Fan blade anomaly detection method, device, equipment, medium and product

Abstract

The application provides a fan blade anomaly detection method, running data of a fan is acquired from a data acquisition and monitoring control system, then feature extraction is performed on the running data through a clustering algorithm to acquire feature data, then the feature data is input into a pre-trained anomaly detection model to obtain a result of whether vibration of the fan blade is abnormal, so that low-cost and high-reliability detection of fan blade anomalies is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fan, in particular to a fan blade anomaly detection method, device, equipment, medium and product. BACKGROUND

[0002] The wind turbine (also known as fan) is an electric power equipment that converts wind energy into mechanical work, and the mechanical work drives the rotor to rotate, and finally outputs alternating current. The fan blade is an important basic part of the fan. Under the driving of the wind, the fan blade can convert wind energy into mechanical energy. When the fan has a bulge, a crack or the like, it may cause the load imbalance of the blade, resulting in fan blade anomaly. The fan blade anomaly may cause the fan to work abnormally, affecting the safety of the fan.

[0003] Generally, the blade anomaly can be detected by collecting the audio signal of the blade. However, this method usually needs to install a specific collection device, and is easily affected by noise and other factors, and the processing difficulty of the collected data is high.

[0004] Therefore, there is an urgent need in the industry for a low-cost fan blade anomaly detection method. SUMMARY

[0005] The present application provides a fan blade anomaly detection method, which can effectively reduce the cost of fan blade anomaly detection and provide a low-cost and reliable fan blade anomaly detection method. The present application also provides a device, equipment and computer readable storage medium corresponding to the above method.

[0006] In a first aspect, the present application provides a fan blade anomaly detection method, which comprises:

[0007] obtaining running data of the fan from a data acquisition and monitoring control system;

[0008] extracting features of the running data by a clustering algorithm to obtain feature data;

[0009] inputting the feature data into a trained anomaly detection model to obtain a detection result of whether the fan blade vibration is abnormal, wherein the anomaly detection model is trained according to the training feature data.

[0010] In some possible implementation manners, the running data comprises engine speed and cabin acceleration.

[0011] In some possible implementation manners, before the feature data is extracted from the running data by the clustering algorithm, the method further comprises:

[0012] cleaning the running data according to the engine speed to obtain cabin acceleration in a plurality of fixed speed intervals.

[0013] In some possible implementation manners, feature extraction is performed on the running data by a clustering algorithm to obtain feature data, including:

[0014] The cabin accelerations in the preset time period are divided into n acceleration bins according to time windows;

[0015] The effective acceleration bins are determined from the n acceleration bins according to a fixed rotation speed interval, and adjacent effective acceleration bins are merged to obtain m target data texts corresponding to the cabin accelerations;

[0016] The frequencies of the m target data texts in each cabin acceleration category are respectively counted;

[0017] A target vector is formed according to the frequencies, and the target vector is used to represent the frequency distribution of each data text;

[0018] The feature data is obtained according to the target vector.

[0019] In some possible implementation manners, the feature data is obtained according to the target vector, including:

[0020] Each target vector is processed by a clustering algorithm to obtain p clusters;

[0021] The cluster centers of each cluster are determined;

[0022] The frequency distribution of each target vector in the p clusters is counted;

[0023] A feature vector formed by the frequency distribution is determined as the feature data.

[0024] In some possible implementation manners, the training step of the anomaly detection model includes:

[0025] Training feature data of the fan is obtained;

[0026] The training feature data is added with a first label and a second label, wherein the first label is used to indicate that the corresponding training feature data is running data of a normal running state of the fan, and the second label is used to indicate that the corresponding training feature data is running data in an abnormal running state of the fan;

[0027] The anomaly detection model is trained according to the training feature data added with the labels to obtain the trained anomaly detection model.

[0028] In some possible implementation manners, the anomaly detection model is a binary classification model.

