Wind turbine generator abnormity alarm method and system based on characteristic parameter identification

Abstract

The invention relates to the technical field of wind turbine generator fault diagnosis, in particular to a wind turbine generator abnormity alarm method and system based on characteristic parameter identification, and the method comprises the steps: obtaining a temperature characteristic, obtaining a power characteristic, obtaining a vibration characteristic, obtaining a plurality of abnormity detection SVDD models, carrying out the abnormity detection of the temperature characteristic, the power characteristic and the vibration characteristic, if the distance detection values smaller than zero do not exist in the distance detection value set, blade image features are obtained, blade similarity total values are calculated, the blade similarity total values are summarized, a blade similarity total value set is obtained, if the distance detection values smaller than zero exist in the distance detection value set, an early warning threshold value is obtained, and the wind turbine generator is maintained. And on the basis of one or more replaced blades and the optimized wind turbine generator set, wind turbine generator set abnormity alarm based on characteristic parameter identification is completed. According to the invention, the abnormity of the wind turbine generator can be rapidly and accurately detected through a characteristic parameter identification technology.

Description

Technical Field

[0001] The present invention relates to the technical field of wind turbine fault diagnosis, and in particular, to a wind turbine abnormal alarm method, system, electronic device and computer-readable storage medium based on feature parameter recognition. Background Art

[0002] Feature parameter recognition refers to a technology that extracts key feature parameters from a large amount of data that can effectively characterize the state of a system or device, and uses these feature parameters to identify and judge the operating state, health status or abnormal conditions of the system or device. A wind turbine is a device that converts wind energy into electrical energy.

[0003] As the core device for converting wind energy into electrical energy, the operating state of a wind turbine is directly related to the efficiency and safety of a wind farm. However, traditional wind turbine monitoring methods mainly rely on manual inspections and simple sensor data analysis, lacking an intelligent abnormal detection and alarm mechanism, unable to provide timely early warnings and maintenance suggestions based on real-time data, and having low monitoring efficiency. Therefore, how to quickly and accurately detect the abnormalities of a wind turbine through feature parameter recognition technology is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] The present invention provides a wind turbine abnormal alarm method and a computer-readable storage medium based on feature parameter recognition, and its main purpose is to quickly and accurately detect the abnormalities of a wind turbine through feature parameter recognition technology, and improve the accuracy and reliability of the system.

[0005] To achieve the above object, an abnormal alarm method for a wind turbine based on feature parameter recognition provided by the present invention includes: receiving a wind turbine acquisition parameter instruction, obtaining a pre-constructed gearbox temperature data set of the wind turbine according to the wind turbine acquisition parameter instruction, and obtaining temperature features according to the gearbox temperature data set; obtaining a power generation power set of the wind turbine, obtaining power features according to the power generation power set, obtaining a vibration frequency set of the wind turbine, and obtaining vibration features according to the vibration frequency set; obtaining a plurality of anomaly detection SVDD models, and using the plurality of anomaly detection SVDD models to perform anomaly detection on the temperature features, power features and vibration features to obtain a distance detection value set; if there is no distance detection value less than zero in the distance detection value set, obtaining an initial blade image set of the wind turbine; sequentially extracting an initial blade image from the initial blade image set, and performing the following operations on each of the extracted initial blade images: segmenting the initial blade image to obtain a plurality of sub-region images, wherein each of the plurality of sub-region images has and only has one blade; sequentially extracting a sub-region image from the plurality of sub-region images, and performing the following operations on each of the extracted sub-region images: obtaining blade image features according to the sub-region image, retrieving standard blade image features in a pre-constructed blade image database according to the blade image features, and calculating a total blade similarity value according to the blade image features and the standard blade image features; summarizing the total blade similarity values to obtain a total blade similarity value set; judging whether there is a total blade similarity value in the total blade similarity value set that is not within a preset similarity value range; if there is a total blade similarity value in the total blade similarity value set that is not within a preset similarity value range, using a pre-constructed alarm device to issue a high-level warning, and replacing the blade corresponding to the total blade similarity value to obtain one or more replaced blades; if there is no total blade similarity value in the total blade similarity value set that is not within a preset similarity value range, returning to the step of sequentially extracting an initial blade image from the initial blade image set until all the initial blade images in the initial blade image set are extracted; if there is a distance detection value less than zero in the distance detection value set, obtaining a warning threshold according to the distance detection value set, and maintaining the wind turbine according to the warning threshold to obtain an optimized wind turbine; completing the abnormal alarm of the wind turbine based on feature parameter recognition based on one or more replaced blades and the optimized wind turbine.

[0006] Optionally, the obtaining a pre-constructed gearbox temperature data set of the wind turbine according to the wind turbine acquisition parameter instruction, and obtaining temperature features according to the gearbox temperature data set includes: obtaining a detection flight path, and obtaining a gearbox temperature image set of the wind turbine by using a pre-constructed unmanned aerial vehicle infrared temperature measurement, the detection flight path, a preset detection time and the wind turbine acquisition parameter instruction; obtaining a temperature data set according to the gearbox temperature image set, and obtaining a temperature average value set according to the temperature data set; and confirming temperature features based on the temperature average value set.

[0007] Optionally, the obtaining of the temperature data set according to the gearbox temperature image set includes: successively extracting a gearbox temperature image from the gearbox temperature image set, and performing the following operations on each of the extracted gearbox temperature images: performing a denoising operation on the gearbox temperature image to obtain a denoised temperature image, and performing a grayscale operation on the denoised temperature image to obtain a grayscale value set; performing the following operations on each grayscale value in the grayscale value set: obtaining a temperature value by using the grayscale value and a pre-constructed temperature-grayscale mapping table, where one temperature value corresponds to one timestamp and one position; summarizing the temperature values to obtain temperature data; and summarizing the temperature data to obtain a temperature data set.

[0008] Optionally, the obtaining of multiple anomaly detection SVDD models includes: obtaining multiple historical detection periods; successively extracting one historical detection period from the multiple historical detection periods, and performing the following operations on each of the extracted historical detection periods: detecting pre-constructed characteristic parameters of the wind turbine by using the historical detection period to obtain a characteristic data set, where the characteristic parameters include: temperature characteristics, power characteristics, and vibration characteristics; normalizing the characteristic data set to obtain a standard characteristic data set; summarizing the standard characteristic data sets to obtain multiple standard characteristic data sets; training a pre-constructed SVDD model by using the multiple standard characteristic data sets to obtain multiple initial SVDD models, where each initial SVDD model corresponds to a standard characteristic data set; testing the accuracy rate set and recall rate set of the multiple initial SVDD models by using pre-constructed test data, where one initial SVDD model corresponds to one accuracy rate and one recall rate; and if it is respectively confirmed that both the accuracy rate set and the recall rate set are within the corresponding preset intervals, then determining the multiple initial SVDD models as multiple anomaly detection SVDD models.

[0009] Optionally, the testing of the accuracy rate set and recall rate set of the multiple initial SVDD models by using the pre-constructed test data includes: calculating the accuracy rate set and recall rate set of the multiple initial SVDD models by using the test data, a pre-constructed accuracy rate formula, and a pre-constructed recall rate formula, where the test data includes multiple unit test data; where the accuracy rate formula is as follows: , where represents the accuracy rate, represents the number of unit test data that are actually abnormal and are predicted to be abnormal by the initial SVDD model, represents the number of unit test data that are actually normal and are predicted to be normal by the initial SVDD model, represents the number of unit test data that are actually normal but are predicted to be abnormal by the initial SVDD model Indicates the number of unit test data that are actually abnormal but predicted as normal by the initial SVDD model. Indicates the recall rate; among them, the recall rate formula is as follows: , where Indicates the recall rate.

[0010] Optionally, the abnormal detection of temperature features, power features, and vibration features is performed using multiple abnormal detection SVDD models to obtain a set of distance detection values, including: obtaining multiple optimal hyperspheres according to multiple abnormal detection SVDD models, where the abnormal detection SVDD models and the optimal hyperspheres are in one-to-one correspondence, and the optimal hypersphere includes: the center of the circle and the radius; extracting one target feature value from the temperature feature, power feature, and vibration feature respectively, and performing the following operations on the extracted target feature values: calculating the distance detection value between the target feature value and the center of the circle using the optimal hypersphere corresponding to the target feature value in the multiple optimal hyperspheres, where the calculation formula of the distance detection value is as follows: , where Indicates the distance detection value, Indicates the target feature value, Indicates the center of the circle of the optimal hypersphere, Indicates the radius of the optimal hypersphere, Indicates the transpose symbol; summarize the distance detection values to obtain a set of distance detection values.

