key analytical object, the cleaning standards for low-frequency sectors can be further relaxed. This flexible and configurable design allows this method to adapt to different types of wind farms and diverse business needs, exhibiting strong engineering adaptability and scalability. Finally, the physical meaning of this method is clear, and the cleaning logic has good interpretability. It closely integrates the physical facts of uneven wind resource distribution and the variation of turbine operating characteristics with wind direction. Every step, from sector division to strategy selection, is based on objective data statistics and engineering experience, rather than relying on black-box models. This feature makes the cleaning process and results easier for engineers in the wind power field to understand, verify, and accept, thereby lowering the threshold for technology implementation and promotion, and facilitating rapid popularization and application within the industry. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a wind power operation data cleaning method for wind direction sector according to an embodiment of the present invention; Figure 2 This is a detailed step diagram of a wind power operation data cleaning method for wind direction sector in an embodiment of the present invention; Figure 3 This is a block diagram of a wind power operation data cleaning system for wind direction sectors according to an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0022] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0023] The present invention will now be described in further detail with reference to the accompanying drawings: See Figure 1 This invention discloses a method for cleaning wind power operation data by wind direction sector, comprising: S1, Obtain the historical operation dataset of the wind turbine, which includes wind speed, power, wind direction and timestamp fields; S2, preprocess the historical operation dataset, remove invalid data with empty key fields or exceeding the physical range, and perform time alignment to obtain preprocessed data; S3, divide the entire wind direction range of 0°-360° into several sectors, classify the preprocessed data into the corresponding sectors, and count the amount of data and the proportion of data in each sector; S4. Based on the relationship between the data volume ratio and the preset classification threshold, several sectors are divided into dominant sectors, secondary dominant sectors and low-frequency sectors. S5, assign corresponding cleaning strategies to the dominant sector, the secondary dominant sector and the low-frequency sector, and obtain the cleaned data of each sector; S6, integrate the cleaned data of each sector and merge them into complete cleaned data based on the timestamp field.

[0024] This invention discloses a wind power operation data cleaning method for wind direction sectors. First, the rationality of the cleaning is significantly improved. Traditional indiscriminate cleaning methods mix all wind direction data together for modeling, causing interference between the operating characteristics of different wind directions. This leads to the cleaning boundary deviating from the actual operating conditions, easily resulting in "false cleaning" and "missed cleaning." This invention, by modeling independently by wind direction sector, can accurately capture the wind speed-power operating pattern of each sector, making the cleaning boundary highly consistent with the actual normal operating mode of the units under that wind direction. This significantly reduces the false positive and false negative rates of abnormal data, and significantly improves the accuracy and reliability of data cleaning. Second, it achieves an intelligent balance between data quality and integrity. The distribution of incoming wind direction in wind farms is often extremely uneven, with abundant dominant wind direction data but scarce low-frequency wind direction data. This invention classifies sectors based on data abundance and configures differentiated cleaning strategies accordingly: for data-rich dominant sectors, a high-precision "fine cleaning" strategy is adopted to strictly remove outliers and produce high-quality data, providing a solid data foundation for core analyses such as unit performance evaluation and power curve optimization; for data-scarce low-frequency sectors, a more lenient "wide cleaning" or even "no cleaning" strategy is adopted to retain valuable original samples to the greatest extent, thereby ensuring data integrity across the entire wind direction range. This makes it possible to comprehensively evaluate the operating characteristics of units under extreme wind directions and complex wind conditions, avoiding the shortcomings of traditional methods that lose key information in low-frequency scenarios due to over-cleaning. Furthermore, this method has high flexibility and wide applicability. Users can flexibly adjust the sector classification threshold and the combination of cleaning strategies for each sector according to the geographical environment of the wind farm, wind resource characteristics, and specific needs of downstream application scenarios. For example, in mountainous wind farms with complex wind conditions, the accuracy of cleaning can be enhanced by increasing the granularity of sector division in the dominant wind direction and raising the anomaly detection threshold. In scenarios where low-frequency wind direction data is a key analytical object, the cleaning standards for low-frequency sectors can be further relaxed. This flexible and configurable design allows this method to adapt to different types of wind farms and diverse business needs, exhibiting strong engineering adaptability and scalability. Finally, the physical meaning of this method is clear, and the cleaning logic has good interpretability. It closely integrates the physical facts of uneven wind resource distribution and the variation of turbine operating characteristics with wind direction. Every step, from sector division to strategy selection, is based on objective data statistics and engineering experience, rather than relying on black-box models. This feature makes the cleaning process and results easier for engineers in the wind power field to understand, verify, and accept, thereby lowering the threshold for technology implementation and promotion, and facilitating rapid popularization and application within the industry.