[0029] In a second aspect, the application provides a fan blade anomaly detection device, which includes:

[0030] The communication module is configured to acquire operation data of the wind turbine from a data acquisition and monitoring control system.

[0031] The extraction module is configured to extract features from the operation data by using a clustering algorithm to obtain feature data.

[0032] The detection module is configured to input the feature data into a trained anomaly detection model to obtain a detection result of whether the vibration of the wind turbine blade is abnormal, wherein the anomaly detection model is trained based on the training feature data.

[0033] In some possible implementations, the operation data includes engine speed and nacelle acceleration.

[0034] In some possible implementations, the device further includes a cleaning module configured to:

[0035] clean the operation data according to the engine speed to obtain nacelle acceleration in a plurality of fixed speed intervals.

[0036] In some possible implementations, the extraction module can be configured to:

[0037] divide the nacelle acceleration in a preset time period into n acceleration bins according to a time window;

[0038] determine effective acceleration bins from the n acceleration bins according to the fixed speed intervals, and merge adjacent effective acceleration bins to obtain m target data texts corresponding to the nacelle acceleration;

[0039] respectively count frequencies of the m target data texts in each nacelle acceleration category;

[0040] form a target vector according to the frequencies, wherein the target vector is configured to represent frequency distribution of each data text;

[0041] obtain the feature data according to the target vector.

[0042] In some possible implementations, the extraction module can be configured to:

[0043] perform clustering processing on each target vector by using a clustering algorithm to obtain p clusters;

[0044] determine a clustering center of each cluster;

[0045] count frequency distribution of each target vector in the p clusters;

[0046] determine a feature vector formed by the frequency distribution as the feature data.

[0047] In some possible implementations, the training step of the anomaly detection model includes:

[0048] Obtain training feature data for the wind turbine;

[0049] A first label and a second label are added to the training feature data. The first label is used to indicate that the corresponding training feature data is the operating data of the wind turbine in normal operating condition, and the second label is used to indicate that the corresponding training feature data is the operating data of the wind turbine in abnormal operating condition.

[0050] The anomaly detection model is trained based on the training feature data after adding labels, and the trained anomaly detection model is obtained.

[0051] In some possible implementations, the anomaly detection model is a binary classification model.

[0052] Thirdly, this application provides an apparatus comprising a processor and a memory. The processor and the memory communicate with each other. The processor is configured to execute instructions stored in the memory to cause the apparatus to perform a wind turbine blade anomaly detection method as described in the first aspect or any implementation thereof.

[0053] Fourthly, this application provides a computer-readable storage medium storing instructions that instruct a device to perform the wind turbine blade anomaly detection method described in the first aspect or any implementation thereof.

[0054] Fifthly, this application provides a computer program product containing instructions that, when run on a device, causes the device to execute the wind turbine blade anomaly detection method described in the first aspect or any implementation thereof.

[0055] Based on the implementation methods provided in the above aspects, this application can be further combined to provide more implementation methods.

[0056] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0057] This application provides a method for detecting anomalies in wind turbine blades. The method acquires wind turbine operating data from a data acquisition and monitoring control system, then extracts features from the operating data using a clustering algorithm to obtain feature data. This feature data is then input into a pre-trained anomaly detection model to determine whether the wind turbine blade vibration is abnormal. Since wind turbine blade vibration data cannot be directly tested, and adding other equipment for data acquisition is costly, this method utilizes the wind turbine's inherent data acquisition and monitoring control system to acquire operating data. By extracting features from the operating data using a clustering algorithm, feature data representing the wind turbine blade vibration can be obtained. This allows for the use of an anomaly detection model to determine whether the wind turbine blade vibration is abnormal, resulting in a low-cost and highly reliable method. Attached Figure Description

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0059] Figure 1 A flow chart of a fan blade anomaly detection method provided by an embodiment of the present application;

[0060] Figure 2 A flow chart of data cleaning provided by an embodiment of the present application;

[0061] Figure 3 A flow chart of extracting a feature vector provided by an embodiment of the present application;

[0062] Figure 4 A flow chart of another fan blade anomaly detection method provided by an embodiment of the present application;

[0063] Figure 5 A schematic diagram of a fan blade anomaly detection device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0064] The schemes in the embodiments provided by the present application will be described below in combination with the drawings in the present application.