[0011] Optionally, the calculation of the total blade similarity value according to the blade image features and the standard blade image features includes: obtaining a set of standard blade feature points according to the standard blade image features; sequentially extracting one standard blade feature point from the set of standard blade feature points, and performing the following operations on the extracted standard blade feature points: obtaining the blade edge image using the pre-constructed edge detection algorithm and the sub-region map, and obtaining a set of blade feature points according to the blade edge image, where the set of blade feature points includes multiple blade feature points, and each blade feature point includes: the position of the feature point, the direction of the feature point, the scale of the feature point, and the descriptor of the feature point; determining the feature points at the same position in the set of blade feature points according to the standard blade feature points, and calculating the feature point similarity between the feature points at the same position and the standard blade feature points; summarizing the feature point similarities to obtain a set of feature point similarities, and obtaining the total blade similarity value according to the set of feature point similarities.

[0012] Optionally, the calculation of the feature point similarity between the feature points at the same position and the standard blade feature points includes: calculating the feature point similarity between the standard blade feature points and the feature points at the same position using the pre-constructed similarity formula, where the similarity formula is as follows: , , , , , where represents the similarity of feature points, represents the coordinates of the standard blade feature points, represents the coordinates of the feature points at the same position, represents the normalization factor, represents the position similarity, represents the direction similarity, represents the direction angle of the standard blade feature points, represents the direction angle of the feature points at the same position, represents the scale similarity, represents the scale of the standard blade feature points, represents the scale of the feature points at the same position, represents the descriptor vector of the standard blade feature points, represents the descriptor vector of the feature points at the same position, represents the descriptor vector.

[0013] Optionally, obtaining the warning threshold according to the set of distance detection values includes: obtaining the number of sliding windows, obtaining the standard deviation of the set of distance detection values; calculating the warning threshold by using the number of sliding windows, the standard deviation, the set of distance detection values and a pre-constructed dynamic sliding window warning threshold formula, where the dynamic sliding window warning threshold formula is as follows: , where represents the warning threshold, represents the standard characteristic parameter value of the wind turbine, represents the standard deviation of the set of distance detection values, represents the total number of the set of distance detection values, represents the number of sliding windows, represents the sliding window movement increment.

[0014] To achieve the above object, the present invention also provides a wind turbine abnormal alarm system based on feature parameter recognition, including: a feature extraction module, configured to receive a wind turbine acquisition parameter instruction, obtain a pre-constructed gearbox temperature data set of the wind turbine according to the wind turbine acquisition parameter instruction, obtain temperature features according to the gearbox temperature data set, obtain a power generation power set of the wind turbine, obtain power features according to the power generation power set, obtain a vibration frequency set of the wind turbine, and obtain vibration features according to the vibration frequency set; an abnormal detection module, configured to obtain a plurality of abnormal detection SVDD models, and perform abnormal detection on the temperature features, power features, and vibration features by using the plurality of abnormal detection SVDD models to obtain a distance detection value set; a blade image analysis module, configured to, if there is no distance detection value less than zero in the distance detection value set, obtain an initial blade image set of the wind turbine, sequentially extract an initial blade image from the initial blade image set, and perform the following operations on each of the extracted initial blade images: segment the initial blade image to obtain a plurality of sub-region images, where each of the plurality of sub-region images has and only has one blade, sequentially extract a sub-region image from the plurality of sub-region images, and perform the following operations on each of the extracted sub-region images: obtain blade image features according to the sub-region image, retrieve standard blade image features in a pre-constructed blade image database according to the blade image features, calculate a total blade similarity value according to the blade image features and the standard blade image features, summarize the total blade similarity values to obtain a total blade similarity value set, determine whether there is a total blade similarity value in the total blade similarity value set that is not within a preset similarity value range, if there is a total blade similarity value in the total blade similarity value set that is not within the preset similarity value range, use a pre-constructed alarm device to issue a high-level warning, and replace the blade corresponding to the total blade similarity value to obtain one or more replaced blades, if there is no total blade similarity value in the total blade similarity value set that is not within the preset similarity value range, return to the step of sequentially extracting an initial blade image from the initial blade image set until all the initial blade images in the initial blade image set are extracted; a wind turbine maintenance module, configured to, if there is a distance detection value less than zero in the distance detection value set, obtain an early warning threshold according to the distance detection value set, perform maintenance on the wind turbine according to the early warning threshold to obtain an optimized wind turbine, and complete the abnormal alarm of the wind turbine based on feature parameter recognition based on one or more replaced blades and the optimized wind turbine.

[0015] To solve the above problems, the present invention also provides an electronic device, where the electronic device includes: a memory, storing at least one instruction; a processor, executing the instruction stored in the memory to implement the above-mentioned wind turbine abnormal alarm method based on feature parameter recognition.

[0016] To solve the above problems, the present invention further provides a computer-readable storage medium, in which at least one instruction is stored, and the at least one instruction is executed by a processor in an electronic device to implement the above-mentioned abnormal alarm method for a wind turbine based on feature parameter recognition.

[0017] To solve the problems described in the background art, the present invention receives the parameter acquisition instructions of the wind turbine, obtains the pre-constructed gearbox temperature data set of the wind turbine according to the parameter acquisition instructions of the wind turbine, and obtains temperature characteristics based on the gearbox temperature data set. Through the unmanned aerial vehicle infrared temperature measurement technology, combined with the preset detection flight path and time, the present invention can efficiently and accurately obtain the temperature image set of the gearbox of the wind turbine. The generation of the temperature data set not only ensures the integrity and consistency of the data, but also confirms the temperature characteristics by calculating the average temperature set, providing high-precision and reliable data support for the abnormal detection and fault warning of the wind turbine. Obtain the power generation power set of the wind turbine, obtain power characteristics based on the power generation power set, obtain the vibration frequency set of the wind turbine, and obtain vibration characteristics based on the vibration frequency set. The power characteristics of the present invention can reflect whether the output of the generator is stable and whether there are abnormal fluctuations, while the vibration characteristics can reveal whether mechanical components have imbalance, looseness or other structural problems. Obtain multiple anomaly detection SVDD models, and use the multiple anomaly detection SVDD models to perform anomaly detection on the temperature characteristics, power characteristics and vibration characteristics respectively to obtain a distance detection value set. By using the support vector data description (SVDD) model for anomaly detection, the present invention can effectively identify data points that deviate from the normal operation mode. By using multiple SVDD models to detect different characteristics respectively, the accuracy and reliability of anomaly detection can be improved. If there is no distance detection value less than zero in the distance detection value set, obtain the initial blade image set of the wind turbine. In the present invention, if the detection results of all characteristic parameters are within the normal range, that is, no abnormality is detected, then the overall health of the wind turbine is further ensured through the analysis of the blade images. The state of the blades has a direct impact on the performance of the wind turbine. Extract an initial blade image from the initial blade image set in sequence, and perform the following operations on each of the extracted initial blade images: segment the initial blade image to obtain multiple sub-region images, where each sub-region image in the multiple sub-region images has and only has one blade. By segmenting the blade image, the present invention can extract each blade separately, facilitating a more precise analysis of the state of each blade, which helps to discover damage or abnormalities of individual blades without being masked by the states of other blades. Extract a sub-region image from the multiple sub-region images in sequence, and perform the following operations on each of the extracted sub-region images: obtain blade image characteristics based on the sub-region image, retrieve standard blade image characteristics in the pre-constructed blade image database according to the blade image characteristics, and calculate the total blade similarity value based on the blade image characteristics and the standard blade image characteristics. By comparing with the standard blade image characteristics, the present invention can quantify the similarity of the blades, thereby judging whether there are damages or other abnormal conditions on the blades. This similarity calculation method based on image characteristics can provide objective and quantitative evaluation results, improving the accuracy of blade state evaluation. Summarize the total blade similarity values to obtain a total blade similarity value set.The present invention aggregates the total similarity values of all blades, enabling a comprehensive understanding of the overall condition of all blades in a wind turbine. This helps maintenance personnel comprehensively evaluate the health status of the wind turbine and formulate corresponding maintenance plans. By determining whether there are blade similarity total values in the blade similarity total value set that are not within the preset similarity total value range, the present invention can quickly screen out blades that may have problems by setting the range of the similarity total value. If there are blade similarity total values in the blade similarity total value set that are not within the preset similarity total value range, a high-level warning is issued using a pre-constructed alarm device, and the blade corresponding to the blade similarity total value is replaced to obtain one or more replacement blades. Through the high-level warning, the present invention can remind maintenance personnel to take emergency measures to prevent the further expansion of the failure and ensure the safe operation of the wind turbine. If there are no blade similarity total values in the blade similarity total value set that are not within the preset similarity total value range, the step of sequentially extracting an initial blade image from the initial blade image set is returned until all the initial blade images in the initial blade image set are extracted. The present invention ensures that all blade images are inspected without omission, which helps comprehensively evaluate the overall condition of the wind turbine blades and improve the reliability of the system. If there are distance detection values less than zero in the distance detection value set, a warning threshold is obtained based on the distance detection value set, and the wind turbine is maintained according to the warning threshold to obtain an optimized wind turbine. By optimizing the wind turbine, the present invention can timely repair potential failures, improve the operation efficiency and safety of the equipment. Based on one or more replacement blades and the optimized wind turbine, the abnormal alarm of the wind turbine based on feature parameter recognition is completed. By comprehensively applying feature parameter recognition and image analysis technologies, the present invention realizes comprehensive and refined abnormal detection and alarm of the wind turbine. Through timely maintenance and blade replacement, the occurrence of failures can be effectively prevented, the service life of the equipment can be extended, and the operation and maintenance costs can be reduced. Therefore, the present invention can quickly and accurately detect the abnormalities of the wind turbine through the feature parameter recognition technology, improving the accuracy and reliability of the system., Description of the Drawings