[0025] See Figure 2 The present invention will be described in detail below with reference to specific embodiments: Step 1: Obtain the historical operation dataset of the wind turbine, which includes wind speed, power, wind direction and timestamp fields; Obtain the raw operational dataset to be cleaned. This dataset should contain at least the following fields: wind speed, power, wind direction, and timestamp. The time range of the data should be representative, typically covering at least one complete annual period to include various wind direction and speed conditions. This invention is applicable not only to data cleaning of a single wind turbine but also to data governance of an entire wind farm cluster. In this case, "wind direction" can refer to the prevailing wind direction representing the wind farm.

[0026] Step two involves preprocessing the historical data set to remove invalid data with empty key fields or data exceeding the physical measurement range, and performing time alignment to obtain preprocessed data. This aims to eliminate obviously invalid data and lay the foundation for subsequent analysis. Specific operations include, but are not limited to: (1) Delete records where the key fields (wind speed, power, wind direction) are NULL or significantly exceed the physical range (such as negative wind speed or power greater than 1.5 times the rated power).

[0027] (2) Time alignment of data to ensure that the correspondence between wind speed, power and wind direction data under the same timestamp is correct.

[0028] (3) Optionally, a simple smoothing process can be performed to eliminate some transient noise.

[0029] Step 3: Divide the entire wind direction range of 0°-360° into several sectors, classify the preprocessed data into the corresponding sectors, and count the amount of data and the proportion of data in each sector. (1) Divide into sectors Divide the omnidirectional wind range from 0° to 360° into equal parts. N Each sector. N The value can be set according to requirements, for example... N =12 (each sector is 30°) or N =16 (each sector is 22.5°). Sector division should ensure coverage of all possible wind directions. Sector division can be uneven; for example, finer sectors (e.g., 10° per sector) can be used near the prevailing wind direction, while wider sectors (e.g., 45° per sector) can be used in non-prevailing wind directions. The division can be based on historical wind rose diagrams.

[0030] (2) Data classification For each valid data record after preprocessing, it is categorized into the above categories based on its wind direction value. N One of the sectors. For example, if N =12, then the wind direction data between 0° and 30° (or 355° and 360° and 0° and 15°, depending on the definition of the 0° starting position) are assigned to sector 1, and so on.

[0031] (3) Data statistics Count the number of valid data records in each sector k ( k = 1, 2, …, N ) number of valid data records M _ k . Calculate the proportion of the data volume of each sector in the total valid data volume P _ k = M _ k / Σ M , where Σ M is the sum of the data volumes of all sectors.

[0032] Step 4, based on the relationship between the proportion of the data volume and the preset classification threshold, divide several sectors into dominant sectors, sub-dominant sectors and low-frequency sectors; According to the P _ k calculated in Step 3, set the classification threshold, and divide the N wind direction sectors into different types. The typical classification is: Dominant sector: sectors with a data proportion P _ k ≥ Th_high. Th_high is the first threshold, which can be set to 10% or 15% for example. Such sectors have rich data and represent the most common incoming wind directions of the wind farm.

[0033] Sub-dominant sector: sectors with a data proportion Th_low ≤ P _ k < Th_high. Th_low is the second threshold, which can be set to 5% or 8% for example. Such sectors have a medium data volume.

[0034] Low-frequency sector: sectors with a data proportion P _ k < Th_low. Such sectors have scarce data and low wind direction occurrence frequencies.