[0065] The terms "first", "second", etc. in the specification and claims of the present application and the above drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the terms used in this way can be interchanged as appropriate, which is only a distinguishing way used in the description of the embodiments of the present application to describe the objects with the same attributes.

[0066] In order to facilitate understanding of the technical solutions of the present application, some technical terms related to the present application will be introduced below.

[0067] The faults of the fan blade mainly include breakage, crack, cracking, etc. When the fan blade fails, it may cause the load imbalance of the blade, cause the fan blade anomaly, and further may cause the fan to work abnormally, affecting the safety of the fan work. Under normal circumstances, the running condition of the fan blade needs to be monitored, and the anomaly of the fan blade can be monitored as early as possible to avoid causing greater loss.

[0068] Typically, abnormalities in wind turbine blades are detected using methods such as audio signals and video images. However, these monitoring methods usually require additional equipment, such as sensors and video acquisition devices, increasing the investment cost of the wind turbine. Furthermore, audio signals and video images are often susceptible to interference from external factors. For example, audio signals are easily affected by noise, increasing the difficulty of data processing. Video images are easily affected by ambient brightness; poor lighting on cloudy or rainy days makes video image processing more challenging.

[0069] In view of this, this application provides a low-cost method for detecting wind turbine blade anomalies, enabling low-cost and high-reliability detection of wind turbine blade anomalies. This method is executed by a processing device, which is a device with data processing capabilities, and can be a terminal or a server. Specifically, the processing device can acquire wind turbine operating data from a data acquisition and monitoring control system, then extract features from the operating data using a clustering algorithm to obtain feature data. This feature data is then input into an anomaly detection model trained based on the feature data to obtain a detection result indicating whether the wind turbine blade vibration is abnormal, thus achieving the detection of wind turbine blade anomalies.

[0070] On the one hand, this method acquires data based on the wind turbine's own data acquisition and monitoring control system, avoiding the increased costs associated with adding equipment. Furthermore, the wind turbine's operating data acquired through the data acquisition and monitoring control system is less susceptible to environmental interference, and the data processing is relatively stable.

[0071] On the other hand, blade abnormalities can be reflected in blade vibration relatively quickly, so judging whether the blades are abnormal by observing the vibration of the wind turbine blades has a high accuracy.

[0072] Next, the wind turbine blade anomaly detection method provided in the embodiments of this application will be described in conjunction with the accompanying drawings.

[0073] See Figure 1 The flowchart shown illustrates a method for detecting abnormal wind turbine blades, which includes the following steps:

[0074] S102: The processing equipment obtains the operating data of the fan from the data acquisition and monitoring control system.

[0075] Supervisory control and data acquisition (SCADA) systems are computer-based distributed control systems (DCS) and power automation monitoring systems. SCADA systems can perform various functions such as data acquisition, equipment control, measurement, and parameter adjustment. Wind turbine generator sets typically include SCADA. In this embodiment, the processing equipment can obtain the wind turbine's operating data from the SCADA system built into the turbine itself, without the need for additional equipment, thus reducing the cost of detecting abnormal wind turbine blades.

[0076] The wind turbine's operating data can include the turbine engine's rotational speed and the nacelle's acceleration. In some possible implementations, the nacelle's acceleration can be decomposed into the x-axis acceleration and y-axis acceleration in the body coordinate system. The body coordinate system refers to the coordinate system fixed on the aircraft; in this embodiment, it is the wind turbine coordinate system. The origin of this coordinate system is located at the wind turbine's center of mass, the positive x-axis points towards the wind turbine's nose, the positive y-axis is perpendicular to the wind turbine's central axis and points to the right side of the wind turbine, and the positive z-axis is determined according to the right-hand rule. The inertial measurement unit (IMU) in SCADA can directly acquire the nacelle acceleration in the body coordinate system.