[0018] Figure 1 It is a schematic flowchart of a method for abnormal alarm of a wind turbine based on feature parameter recognition provided by an embodiment of the present invention; Figure 2 It is a functional module diagram of a system for abnormal alarm of a wind turbine based on feature parameter recognition provided by an embodiment of the present invention; Figure 3 It is a schematic structural diagram of an electronic device for implementing the method for abnormal alarm of a wind turbine based on feature parameter recognition provided by an embodiment of the present invention.

[0019] Description of the reference numerals: 1. Electronic device; 10. Processor; 11. Memory; 12. Bus.

[0020] The realization, functional features, and advantages of the present invention will be further described in conjunction with embodiments with reference to the accompanying drawings. Detailed implementation manners

[0021] It should be understood that the specific embodiments described herein are merely used to explain the present invention and are not used to limit the present invention.

[0022] An embodiment of the present application provides a method for abnormal alarm of a wind turbine based on feature parameter recognition. The execution subject of the method for abnormal alarm of a wind turbine based on feature parameter recognition includes, but is not limited to, at least one of electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided in this embodiment of the present application. In other words, the method for abnormal alarm of a wind turbine based on feature parameter recognition can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to: a single server, a server cluster, a cloud server, or a cloud server cluster, etc.

[0023] Referring to Figure 1 As shown, it is a schematic flowchart of a method for abnormal alarm of a wind turbine based on feature parameter recognition provided by an embodiment of the present invention. In this embodiment, the method for abnormal alarm of a wind turbine based on feature parameter recognition includes: S1. Receive a wind turbine collection parameter instruction, obtain a pre-constructed gearbox temperature data set of the wind turbine according to the wind turbine collection parameter instruction, and obtain temperature features according to the gearbox temperature data set.

[0024] It should be explained that the wind turbine collection parameter instruction refers to an instruction used to collect various operation parameters and status data from the wind turbine, and is generally issued by a computer control system.

[0025] Specifically, the obtaining of the pre-constructed gearbox temperature data set of the wind turbine according to the wind turbine collection parameter instruction and the obtaining of temperature features according to the gearbox temperature data set include: obtaining a detection flight path, and obtaining a gearbox temperature image set of the wind turbine by using the pre-constructed unmanned aerial vehicle infrared temperature measurement, the detection flight path, a preset detection time, and the wind turbine collection parameter instruction; obtaining a temperature data set according to the gearbox temperature image set, and obtaining a temperature average value set according to the temperature data set; and confirming temperature features based on the temperature average value set.

[0026] It should be noted that the step of obtaining the detection flight path is as follows: obtaining the position set of the wind turbine generator set, and planning the detection flight path of the unmanned aerial vehicle according to the position set. Infrared temperature measurement by the unmanned aerial vehicle means that the unmanned aerial vehicle is equipped with infrared temperature measurement equipment, which can measure the temperature distribution of the wind turbine generator set remotely and non-contactingly. The preset detection time refers to a preset time for measuring the temperature of each wind turbine in the wind turbine generator set at the preset time point. The gearbox temperature image set refers to the set of infrared temperature images of the gearbox of the wind turbine generator set. The temperature average value set refers to the set of temperature average values. The step of confirming the temperature feature based on the temperature average value set is: taking the temperature average value set as the temperature feature.

[0027] Specifically, the step of obtaining the temperature data set according to the gearbox temperature image set includes: sequentially extracting a gearbox temperature image from the gearbox temperature image set, and performing the following operations on each of the extracted gearbox temperature images: performing a denoising operation on the gearbox temperature image to obtain a denoised temperature image, and performing a grayscale operation on the denoised temperature image to obtain a grayscale value set; performing the following operations on each grayscale value in the grayscale value set: obtaining a temperature value by using the grayscale value and a pre-constructed temperature-grayscale mapping table, where a temperature value corresponds to a timestamp and a position; summarizing the temperature values to obtain temperature data; and summarizing the temperature data to obtain a temperature data set.

[0028] It should be noted that the step of performing a denoising operation on the gearbox temperature image to obtain a denoised temperature image is: performing a denoising operation on the gearbox temperature image by using median filtering to obtain a denoised temperature image. Median filtering refers to a filtering method that replaces the gray value of the central pixel by taking the median of the gray values of the pixels in the neighborhood. The median filtering in the embodiments of the present invention is used to remove random noise in the gearbox temperature image, making the image clearer and the extraction of temperature data more accurate.

[0029] It can be understood that the grayscale operation refers to the operation of converting a color image into a grayscale image, and the value of each pixel point represents the grayscale value of that point. The temperature-grayscale mapping table refers to a pre-constructed table that records the correspondence between grayscale values and temperature values. The temperature data set refers to the set obtained by summarizing the temperature data at all time points.

[0030] S2. Obtain the power generation power set of the wind turbine generator set, obtain the power feature according to the power generation power set, obtain the vibration frequency set of the wind turbine generator set, and obtain the vibration feature according to the vibration frequency set.

[0031] It should be noted that the step of obtaining the power generation power set of the wind turbine is as follows: Install a current sensor and a voltage sensor at the generator outlet of the wind turbine to obtain a current set and a voltage set, and obtain the power generation power set according to the current set and the voltage set. The steps of obtaining the power characteristics according to the power generation power set and obtaining the vibration characteristics according to the vibration frequency set are as follows: Obtain the power average value set according to the power data set, confirm the power characteristics according to the power average value set, obtain the vibration average value set according to the vibration frequency set, and confirm the vibration characteristics according to the vibration average value set. The steps of obtaining the temperature characteristics are the same as above and will not be elaborated here.

[0032] S3. Obtain a plurality of anomaly detection SVDD models, and use the plurality of anomaly detection SVDD models to perform anomaly detection on the temperature characteristics, power characteristics and vibration characteristics to obtain a set of distance detection values.