[0035] In addition to simply relying on the data volume ratio P _ k , it is also possible to perform comprehensive scoring classification by combining the data quality of the sectors (such as the proportion of initial abnormal points) and the importance weights of downstream applications.

[0036] Step 5, assign corresponding cleaning strategies to the dominant sectors, sub-dominant sectors and low-frequency sectors to obtain the cleaned data for each sector; Adopt cleaning methods with different strictness levels according to the characteristics of different sectors and the requirements for data integrity.

[0037] Execute the first cleaning strategy for the dominant sectors; This strategy aims to obtain the highest quality data and can employ a more rigorous model. For example, a rigorous cleaning method based on kernel density estimation (KDE) can be used: a two-dimensional kernel density estimation model is constructed using only wind speed-power data pairs within the same sector; a high probability density threshold T1 is set (e.g., corresponding to the 5th percentile of the cumulative probability distribution), and data points whose estimated probability density values ​​are lower than T1 are identified as outliers and removed. This method can accurately identify and remove marginal outliers when there is sufficient data.

[0038] For dominant sectors, in addition to KDE, strict clustering algorithms (such as OPTICS) or machine learning anomaly detection models can also be used; for secondary dominant sectors, iterative cleaning methods can be used to gradually tighten the cleaning criteria; for low-frequency sectors, data augmentation techniques (such as interpolation or generative models based on adjacent sector data) can be used to supplement a small amount of reasonable data after cleaning, rather than simply retaining it.

[0039] The second cleaning strategy is executed on the secondary dominant sector; This strategy seeks a balance between data quality and integrity, and can employ a compromise hybrid approach. For example, a combination of physical boundaries and statistical methods can be used: First, the theoretical power curve of the wind turbine is applied to set a reasonable upper and lower limit envelope for preliminary screening, eliminating data that clearly exceeds the physical limits; then, the remaining data is cleaned a second time using relatively lenient statistical methods (such as the 2.5σ principle) or a lower KDE threshold T2 (such as the 2% quantile).

[0040] The third cleaning strategy is applied to low-frequency sectors; This strategy prioritizes preserving sample integrity to the greatest extent possible. Extremely lenient cleaning methods can be employed; for example, only data that clearly violates physical laws (such as negative power or excessively exceeding rated power) can be removed, while retaining all physically plausible data points, even if they may contain some noise. In specific application scenarios with extremely high requirements for the integrity of low-frequency wind direction data, it may even be possible to choose not to perform any cleaning operations on the sector data and directly retain the raw data for analysis.

[0041] Step 6: Integrate the cleaned data of each sector and merge them into complete cleaned data based on the timestamp field.

[0042] The data from all sectors, after being cleaned separately, were re-merged according to their timestamps to form a complete and unified dataset. This dataset not only improves the overall quality but also retains data samples from all wind directions, making it suitable for various advanced analyses and applications.

[0043] The working principle of this invention is as follows: (1) Overall concept of “differentiated cleaning of sectors by wind direction”: Protect the basic idea and method framework of dividing wind power operation data into sectors according to the incoming wind direction and adopting different cleaning strategies for different sectors.

[0044] (2) Sector classification method: Based on the key statistical feature of the amount or proportion of data in each wind direction sector, the specific classification steps and threshold setting logic are used to classify sectors into at least two (preferably three: dominant, secondary dominant, and low frequency) types.

[0045] (3) Specific implementation of the differentiated cleaning strategy: A. For sectors with high data abundance (such as dominant sectors), adopt the first cleaning strategy (such as high threshold density estimation). B. For sectors in the data abundance (such as the secondary dominant sectors), adopt a second cleaning strategy with a lower degree of strictness than the first strategy (such as a hybrid strategy based on theoretical boundaries and statistics). C. For sectors with low data abundance (such as low-frequency sectors), adopt a third cleaning strategy with the core objective of preserving data integrity (such as only removing physically invalid values ​​or not cleaning at all).