[0077] S104: The processing equipment cleans the operating data.

[0078] The processing equipment can clean the operating data based on the acquired engine speed to obtain nacelle acceleration within multiple fixed speed ranges. Specifically, the processing equipment can discard nacelle accelerations at lower engine speeds and those at higher engine speeds. When the engine speed is low, the unit operation is unstable, and the nacelle acceleration is significantly affected. When the engine speed is high, the unit is more affected by the electrical system, which also leads to unstable nacelle acceleration. Therefore, the processing equipment can clean the operating data based on engine speed to obtain more stable and less disturbed nacelle accelerations within a fixed engine speed range.

[0079] Specifically, such as Figure 2As shown, the processing equipment reads data from a single generator set over a certain period of time, including engine speed, nacelle y-axis acceleration, and nacelle x-axis acceleration. The processing equipment deletes control rows from the data. When the number of data entries exceeds a preset value, such as p1 (in this embodiment, pi represents the preset threshold), the acceleration data is filtered. The filtering criteria can be: nacelle y-axis acceleration greater than or equal to p2 and less than or equal to p3; engine speed greater than or equal to p4 and less than or equal to p5; and nacelle x-axis acceleration greater than or equal to p6 and less than or equal to p7. When the number of data entries exceeds p8, data meeting the criteria within a specific engine speed range (greater than or equal to p4 and less than or equal to p5) is obtained.

[0080] S106: The processing equipment extracts features from the running data using a clustering algorithm to obtain feature data.

[0081] In some possible implementations, the operating data can be the engine speed within a fixed speed range, and the processing equipment can discretize and quantize it according to a certain time window interval.

[0082] Clustering algorithms divide a dataset into different clusters according to specific criteria, aiming to maximize the similarity of data objects within the same cluster and maximize the differences between data objects in different clusters.

[0083] Clustering algorithms can be categorized into several types, including hierarchical methods, which can be further divided into merge-based hierarchical clustering and split-based hierarchical clustering. Merge-based hierarchical clustering starts from the bottom layer, merging the most similar clusters to form clusters in the next higher level. The process stops when all data points are merged into one cluster or when a certain termination condition is met. Split-based hierarchical clustering uses a top-down approach, starting with a cluster containing all data points. The root node is then split into sub-clusters, and each sub-cluster recursively continues splitting until a single-node cluster containing only one data point appears. This type of clustering algorithm has higher computational complexity and is suitable for small orders of magnitude.

[0084] Partition-based methods cluster a set of scattered data points, aiming to ensure that points within each cluster are sufficiently close and points between clusters are sufficiently far apart. Specifically, this involves first determining how many clusters the data points will ultimately form, then selecting a few points as initial centroids, and finally iteratively relocating the data points using pre-defined heuristic algorithms until the desired effect of "points within each cluster being sufficiently close and points between clusters being sufficiently far apart" is achieved. Examples include the k-means algorithm and its variants such as k-medoids, k-modes, k-medians, and kernel k-means.

[0085] Density-based methods cluster regions based on their density. Clustering continues when the density (number of objects or data points) of neighboring regions exceeds a certain threshold. An example is DBSCAN (Density-Based Spatial Clustering of Applications with Noise).

[0086] Grid-based methods divide the data space into grid cells, map the data object set to the grid cells, calculate the density of each cell, and determine whether each grid cell is a high-density cell based on a preset threshold. Neighboring dense cells are grouped to form a "class".

[0087] Model-based methods assume a model for each cluster and seek the best fit of the data to a given model.

[0088] Fuzzy clustering (FCM fuzzy clustering) is an algorithm that uses fuzzy set theory to perform cluster analysis and determines the degree to which each data point belongs to a certain cluster based on its membership degree.