[0033] Specifically, the step of obtaining a plurality of anomaly detection SVDD models includes: obtaining a plurality of historical detection periods; sequentially extracting one historical detection period from the plurality of historical detection periods, and performing the following operations on each of the extracted historical detection periods: using the historical detection period to detect the pre-constructed characteristic parameters of the wind turbine to obtain a characteristic data set, where the characteristic parameters include: temperature characteristics, power characteristics and vibration characteristics; standardizing the characteristic data set to obtain a standard characteristic data set; summarizing the standard characteristic data sets to obtain a plurality of standard characteristic data sets; training the pre-constructed SVDD model with the plurality of standard characteristic data sets to obtain a plurality of initial SVDD models, where the initial SVDD models correspond to the standard characteristic data sets one by one; using the pre-constructed test data to test the accuracy rate set and the recall rate set of the plurality of initial SVDD models, where one initial SVDD model corresponds to one accuracy rate and one recall rate; if it is respectively confirmed that both the accuracy rate set and the recall rate set are within the corresponding preset intervals, then the plurality of initial SVDD models are confirmed as a plurality of anomaly detection SVDD models.

[0034] It should be noted that a plurality of historical detection periods refer to the operation data of the wind turbine collected at different time points or time periods, and are selected under different operating conditions and environmental factors to ensure the diversity and representativeness of the data. The step of standardizing the characteristic data set is as follows: Use the minimum-maximum normalization formula to convert the characteristic data to a common scale to eliminate the deviation caused by different dimensions and value ranges of different characteristics. The minimum-maximum normalization formula is as follows: , where represents the standardized characteristic data, represents the characteristic data, represents the minimum characteristic value in the characteristic data, represents the maximum characteristic value in the characteristic data.

[0035] Specifically, the multiple standard feature data sets refer to the feature data sets after standardization processing. The feature data set refers to the set of temperature features, power features, and vibration features extracted from the historical detection period. The SVDD model is a one-class classification method used to construct a minimum hypersphere to include normal data points, thereby realizing an anomaly detection model. The multiple initial SVDD models refer to multiple SVDD models trained using multiple standard feature data sets, and each standard feature data set corresponds to an initial SVDD model.

[0036] Specifically, testing the accuracy set and recall set of multiple initial SVDD models using pre-constructed test data includes: calculating the accuracy set and recall set of multiple initial SVDD models using the test data, the pre-constructed accuracy formula, and the pre-constructed recall formula, where the test data includes multiple unit test data. Among them, the accuracy formula is as follows: , where represents the accuracy, represents the number of unit test data that are actually abnormal and are predicted as abnormal by the initial SVDD model, represents the number of unit test data that are actually normal and are predicted as normal by the initial SVDD model, represents the number of unit test data that are actually normal but are predicted as abnormal by the initial SVDD model represents the number of unit test data that are actually abnormal but are predicted as normal by the initial SVDD model, represents the recall rate; among them, the recall rate formula is as follows: , where represents the recall rate.

[0037] It should be explained that the test data refers to a set of predefined data sets used to evaluate the performance of machine learning models. The test data described in the embodiments of the present invention includes multiple unit test data, and each unit test data represents a sample used to test the classification accuracy of the initial SVDD model for the state of the wind turbine. The accuracy set refers to the proportion of the number of samples correctly predicted by the initial SVDD model to the total number of samples, including the set of correctly predicted positive and negative class samples. The recall set refers to the proportion of the number of positive class samples correctly identified by the initial SVDD model to the total number of actual positive class samples, reflecting the set of sensitivity of the initial SVDD model to positive class samples. The unit test data refers to a single sample in the test data set, and each unit test data contains the characteristic parameters of the wind turbine.

[0038] Further, the use of multiple anomaly detection SVDD models to perform anomaly detection on temperature features, power features, and vibration features to obtain a set of distance detection values includes: obtaining multiple optimal hyperspheres according to multiple anomaly detection SVDD models, where the anomaly detection SVDD models and the optimal hyperspheres are in one-to-one correspondence, and the optimal hyperspheres include: the center of the circle and the radius; respectively extracting a target feature value from the temperature feature, power feature, and vibration feature, and performing the following operations on the extracted target feature values: calculating the distance detection value between the target feature value and the center of the circle using the optimal hypersphere corresponding to the target feature value among the multiple optimal hyperspheres, where the calculation formula for the distance detection value is as follows: , where, represents the distance detection value, represents the target feature value, represents the center of the optimal hypersphere, represents the radius of the optimal hypersphere, represents the transpose symbol; summarize the distance detection values to obtain a set of distance detection values.

[0039] It should be explained that multiple anomaly detection SVDD models refer to multiple models for anomaly detection obtained by training multiple SVDD models, each model for different feature datasets or different operating conditions. The set of distance detection values refers to the set of a series of distance detection values calculated when performing anomaly detection on temperature features, power features, and vibration features. Multiple optimal hyperspheres refer to corresponding to multiple anomaly detection SVDD models. Each SVDD model will obtain an optimal hypersphere during the training process, and this hypersphere can minimize the set of regions containing normal data.

[0040] Importantly, the step of obtaining multiple optimal hyperspheres according to multiple anomaly detection SVDD models is: when training the SVDD model, the goal is to find a hypersphere such that most normal data points are inside the hypersphere and abnormal data points are outside the hypersphere. The step of obtaining multiple optimal hyperspheres according to multiple anomaly detection SVDD models in the embodiments of the present invention is the prior art and will not be elaborated here.

[0041] It can be understood that the target feature value refers to a single feature value extracted from the temperature feature, power feature, and vibration feature for anomaly detection. The transpose symbol refers to the transpose operation of a vector or matrix. In the embodiments of the present invention, in the calculation formula for the distance detection value, represents the vector the square of the Euclidean distance of. In the embodiments of the present invention, one feature corresponds to one anomaly detection SVDD model and one optimal hypersphere, and one feature corresponds to multiple target feature values.

[0042] S4. If there is no distance detection value less than zero in the set of distance detection values, obtain the initial blade image set of the wind turbine generator set.

[0043] It should be explained that the step of "if there is no distance detection value less than zero in the set of distance detection values" means that all eigenvalues are located within their respective hyperspheres, that is, all monitored characteristic parameters are within the normal range and no abnormality is detected. The step of "obtain the initial blade image set of the wind turbine generator set" is to use a photographing device to photograph each wind turbine in the wind turbine generator set to obtain the initial blade image set.

[0044] S5. Sequentially extract an initial blade image from the initial blade image set, and perform the following operations on each of the extracted initial blade images: segment the initial blade image to obtain multiple sub-region images.

[0045] Specifically, each of the multiple sub-region images has and only has one blade.

[0046] It should be explained that the initial blade image set refers to the set of blade images collected from the wind turbine generator set. The initial blade image refers to a single image sequentially extracted from the initial blade image set, and each image captures the whole of a certain blade of the wind turbine generator set. The step of "segment the initial blade image to obtain multiple sub-region images" is to use a segmentation algorithm to segment the initial blade image, and each sub-region image corresponds to one blade to obtain multiple sub-region images.

[0047] S6. Sequentially extract a sub-region image from the multiple sub-region images, and perform the following operations on each of the extracted sub-region images: obtain the blade image features according to the sub-region image, retrieve the standard blade image features in the pre-constructed blade image database according to the blade image features, calculate the total blade similarity value according to the blade image features and the standard blade image features, and summarize the total blade similarity values to obtain the total blade similarity value set.

[0048] In detail, the calculation of the total blade similarity value according to the blade image features and the standard blade image features includes: obtaining the standard blade feature point set according to the standard blade image features; sequentially extracting a standard blade feature point from the standard blade feature point set, and performing the following operations on each of the extracted standard blade feature points: using the pre-constructed edge detection algorithm and the sub-region image to obtain the blade edge image, and obtaining the blade feature point set according to the blade edge image, where the blade feature point set includes multiple blade feature points, and each blade feature point includes: feature point position, feature point direction, feature point scale, and feature point descriptor; determining the same-position feature points in the blade feature point set according to the standard blade feature points, and calculating the feature point similarity between the same-position feature points and the standard blade feature points; summarizing the feature point similarities to obtain the feature point similarity set, and obtaining the total blade similarity value according to the feature point similarity set.