[0046] (4) The entire process: Protect the complete steps from data acquisition, wind direction sector division and statistics, sector classification, to differential cleaning and final data integration.

[0047] See Figure 3 This invention discloses a wind power operation data cleaning system for wind direction sectors, comprising: The data acquisition module is used to acquire the historical operating dataset of the wind turbine, which includes wind speed, power, wind direction and timestamp fields; The preprocessing module is used to preprocess the historical running dataset, remove invalid data with empty key fields or data that exceeds the physical range, and perform time alignment to obtain preprocessed data. The sector data statistics module is used to divide the entire wind direction range of 0°-360° into several sectors, classify the preprocessed data into the corresponding sectors, and count the amount of data and the proportion of data in each sector. The sector partitioning unit is used to divide several sectors into dominant sectors, secondary dominant sectors, and low-frequency sectors based on the relationship between the data volume ratio and the preset classification threshold. The partition cleaning module is used to assign corresponding cleaning strategies to the dominant sector, the secondary dominant sector and the low-frequency sector, and obtain the cleaned data of each sector. The post-cleaning data integration module is used to integrate the post-cleaning data of each sector and merge it into complete post-cleaning data based on the timestamp field.

[0048] A third objective of this invention is to provide 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 computer program to implement the wind power operation data cleaning method for the wind direction sector.

[0049] The wind power operation data cleaning method for wind direction sector includes the following steps: Obtain the historical operation dataset of the wind turbine, which includes wind speed, power, wind direction, and timestamp fields; The historical operation dataset is preprocessed to remove invalid data with empty key fields or data exceeding the physical range, and time alignment is performed to obtain preprocessed data. Based on the 0°-360° all-wind-direction range, several sectors are divided, the preprocessed data is classified into the corresponding sectors, and the data volume and data volume ratio of each sector are counted. Based on the relationship between the data volume ratio and the preset classification threshold, several sectors are divided into dominant sectors, secondary dominant sectors and low-frequency sectors. The dominant sector, the secondary dominant sector and the low-frequency sector are assigned corresponding cleaning strategies to obtain the cleaned data of each sector. The cleaned data from each sector is integrated and merged into complete cleaned data based on the timestamp field.

[0050] The fourth objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements a method for cleaning wind power operation data by wind direction sector.

[0051] The wind power operation data cleaning method for wind direction sector includes the following steps: Obtain the historical operation dataset of the wind turbine, which includes wind speed, power, wind direction, and timestamp fields; The historical operation dataset is preprocessed to remove invalid data with empty key fields or data exceeding the physical range, and time alignment is performed to obtain preprocessed data. Based on the 0°-360° all-wind-direction range, several sectors are divided, the preprocessed data is classified into the corresponding sectors, and the data volume and data volume ratio of each sector are counted. Based on the relationship between the data volume ratio and the preset classification threshold, several sectors are divided into dominant sectors, secondary dominant sectors and low-frequency sectors. The dominant sector, the secondary dominant sector and the low-frequency sector are assigned corresponding cleaning strategies to obtain the cleaned data of each sector. The cleaned data from each sector is integrated and merged into complete cleaned data based on the timestamp field.

[0052] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0053] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0054] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0055] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0056] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for cleaning wind power operation data by wind direction sector, characterized in that, include: Obtain the historical operation dataset of the wind turbine, which includes wind speed, power, wind direction, and timestamp fields; The historical operation dataset is preprocessed to remove invalid data with empty key fields or data exceeding the physical range, and time alignment is performed to obtain preprocessed data. Based on the 0°-360° all-wind-direction range, several sectors are divided, the preprocessed data is classified into the corresponding sectors, and the data volume and data volume ratio of each sector are counted. Based on the relationship between the data volume ratio and the preset classification threshold, several sectors are divided into dominant sectors, secondary dominant sectors and low-frequency sectors. The dominant sector, the secondary dominant sector and the low-frequency sector are assigned corresponding cleaning strategies to obtain the cleaned data of each sector. The cleaned data from each sector is integrated and merged into complete cleaned data based on the timestamp field.