[0089] Specifically, the processing device can extract features from the operational data using the BM25 algorithm to obtain feature data. The BM25 algorithm can be used for search relevance scoring. In this embodiment, the processing device divides the nacelle acceleration over a preset time period into n acceleration modules according to time windows, for example, with a 5-minute interval. Then, the processing device determines the valid acceleration modules from the n acceleration modules based on a fixed speed range. The validity of an acceleration module can be determined based on whether the engine speed is within a fixed range. For example, if the amount of data within the preset range of engine speed is greater than 80% and greater than 40 records within the time window, then the acceleration module within that time window is considered a valid acceleration module. The processing device sequentially determines the validity of the acceleration modules in each time window. If the data in adjacent windows are all valid, they can be merged to form m target data texts corresponding to the nacelle acceleration. In this embodiment, since the BM25 algorithm has a weight for text length when scoring search relevance, when the length of the merged data is greater than a certain value (e.g., 5), the frequency of m target data texts in each cabin acceleration category is counted to construct a TF-IDF vector.

[0090] In the BM25 algorithm, the TF-IDF score represents the importance of the target data text in each acceleration category. The TF-IDF score consists of two parts: TF represents the standard frequency of the target data text, and IDF represents the frequency of the inverse document.

[0091] In some possible implementations, the nacelle acceleration range can be from -0.3 to 0.3, where "-" indicates direction. The processing device divides the range into 0.01 increments, obtaining -0.3 to -0.29 as the first category, -0.29 to -0.28 as the second category, and so on, up to 0.29 to 0.3 as the sixty-first category. The processing device counts the frequency of each data text in each category, obtaining the frequency distribution corresponding to each data text. In this embodiment, the processing device obtains a 61-dimensional TF-IDF vector for each data text. Each wind turbine, after data processing, generates at least one TF-IDF vector.

[0092] The processing equipment can cluster the target TF-IDF vector using a clustering algorithm (e.g., k-means) to obtain p clusters. The k-means algorithm is a partitioning method, thus determining the cluster centers of each cluster and statistically analyzing the frequency distribution of the TF-IDF vector for each unit across all clusters. The feature vector formed by this frequency distribution is then used as the feature data. In some possible implementations, when there are 5 clusters, a unit can have 5 TF-IDF vectors, with their frequency distributions in the 5 clusters being 0.1, 0.2, 0.3, 0.3, and 0.1, respectively, thus forming the feature data for that unit.

[0093] In this embodiment, a feature extraction method for blade vibration data based on the BM25 algorithm and K-means clustering algorithm can be used, such as... Figure 3 As shown, cleaned data is imported into the processing device, the TFTIDF vector is calculated using the BM25 principle, and then the k-means algorithm is applied to calculate the class distribution of the TFTIDF vector, forming a feature vector. In this way, discrete data of different lengths within a specific speed range can be transformed into effective feature vectors of the same length, facilitating subsequent data processing.

[0094] S108: The processing device inputs the feature data into the trained anomaly detection model to obtain the detection result of whether the wind turbine blade vibration is abnormal.

[0095] An anomaly detection model is a machine learning model that can determine whether the vibration of wind turbine blades is abnormal based on feature data. The anomaly detection model can be trained using training feature data. Specifically, the processing device can add labels to the training feature data, which can be obtained through steps S102 to S106. When acquiring training data, the corresponding wind turbine blade vibration state is also acquired. Thus, a first label and a second label can be added to the training feature data. The first label corresponds to the training feature data when the wind turbine is operating normally, and the second label corresponds to the training feature data when the wind turbine is operating abnormally.

[0096] The machine learning model can be a logistic regression (LR) model, which is a binary classification model. Since the training feature data contains far more data representing normal operating conditions than abnormal operating conditions, an imbalanced data distribution may occur during machine learning model training. Logistic regression is effective in addressing this imbalanced data distribution problem. During training, when the wind turbine operating state corresponding to the training feature data is abnormal, the training label is marked as "1"; otherwise, it is marked as "0". When applying the model for anomaly detection, the model outputs a value between 0 and 1. In some possible implementations, the anomaly detection model can also output the probability of the feature data being 0 and the probability of it being 1, thus performing anomaly detection.