[0049] It should be explained that the standard leaf image feature refers to the feature information of the leaf image stored in the leaf image database. The standard leaf feature point set refers to the set of feature points extracted from the standard leaf image, and each feature point includes position, direction, scale, and descriptor. The step of obtaining the leaf feature point set according to the leaf edge image is as follows: using a feature point detection algorithm to extract feature points on the leaf edge image to obtain the leaf feature point set. The feature point scale refers to the scale size of the feature point in the image. The feature point descriptor refers to the vector describing the features of the area around the feature point. The same-position feature points refer to the feature points found in the standard leaf feature point set that have the same position as the actual leaf feature points.

[0050] Importantly, the feature point similarity refers to the degree of similarity between the actual leaf feature points and the standard leaf feature points. The feature point similarity set refers to the set of similarity values of all the same-position feature points. The step of obtaining the total leaf similarity value according to the feature point similarity set is as follows: calculating the average value of the feature point similarity set to obtain the total leaf similarity value.

[0051] Exemplarily, the feature point similarity set is [0.85, 0.90, 0.88, 0.92, 0.87]. Calculate the average value of the feature point similarity set, and use the following formula to calculate the average value: , where represents the average value.

[0052] Specifically, the calculation of the feature point similarity between the same-position feature points and the standard leaf feature points includes: using a pre-constructed similarity formula to calculate the feature point similarity between the standard leaf feature points and the same-position feature points, where the similarity formula is as follows: , , , , , where represents the feature point similarity, represents the coordinates of the standard leaf feature point, represents the coordinates of the same-position feature point, represents the normalization factor, represents the position similarity, represents the direction similarity, represents the direction angle of the standard leaf feature point, represents the direction angle of the same-position feature point, represents the scale similarity, represents the scale of the standard leaf feature point, represents the scale of the same-position feature point, The descriptor vector representing the standard blade feature points, The descriptor vector representing the feature points at the same position, The descriptor vector.

[0053] It should be explained that the steps for the coordinates of the standard blade feature points are as follows: taking the lower left corner of the standard blade image as the origin, the horizontal right direction as the x-axis, and the vertical upward direction as the y-axis, constructing a two-dimensional reference coordinate system, obtaining the width and height of the standard blade image, and the value range of the abscissa of the standard blade feature points is , representing the width of the standard blade image, and the value range of the ordinate of the standard blade feature points is , representing the height of the standard blade image.

[0054] S7. Determine whether there is a blade similarity total value in the blade similarity total value set that is not within the preset similarity total value interval. If there is a blade similarity total value in the blade similarity total value set that is not within the preset similarity total value interval, use the pre-constructed alarm device to issue a high-level warning, and replace the blade corresponding to the blade similarity total value to obtain one or more replacement blades. If there is no blade similarity total value in the blade similarity total value set that is not within the preset similarity total value interval, return to the step of sequentially extracting an initial blade image from the initial blade image set until all the initial blade images in the initial blade image set are extracted.

[0055] It should be explained that the alarm device refers to a device that issues a warning signal when an abnormal situation is detected to remind personnel to pay attention to the abnormal situation. A high-level warning refers to a special warning signal issued by the alarm device to identify a serious or urgent abnormal situation. The step of replacing the blade corresponding to the blade similarity total value to obtain one or more replacement blades is as follows: professional equipment and technical personnel are required to replace the abnormal wind turbine to obtain one or more replacement blades. The similarity total value interval refers to a preset numerical range used to determine whether the blade similarity total value is normal.

[0056] S8. If there is a distance detection value less than zero in the distance detection value set, obtain the warning threshold according to the distance detection value set, and maintain the wind turbine according to the warning threshold to obtain an optimized wind turbine.

[0057] It should be explained that maintenance refers to the operation of replacing components in the wind turbine that have reached the service life or have failed or optimizing control parameters to improve the operating efficiency of the wind turbine.

[0058] Specifically, obtaining the warning threshold according to the set of distance detection values includes: obtaining the number of sliding windows and the standard deviation of the set of distance detection values; calculating the warning threshold by using the number of sliding windows, the standard deviation, the set of distance detection values, and a pre-constructed dynamic sliding window warning threshold formula. The dynamic sliding window warning threshold formula is as follows: , where represents the warning threshold, represents the standard characteristic parameter value of the wind turbine, represents the standard deviation of the set of distance detection values, represents the total number of the set of distance detection values, represents the number of sliding windows, represents the sliding window movement increment.

[0059] It should be explained that the step of obtaining the number of sliding windows is: according to the historical operation data of the wind turbine, analyzing the time period of the set of distance detection values when a fault or anomaly occurs, and obtaining the number of sliding windows according to the time period. Exemplarily, if sampling is performed once per minute and it is desired to analyze the data of the recent 10 minutes, the sliding window size can be set to 10.

[0060] It can be understood that the standard deviation of the set of distance detection values is a statistical indicator that measures the volatility of the set of distance detection values. The sliding window movement increment is the step size by which the sliding window moves each time when calculating the warning threshold. The step of obtaining the standard deviation of the set of distance detection values is: using the following formula: , where represents the standard deviation, represents the index of the distance detection value, represents the th value in the set of distance detection values, represents the average value of the set of distance detection values.

[0061] S9. Based on one or more replaced blades and an optimized wind turbine, an abnormal alarm of the wind turbine based on feature parameter recognition is completed.

[0062] It should be explained that optimizing the wind turbine refers to the wind turbine after maintaining a faulty wind turbine. Replacing the blade refers to the blade obtained by replacing a problematic blade.

[0063] To solve the problems described in the background art, the present invention receives the parameter acquisition instructions of the wind turbine, obtains the pre-constructed gearbox temperature data set of the wind turbine according to the parameter acquisition instructions of the wind turbine, and obtains the temperature characteristics according to the gearbox temperature data set. Through the unmanned aerial vehicle infrared temperature measurement technology, combined with the preset detection flight path and time, the present invention can efficiently and accurately obtain the temperature image set of the gearbox of the wind turbine. The generation of the temperature data set not only ensures the integrity and consistency of the data, but also confirms the temperature characteristics by calculating the average temperature set, providing high-precision and reliable data support for the abnormal detection and fault warning of the wind turbine. Obtain the power generation power set of the wind turbine, obtain the power characteristics according to the power generation power set, obtain the vibration frequency set of the wind turbine, and obtain the vibration characteristics according to the vibration frequency set. The power characteristics of the present invention can reflect whether the output of the generator is stable and whether there are abnormal fluctuations, while the vibration characteristics can reveal whether there are imbalances, looseness or other structural problems in the mechanical components. Obtain multiple anomaly detection SVDD models, and use the multiple anomaly detection SVDD models to perform anomaly detection on the temperature characteristics, power characteristics and vibration characteristics to obtain a distance detection value set. By using the support vector data description (SVDD) model for anomaly detection, the present invention can effectively identify the data points deviating from the normal operation mode. By using multiple SVDD models to detect different characteristics respectively, the accuracy and reliability of anomaly detection can be improved. If there is no distance detection value less than zero in the distance detection value set, obtain the initial blade image set of the wind turbine. In the present invention, if the detection results of all characteristic parameters are within the normal range, that is, no anomaly is detected, then further analyze the blade images to ensure the overall health of the wind turbine. The state of the blades has a direct impact on the performance of the wind turbine. Extract an initial blade image from the initial blade image set in sequence, and perform the following operations on each of the extracted initial blade images: Segment the initial blade image to obtain multiple sub-region images, where each sub-region image in the multiple sub-region images has and only has one blade. By segmenting the blade image, the present invention can extract each blade separately, facilitating more accurate analysis of the state of each blade, which helps to discover damage or anomalies of individual blades without being masked by the states of other blades. Extract a sub-region image from the multiple sub-region images in sequence, and perform the following operations on each of the extracted sub-region images: Obtain the blade image characteristics according to the sub-region image, retrieve the standard blade image characteristics in the pre-constructed blade image database according to the blade image characteristics, and calculate the total blade similarity value according to the blade image characteristics and the standard blade image characteristics. By comparing with the standard blade image characteristics, the present invention can quantify the similarity of the blades, thereby judging whether there are damages or other abnormal conditions on the blades. This similarity calculation method based on image characteristics can provide objective and quantitative evaluation results, improving the accuracy of blade state evaluation. Summarize the total blade similarity values to obtain a total blade similarity value set.The present invention summarizes the total similarity values of all blades, enabling a comprehensive understanding of the overall condition of all blades of a wind turbine. This helps the operation and maintenance personnel comprehensively evaluate the health status of the wind turbine and formulate corresponding maintenance plans. It determines whether there are blade similarity values in the total blade similarity value set that are not within the preset similarity value range. By setting the range of the total similarity value, the present invention can quickly screen out blades that may have problems. If there are blade similarity values in the total blade similarity value set that are not within the preset similarity value range, a high-level warning is issued using a pre-constructed alarm device, and the blade corresponding to the blade similarity value is replaced to obtain one or more replacement blades. Through the high-level warning, the present invention can remind the operation and maintenance personnel to take emergency measures to prevent the further expansion of the fault and ensure the safe operation of the wind turbine. If there are no blade similarity values in the total blade similarity value set that are not within the preset similarity value range, the step of sequentially extracting an initial blade image from the initial blade image set is returned until all the initial blade images in the initial blade image set are extracted. The present invention ensures that all blade images are inspected without omission, which helps comprehensively evaluate the overall condition of the wind turbine blades and improve the reliability of the system. If there are distance detection values less than zero in the distance detection value set, the warning threshold is obtained according to the distance detection value set, and the wind turbine is maintained according to the warning threshold to obtain an optimized wind turbine. By optimizing the wind turbine, the present invention can timely repair potential faults, improve the operation efficiency and safety of the equipment. Based on one or more replacement blades and the optimized wind turbine, the abnormal alarm of the wind turbine based on feature parameter recognition is completed. By comprehensively applying feature parameter recognition and image analysis technologies, the present invention realizes comprehensive and refined abnormal detection and alarm of the wind turbine. Through timely maintenance and blade replacement, the occurrence of faults can be effectively prevented, the service life of the equipment can be extended, and the operation and maintenance costs can be reduced. Therefore, the present invention can quickly and accurately detect the abnormalities of the wind turbine through feature parameter recognition technology, improving the accuracy and reliability of the system.,