2. The wind power operation data cleaning method for wind direction sector according to claim 1, characterized in that, Based on the 0°-360° all-directional wind range, several sectors are divided as follows: The wind direction range of 0°-360° is divided into 12 sectors, with each sector being 30°; or the wind direction range of 0°-360° is divided into 16 sectors, with each sector being 22.5°; or the historical wind rose diagram is used to divide the dominant wind direction range into sectors every 10° and the non-dominant wind direction range into sectors every 45°.

3. The wind power operation data cleaning method for wind direction sector according to claim 1, characterized in that, Based on the relationship between the data volume ratio and the preset classification threshold, several sectors are specifically divided into dominant sectors, secondary dominant sectors, and low-frequency sectors as follows: A preset classification threshold is set based on the data volume ratio, and the preset classification threshold includes a first threshold and a second threshold; the first threshold > the second threshold. When the data volume ratio is greater than the first threshold, the corresponding sector will be assigned to the dominant sector; When the first threshold > data volume ratio > second threshold, the corresponding sector will be assigned to the secondary dominant sector; When the second threshold exceeds the data volume ratio, the corresponding sector will be classified as a low-frequency sector.

4. The wind power operation data cleaning method for wind direction sector according to claim 3, characterized in that, The first threshold is set to 10%~15%, and the second threshold is set to 5%~8%.

5. The wind power operation data cleaning method for wind direction sector according to claim 3, characterized in that, The specific cleaning strategies assigned to the dominant sector, secondary dominant sector, and low-frequency sector are as follows: A first cleaning strategy based on a strict cleaning method using kernel density estimation is applied to the dominant sector; a second cleaning strategy combining physical boundaries and statistics is applied to the secondary dominant sector. A third cleaning strategy of lenient cleaning is applied to the low-frequency sectors.

6. The wind power operation data cleaning method for wind direction sector according to claim 1, characterized in that, The first cleaning strategy is as follows: A two-dimensional kernel density estimation model is constructed using wind speed-power data pairs within the corresponding sector; a probability density threshold is set, and data points whose probability density values ​​estimated by the two-dimensional kernel density estimation model are lower than the probability density threshold are identified as outliers and removed. The second cleaning strategy is as follows: The theoretical power curve of the wind turbine is used to set upper and lower limit envelopes for preliminary screening, and data that obviously exceeds the physical limits is removed; the remaining data are then cleaned a second time using a relaxed statistical method or the KDE threshold T2. The third cleaning strategy is specifically as follows: Data that violates physical laws is removed, and all physically existing data points are retained. In specific application scenarios where the integrity of low-frequency wind direction data is required, the data of that sector is not cleaned, and the original data is retained for analysis.

7. The wind power operation data cleaning method for wind direction sectors according to claim 1, characterized in that, The time range for obtaining the historical operating dataset of the wind turbine covers at least one complete annual cycle.

8. A wind power operation data cleaning system for wind direction sectors, characterized in that, include: The data acquisition module is used to acquire the historical operating dataset of the wind turbine, which includes wind speed, power, wind direction and timestamp fields; The preprocessing module is used to preprocess the historical running dataset, remove invalid data with empty key fields or data that exceeds the physical range, and perform time alignment to obtain preprocessed data. The sector data statistics module is used to divide the entire wind direction range of 0°-360° into several sectors, classify the preprocessed data into the corresponding sectors, and count the amount of data and the proportion of data in each sector. The sector partitioning unit is used to divide several sectors into dominant sectors, secondary dominant sectors, and low-frequency sectors based on the relationship between the data volume ratio and the preset classification threshold. The partition cleaning module is used to assign corresponding cleaning strategies to the dominant sector, the secondary dominant sector and the low-frequency sector, and obtain the cleaned data of each sector. The post-cleaning data integration module is used to integrate the post-cleaning data of each sector and merge it into complete post-cleaning data based on the timestamp field.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the wind power operation data cleaning method for wind direction sectors as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program that, when executed by a processor, implements the wind power operation data cleaning method for wind direction sectors as described in any one of claims 1-7.

Images