[0097] In summary, the embodiments of this application provide a method for detecting abnormalities in wind turbine blades. This method can collect real-time operating data from wind turbine generator sets and can be applied to megawatt-level wind turbine generator sets. Specifically, as... Figure 4 As shown, the processing equipment can obtain the wind turbine's operating data from the wind turbine's data acquisition and monitoring control system without the need to install additional acquisition equipment. Then, it uses a clustering algorithm to extract features from the operating data, obtains feature data, and inputs the feature data into the trained anomaly detection model to obtain the detection result of whether the wind turbine blade vibration is abnormal.

[0098] Furthermore, the processing equipment can also effectively clean the data, improve the quality of feature extraction, and reduce the impact of factors such as noise.

[0099] The above combination Figure 1 The method for detecting abnormal wind turbine blades provided in the embodiments of this application has been described in detail. Next, the device for detecting abnormal wind turbine blades provided in the embodiments of this application will be described in conjunction with the accompanying drawings.

[0100] See Figure 5 The diagram shows the structure of a wind turbine blade anomaly detection device. The device 400 includes: a communication module 502, an extraction module 504, and a detection module 506.

[0101] Secondly, this application provides a wind turbine blade abnormality detection device, the device comprising:

[0102] The communication module 502 is used to acquire the wind turbine's operating data from the data acquisition and monitoring control system;

[0103] Extraction module 504 is used to extract features from running data using a clustering algorithm to obtain feature data;

[0104] The detection module 506 is used to input feature data into the trained anomaly detection model to obtain the detection result of whether the wind turbine blade vibration is abnormal. The anomaly detection model is trained based on the training feature data.

[0105] In some possible implementations, the operating data includes engine speed and nacelle acceleration.

[0106] In some possible implementations, the device also includes a cleaning module for:

[0107] The operating data is cleaned based on the engine speed to obtain the nacelle acceleration in multiple fixed speed ranges.

[0108] In some possible implementations, the extraction module can be used for:

[0109] The cabin acceleration within a preset time period is divided into n acceleration compartments according to time windows.

[0110] Based on a fixed speed range, the effective acceleration chambers are determined from n acceleration chambers, and adjacent effective acceleration chambers are merged to obtain m target data texts corresponding to the nacelle acceleration;

[0111] Calculate the frequency of each of the m target data texts in each cabin acceleration category;

[0112] A target vector is formed based on the frequency, and the target vector is used to characterize the frequency distribution of each data text.

[0113] Feature data is obtained from the target vector.

[0114] In some possible implementations, the extraction module can be used for:

[0115] Clustering algorithms are used to cluster each target vector to obtain p clusters;

[0116] Determine the cluster centers of each cluster;

[0117] Calculate the frequency distribution of each target vector in p clusters;

[0118] The eigenvectors formed by the frequency distribution are determined as the feature data.

[0119] In some possible implementations, the training steps for the anomaly detection model include:

[0120] Obtain training feature data for the wind turbine;

[0121] A first label and a second label are added to the training feature data. The first label is used to indicate that the corresponding training feature data is the operating data of the wind turbine in normal operating condition, and the second label is used to indicate that the corresponding training feature data is the operating data of the wind turbine in abnormal operating condition.

[0122] The anomaly detection model is trained based on the training feature data after adding labels, and the trained anomaly detection model is obtained.

[0123] In some possible implementations, the anomaly detection model is a binary classification model.

[0124] The wind turbine blade anomaly detection device 500 according to the embodiments of this application can correspondingly execute the method described in the embodiments of this application, and the above and other operations and / or functions of each module of the wind turbine blade anomaly detection device 500 are respectively for implementing Figure 1 For the sake of brevity, the corresponding processes of each method in the code will not be elaborated here.