[0064] As Figure 2 shown, it is a functional module diagram of an abnormal alarm system for a wind turbine based on feature parameter recognition provided by an embodiment of the present invention.

[0065] The abnormal alarm system 100 of a wind turbine based on feature parameter recognition according to the present invention can be installed in an electronic device. According to the implemented functions, the abnormal alarm system 100 of a wind turbine based on feature parameter recognition can include a feature extraction module 101, an anomaly detection module 102, a blade image analysis module 103, and a wind turbine maintenance module 104. The modules described in the present invention can also be referred to as units, which refer to a series of computer program segments that can be executed by a processor of an electronic device and can complete fixed functions, and are stored in the memory of the electronic device; the feature extraction module 101 is configured to receive a wind turbine acquisition parameter instruction, obtain a pre-constructed gearbox temperature data set of the wind turbine according to the wind turbine acquisition parameter instruction, obtain temperature features according to the gearbox temperature data set, obtain a power generation power set of the wind turbine, obtain power features according to the power generation power set, obtain a vibration frequency set of the wind turbine, and obtain vibration features according to the vibration frequency set; the anomaly detection module 102 is configured to obtain a plurality of anomaly detection SVDD models, and perform anomaly detection on temperature features, power features, and vibration features by using the plurality of anomaly detection SVDD models to obtain a distance detection value set; the blade image analysis module 103 is configured to, if there is no distance detection value less than zero in the distance detection value set, obtain an initial blade image set of the wind turbine, sequentially extract an initial blade image from the initial blade image set, and perform the following operations on each of the extracted initial blade images: segment the initial blade image to obtain a plurality of sub-region images, where each of the plurality of sub-region images has and only has one blade, sequentially extract a sub-region image from the plurality of sub-region images, and perform the following operations on each of the extracted sub-region images: obtain blade image features according to the sub-region image, retrieve standard blade image features in a pre-constructed blade image database according to the blade image features, calculate a total blade similarity value according to the blade image features and the standard blade image features, summarize the total blade similarity values to obtain a total blade similarity value set, determine whether there is a total blade similarity value in the total blade similarity value set that is not within a preset similarity value range, if there is a total blade similarity value in the total blade similarity value set that is not within a preset similarity value range, then use a pre-constructed alarm device to issue a high-level warning, and replace the blade corresponding to the total blade similarity value to obtain one or more replacement blades, if there is no total blade similarity value in the total blade similarity value set that is not within a preset similarity value range, then return to the step of sequentially extracting an initial blade image from the initial blade image set until all the initial blade images in the initial blade image set are extracted; the wind turbine maintenance module 104 is configured to, if there is a distance detection value less than zero in the distance detection value set, obtain a warning threshold according to the distance detection value set, perform maintenance on the wind turbine according to the warning threshold to obtain an optimized wind turbine, and complete the abnormal alarm of the wind turbine based on feature parameter recognition based on one or more replacement blades and the optimized wind turbine.

[0066] Specifically, when the modules in the wind turbine abnormal alarm system 100 based on feature parameter recognition in the embodiments of the present invention are used, they adopt the same technical means as those in the Figure 1 wind turbine abnormal alarm method based on feature parameter recognition described above, and can produce the same technical effects, which will not be elaborated here.

[0067] As Figure 3 shown, it is a schematic structural diagram of an electronic device for implementing the wind turbine abnormal alarm method based on feature parameter recognition provided by an embodiment of the present invention.

[0068] The electronic device 1 may include a processor 10, a memory 11, and a bus 12, and may further include a computer program stored in the memory 11 and executable on the processor 10, such as a wind turbine abnormal alarm method program based on feature parameter recognition.

[0069] Among them, the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (such as SD or DX memory, etc.), magnetic memory, magnetic disk, optical disk, etc. The memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as the mobile hard disk of the electronic device 1. The memory 11 may also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 1. Further, the memory 11 also includes the internal storage unit of the electronic device 1 and the external storage device. The memory 11 can not only be used to store application software installed in the electronic device 1 and various types of data, such as the code of the wind turbine abnormal alarm method program based on feature parameter recognition, but also be used to temporarily store data that has been output or will be output.

[0070] In some embodiments, the processor 10 may be composed of an integrated circuit. For example, it may be composed of a single packaged integrated circuit, or may be composed of multiple packaged integrated circuits with the same or different functions, including a combination of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control core (Control Unit) of the electronic device, connecting various components of the entire electronic device through various interfaces and circuits, and by running or executing programs or modules stored in the memory 11 (such as the program for abnormal alarm method of wind turbine based on feature parameter recognition, etc.), and by calling the data stored in the memory 11, to execute various functions of the electronic device 1 and process data.

[0071] The bus 12 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. This bus 12 can be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to enable connection communication between the memory 11 and at least one processor 10, etc.

[0072] Figure 3 Only the electronic device with components is shown. Those skilled in the art can understand that Figure 3 the shown structure does not constitute a limitation on the electronic device 1, and it may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0073] For example, although not shown, the electronic device 1 may further include a power source (such as a battery) for powering each component. Preferably, the power source can be logically connected to the at least one processor 10 through a power management device, so as to implement functions such as charge management, discharge management, and power consumption management through the power management device. The power source may also include any components such as one or more DC or AC power sources, a recharge device, a power failure detection circuit, a power converter or inverter, and a power status indicator. The electronic device 1 may also include various sensors, a Bluetooth module, a Wi-Fi module, etc., which will not be elaborated here.

[0074] Furthermore, the electronic device 1 may further include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0075] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), and optionally, the user interface may also be a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display the information processed in the electronic device 1 and to display a visual user interface.

[0076] The program of the wind turbine abnormal alarm method based on feature parameter recognition stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve the following: receiving the wind turbine acquisition parameter instruction, obtaining the pre-constructed gearbox temperature data set of the wind turbine according to the wind turbine acquisition parameter instruction, and obtaining the temperature feature according to the gearbox temperature data set; obtaining the power generation power set of the wind turbine, obtaining the power feature according to the power generation power set, obtaining the vibration frequency set of the wind turbine, and obtaining the vibration feature according to the vibration frequency set; obtaining multiple anomaly detection SVDD models, and using the multiple anomaly detection SVDD models to perform anomaly detection on the temperature feature, power feature and vibration feature respectively to obtain a set of distance detection values; if there is no distance detection value less than zero in the set of distance detection values, obtaining the initial blade image set of the wind turbine; sequentially extracting an initial blade image from the initial blade image set, and performing the following operations on each of the extracted initial blade images: segmenting the initial blade image to obtain a plurality of sub-region images, wherein each sub-region image in the plurality of sub-region images has and only has one blade; sequentially extracting a sub-region image from the plurality of sub-region images, and performing the following operations on each of the extracted sub-region images: obtaining the blade image feature according to the sub-region image, retrieving the standard blade image feature in the pre-constructed blade image database according to the blade image feature, and calculating the total blade similarity value according to the blade image feature and the standard blade image feature; summarizing the total blade similarity values to obtain a set of total blade similarity values; judging whether there is a total blade similarity value in the set of total blade similarity values that is not within the preset similarity value range; if there is a total blade similarity value in the set of total blade similarity values that is not within the preset similarity value range, using the pre-constructed alarm device to issue a high-level warning, and replacing the blade corresponding to the total blade similarity value to obtain one or more replaced blades; if there is no total blade similarity value in the set of total blade similarity values that is not within the preset similarity value range, returning to the step of sequentially extracting an initial blade image from the initial blade image set until all the initial blade images in the initial blade image set are extracted; if there is a distance detection value less than zero in the set of distance detection values, obtaining the warning threshold according to the set of distance detection values, and maintaining the wind turbine according to the warning threshold to obtain an optimized wind turbine; completing the wind turbine abnormal alarm based on feature parameter recognition based on one or more replaced blades and the optimized wind turbine.

[0077] Specifically, the specific implementation method of the above instructions by the processor 10 can refer to Figures 1 to 3 the description of the relevant steps in the corresponding embodiment, which will not be elaborated here.

[0078] Furthermore, if the modules / units integrated in the electronic device 1 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM, Read-Only Memory).

[0079] The present invention also provides a computer-readable storage medium storing a computer program, which when executed by a processor of an electronic device, can achieve: receiving a wind turbine set acquisition parameter instruction, obtaining a pre-constructed gearbox temperature data set of the wind turbine set according to the wind turbine set acquisition parameter instruction, and obtaining temperature characteristics according to the gearbox temperature data set; obtaining a power generation power set of the wind turbine set, obtaining power characteristics according to the power generation power set, obtaining a vibration frequency set of the wind turbine set, and obtaining vibration characteristics according to the vibration frequency set; obtaining a plurality of anomaly detection SVDD models, and performing anomaly detection on the temperature characteristics, power characteristics and vibration characteristics by using the plurality of anomaly detection SVDD models to obtain a distance detection value set; if there is no distance detection value less than zero in the distance detection value set, obtaining an initial blade image set of the wind turbine set; sequentially extracting an initial blade image from the initial blade image set, and performing the following operations on each of the extracted initial blade images: segmenting the initial blade image to obtain a plurality of sub-region images, wherein each sub-region image in the plurality of sub-region images has and only has one blade; sequentially extracting a sub-region image from the plurality of sub-region images, and performing the following operations on each of the extracted sub-region images: obtaining blade image characteristics according to the sub-region image, retrieving standard blade image characteristics in a pre-constructed blade image database according to the blade image characteristics, and calculating a total blade similarity value according to the blade image characteristics and the standard blade image characteristics; summarizing the total blade similarity values to obtain a total blade similarity value set; judging whether there is a total blade similarity value in the total blade similarity value set that is not within a preset total similarity value interval; if there is a total blade similarity value in the total blade similarity value set that is not within a preset total similarity value interval, using a pre-constructed alarm device to issue a high-level warning, and replacing the blade corresponding to the total blade similarity value to obtain one or more replaced blades; if there is no total blade similarity value in the total blade similarity value set that is not within a preset total similarity value interval, returning to the step of sequentially extracting an initial blade image from the initial blade image set until all the initial blade images in the initial blade image set are extracted; if there is a distance detection value less than zero in the distance detection value set, obtaining a warning threshold according to the distance detection value set, and performing maintenance on the wind turbine set according to the warning threshold to obtain an optimized wind turbine set; completing the anomaly alarm of the wind turbine set based on feature parameter recognition based on one or more replaced blades and the optimized wind turbine set.

[0080] In several embodiments provided by the present invention, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative, and there may be other division methods in actual implementation.

[0081] The module described as a separation component may or may not be physically separated. The component shown as a module may or may not be a physical unit, that is, it may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0082] In addition, each functional module in various embodiments of the present invention may be integrated in a processing unit, may exist separately as individual physical units, or two or more units may be integrated in one unit. The above integrated unit may be implemented in the form of hardware or in the form of a combination of hardware and software functional modules.

[0083] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A wind turbine abnormality alarm method based on characteristic parameter identification, characterized in that: The method comprises: receiving a wind turbine generator set acquisition parameter instruction, acquiring a pre-constructed gearbox temperature data set of the wind turbine generator set according to the wind turbine generator set acquisition parameter instruction, and acquiring temperature characteristics according to the gearbox temperature data set; acquiring a power generation set of the wind turbine generator set, acquiring power characteristics according to the power generation set, acquiring a vibration frequency set of the wind turbine generator set, and acquiring vibration characteristics according to the vibration frequency set; acquiring multiple anomaly detection SVDD models, and using multiple anomaly detection SVDD models to perform anomaly detection on temperature characteristics, power characteristics and vibration characteristics to obtain a distance detection value set; if there is no distance detection value less than zero in the distance detection value set, then acquiring an initial blade image set of the wind turbine generator set; extracting an initial blade image from the initial blade image set in turn, and performing the following operations on the extracted initial blade images: segmenting the initial blade image to obtain multiple sub-region maps, wherein each of the multiple sub-region maps has and only has one blade; extracting a sub-region map from the multiple sub-region maps in turn, and performing the following operations on the extracted sub-region maps: acquiring blade image characteristics according to the sub-region map, and according to the blade image characteristics, The standard blade image features are retrieved from the established blade image database, and the total blade similarity value is calculated according to the blade image features and the standard blade image features; the total blade similarity values ​​are summarized to obtain a blade similarity total value set; it is determined whether there is a blade similarity total value that is not in a preset similarity total value interval in the blade similarity total value set; if there is a blade similarity total value that is not in the preset similarity total value interval in the blade similarity total value set, a high-level warning is issued by using a pre-built alarm device, and the blade corresponding to the blade similarity total value is replaced to obtain one or more replacement blades; if there is no blade similarity total value that is not in the preset similarity total value interval in the blade similarity total value set, the step of extracting an initial blade image from the initial blade image set in turn is returned until the initial blade image in the initial blade image set is extracted; if there is a distance detection value less than zero in the distance detection value set, an early warning threshold is obtained according to the distance detection value set, and the wind turbine is maintained according to the early warning threshold to obtain an optimized wind turbine; based on one or more replacement blades and the optimized wind turbine, an abnormal alarm of the wind turbine based on feature parameter identification is completed.

2. The abnormal alarm method for wind turbines based on characteristic parameter identification according to claim 1, characterized in that: The method of obtaining a pre-built gearbox temperature data set of the wind turbine according to the wind turbine acquisition parameter instruction and obtaining temperature characteristics according to the gearbox temperature data set includes: obtaining a detection flight path, using a pre-built drone infrared temperature measurement, a detection flight path, a preset detection time and a wind turbine acquisition parameter instruction to obtain a gearbox temperature image set of the wind turbine; obtaining a temperature data set according to the gearbox temperature image set, and obtaining a temperature average value set according to the temperature data set; and confirming the temperature characteristics based on the temperature average value set.

3. The abnormal alarm method for wind turbines based on characteristic parameter identification according to claim 2, characterized in that: The method of obtaining a temperature data set according to a gearbox temperature image set includes: extracting one gearbox temperature image from the gearbox temperature image set in turn, and performing the following operations on the extracted gearbox temperature images: performing a denoising operation on the gearbox temperature image to obtain a denoised temperature image, performing a grayscale operation on the denoised temperature image to obtain a grayscale value set; performing the following operations on each grayscale value in the grayscale value set: obtaining a temperature value using a grayscale value and a pre-constructed temperature-grayscale mapping table, wherein one temperature value corresponds to a timestamp and a position; summarizing the temperature values ​​to obtain temperature data; and summarizing the temperature data to obtain a temperature data set.