[0125] This application provides an apparatus for implementing a method for detecting abnormal wind turbine blades. The apparatus includes a processor and a memory. The processor and the memory communicate with each other. The processor executes instructions stored in the memory to cause the apparatus to perform the method for detecting abnormal wind turbine blades.

[0126] This application provides a computer-readable storage medium storing instructions that, when executed on a device, cause the device to perform the aforementioned wind turbine blade anomaly detection method.

[0127] This application provides a computer program product containing instructions that, when run on a device, causes the device to perform the aforementioned wind turbine blade anomaly detection method.

[0128] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0129] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0130] In the above embodiments, the implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, in the form of a computer program product.

[0131] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A method for detecting abnormalities in wind turbine blades, characterized in that, The method includes: The operating data of the wind turbine is acquired from the data acquisition and monitoring control system, and the operating data includes engine speed and nacelle acceleration. The cabin acceleration within a preset time period is divided into n acceleration compartments according to time windows. Based on a fixed rotational speed range, the effective acceleration chambers are determined from the n acceleration chambers, and adjacent effective acceleration chambers are merged to obtain m target data texts corresponding to the nacelle acceleration; Calculate the frequency of the m target data texts in each cabin acceleration category; A target vector is formed based on the frequency, and the target vector is used to characterize the frequency distribution of each data text. Clustering algorithms are used to cluster each target vector to obtain p clusters; Determine the cluster center of each of the p cluster classes; Calculate the frequency distribution of each target vector in the p clusters; The eigenvectors formed by the frequency distribution are determined as feature data; The feature data is input into the trained anomaly detection model to obtain the detection result of whether the wind turbine blade vibration is abnormal. The anomaly detection model is trained based on the training feature data.

2. The method according to claim 1, characterized in that, Before dividing the cabin acceleration for the preset time period into cabins according to time windows, the method further includes: The operating data is cleaned based on the engine speed to obtain the nacelle acceleration in multiple fixed speed ranges.

3. The method according to any one of claims 1 to 2, characterized in that, The training steps of the anomaly detection model include: Obtain the training feature data of the wind turbine; A first label and a second label are added to the training feature data, wherein the first label is used to indicate that the corresponding training feature data is the operating data of the wind turbine in normal operating state, and the second label is used to indicate that the corresponding training feature data is the operating data of the wind turbine in abnormal operating state. The anomaly detection model is trained based on the training feature data with added labels, and the trained anomaly detection model is obtained.

4. The method according to claim 1, characterized in that, The anomaly detection model is a binary classification model.

5. A wind turbine blade anomaly detection device, characterized in that, The device includes: A communication module is used to acquire the operating data of the wind turbine from the data acquisition and monitoring control system, the operating data including engine speed and nacelle acceleration; An extraction module is used to divide the cabin acceleration within a preset time period into n acceleration compartments according to a time window; determine the effective acceleration compartments from the n acceleration compartments based on a fixed speed range and merge adjacent effective acceleration compartments to obtain m target data texts corresponding to the cabin acceleration; count the frequency of the m target data texts in each cabin acceleration category; form target vectors based on the frequencies, which are used to characterize the frequency distribution of each data text; perform clustering processing on each target vector using a clustering algorithm to obtain p clusters; determine the cluster centers of each cluster in the p clusters; count the frequency distribution of each target vector in the p clusters; and determine the feature vectors formed by the frequency distribution as feature data. The detection module is used to input the feature data into the trained anomaly detection model to obtain the detection result of whether the vibration of the wind turbine blade is abnormal. The anomaly detection model is trained based on the training feature data.

6. A wind turbine blade anomaly detection device, characterized in that, The device includes a processor and a memory; The processor is configured to execute instructions stored in the memory to cause the device to perform the method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, Includes instructions that instruct the device to perform the method as described in any one of claims 1 to 4.

8. A computer program product, characterized in that, When the computer program product is run on a computer, it causes the computer to perform the method as described in any one of claims 1 to 4.

Images