4. The abnormal alarm method for wind turbines based on characteristic parameter identification according to claim 3 is characterized in that: The method of obtaining multiple anomaly detection SVDD models includes: obtaining multiple historical detection time periods; extracting one historical detection time period from the multiple historical detection time periods in turn, and performing the following operations on the extracted historical detection time periods: using the historical detection time period to detect pre-constructed characteristic parameters of the wind turbine to obtain a characteristic data set, wherein the characteristic parameters include: temperature characteristics, power characteristics and vibration characteristics; standardizing the characteristic data set to obtain a standard characteristic data set; summarizing the standard characteristic data set to obtain multiple standard characteristic data sets; using the multiple standard characteristic data sets to train the pre-constructed SVDD model to obtain multiple initial SVDD models, wherein the initial SVDD model corresponds to the standard characteristic data set one by one; using pre-constructed test data to test the accuracy set and recall rate set of the multiple initial SVDD models, wherein one initial SVDD model corresponds to one accuracy rate and one recall rate; if it is respectively confirmed that the accuracy rate set and the recall rate set are both in the corresponding preset intervals, the multiple initial SVDD models are confirmed as multiple anomaly detection SVDD models.

5. The abnormal alarm method for wind turbines based on characteristic parameter identification according to claim 4, characterized in that: The method of testing the accuracy set and the recall set of the multiple initial SVDD models using the pre-constructed test data includes: calculating the accuracy set and the recall set of the multiple initial SVDD models using the test data, the pre-constructed accuracy formula and the pre-constructed recall formula, wherein the test data includes multiple unit test data; wherein the accuracy formula is as follows: ,in, Indicates accuracy, Represents the number of unit test data that are actually abnormal and predicted to be abnormal by the initial SVDD model, It represents the number of unit test data that are actually normal and predicted to be normal by the initial SVDD model. Indicates the number of unit test data that are actually normal but predicted as abnormal by the initial SVDD model It indicates the number of unit test data that are actually abnormal but predicted to be normal by the initial SVDD model. represents the recall rate; the recall rate formula is as follows: ,in, Represents the recall rate.

6. The abnormal alarm method for wind turbines based on characteristic parameter identification according to claim 5, characterized in that: The method uses multiple anomaly detection SVDD models to perform anomaly detection on temperature features, power features, and vibration features to obtain a distance detection value set, including: obtaining multiple optimal hyperspheres according to the multiple anomaly detection SVDD models, wherein the anomaly detection SVDD models correspond to the optimal hyperspheres one by one, and the optimal hyperspheres include: a center and a radius; extracting a target feature value from the temperature feature, the power feature, and the vibration feature respectively, and performing the following operations on the extracted target feature values: calculating the distance detection value between the target feature value and the center of the circle using the optimal hypersphere corresponding to the target feature value in the multiple optimal hyperspheres, wherein the calculation formula of the distance detection value is as follows: ,in, Indicates the distance detection value, represents the target feature value, represents the center of the optimal hypersphere, represents the radius of the optimal hypersphere, Represents the transposed symbol; summarizes the distance detection values ​​to obtain a distance detection value set.

7. The abnormal alarm method for wind turbines based on characteristic parameter identification according to claim 6, characterized in that: The method for calculating the total leaf similarity value based on leaf image features and standard leaf image features includes: obtaining a standard leaf feature point set based on the standard leaf image features; extracting a standard leaf feature point from the standard leaf feature point set in turn, and performing the following operations on the extracted standard leaf feature points: obtaining a leaf edge image using a pre-constructed edge detection algorithm and a sub-region map, and obtaining a leaf feature point set based on the leaf edge image, wherein the leaf feature point set includes multiple leaf feature points, and each leaf feature point includes: a feature point position, a feature point direction, a feature point scale, and a feature point descriptor; determining feature points at the same position in the leaf feature point set based on the standard leaf feature points, and calculating the feature point similarity between the feature points at the same position and the standard leaf feature points; summarizing the feature point similarities to obtain a feature point similarity set, and obtaining the total leaf similarity value based on the feature point similarity set.

8. The abnormal alarm method for wind turbines based on characteristic parameter identification according to claim 7, characterized in that: The calculating of the feature point similarity between the feature point at the same position and the feature point at the standard leaf comprises: using a pre-built similarity formula to calculate the feature point similarity between the feature point at the standard leaf and the feature point at the same position, wherein the similarity formula is as follows: , , , , ,in, Represents the similarity of feature points, represents the coordinates of the standard blade feature points, Represents the coordinates of feature points at the same position, represents the normalization factor, represents the location similarity, represents the direction similarity, represents the orientation angle of the standard blade feature point, Represents the direction angle of feature points at the same position, represents the scale similarity, represents the scale of the standard leaf feature points, Represents the scale of feature points at the same position, Descriptor vector representing the standard leaf feature points, Descriptor vector representing feature points at the same position, Represents the descriptor vector.

9. The abnormal alarm method for wind turbines based on characteristic parameter identification according to claim 8, characterized in that: The step of obtaining the warning threshold according to the distance detection value set includes: obtaining the number of sliding windows, obtaining the standard deviation of the distance detection value set; and calculating the warning threshold using the number of sliding windows, the standard deviation, the distance detection value set, and a pre-built dynamic sliding window warning threshold formula, wherein the dynamic sliding window warning threshold formula is as follows: ,in, Indicates the warning threshold, Indicates the standard characteristic parameter value of the wind turbine. represents the standard deviation of the distance detection value set, Represents the total number of distance detection value sets, represents the number of sliding windows, Indicates the sliding window movement increment.

10. A wind turbine abnormality alarm system based on characteristic parameter identification, characterized in that: The system includes: a feature extraction module, which is used to receive a wind turbine parameter acquisition instruction, obtain a pre-constructed gearbox temperature data set of the wind turbine according to the wind turbine parameter acquisition instruction, obtain temperature characteristics according to the gearbox temperature data set, obtain a power generation set of the wind turbine, obtain power characteristics according to the power generation set, obtain a vibration frequency set of the wind turbine, and obtain vibration characteristics according to the vibration frequency set; an anomaly detection module, which is used to obtain multiple anomaly detection SVDD models, use multiple anomaly detection SVDD models to perform anomaly detection on temperature characteristics, power characteristics and vibration characteristics to obtain a distance detection value set; a blade image analysis module, which is used to obtain an initial blade image set of the wind turbine if there is no distance detection value less than zero in the distance detection value set, extract an initial blade image from the initial blade image set in turn, and perform the following operations on the extracted initial blade images: segment the initial blade image to obtain multiple sub-region maps, wherein each of the multiple sub-region maps has and only has one blade, extract a sub-region map from the multiple sub-region maps in turn, and perform the following operations on the extracted sub-region maps: obtain blade image characteristics according to the sub-region map, according to the blade The standard leaf image features are retrieved from a pre-constructed leaf image database, a total leaf similarity value is calculated according to the leaf image features and the standard leaf image features, the total leaf similarity values ​​are summarized to obtain a leaf similarity total value set, and it is determined whether there is a leaf similarity total value that is not in a preset similarity total value interval in the leaf similarity total value set; if there is a leaf similarity total value that is not in the preset similarity total value interval in the leaf similarity total value set, a high-level warning is issued using a pre-constructed alarm device, and the leaf corresponding to the leaf similarity total value is replaced to obtain one or more replacement leaves. If there is no total blade similarity value that is not in the preset total similarity value interval in the blade similarity total value set, then return to the step of extracting an initial blade image from the initial blade image set in sequence until the initial blade image in the initial blade image set is extracted; a wind turbine maintenance module is used to obtain a warning threshold according to the distance detection value set if there is a distance detection value less than zero in the distance detection value set, maintain the wind turbine according to the warning threshold, obtain an optimized wind turbine, and complete an abnormal alarm of the wind turbine based on characteristic parameter identification based on one or more replacement blades and optimized wind turbines.

Images