A wind turbine blade anomaly detection method and system

Abstract

The application relates to a wind turbine blade anomaly detection method and system, which comprises the following steps: collecting vibration data in a wind turbine blade and operation parameters in a wind turbine as initial data sets; adopting a moving average filtering method and a linear interpolation algorithm to pre-process data in the initial data sets to obtain target data sets; adopting a multi-modal feature extraction mode to extract features of data in the target data sets; adopting a three-layer mode of residual cleaning, time series smoothing and probability output to perform hierarchical processing on the data after feature extraction, and performing anomaly detection under abnormal working conditions; and introducing a double-factor confirmation mechanism and an abnormal confidence score system to judge abnormal conditions of the wind turbine blade. Through the pre-processing of data by adopting the moving average filtering method and the linear interpolation algorithm, the feature extraction of data and the hierarchical processing of data, the wind turbine blade anomaly detection method and system can more accurately detect the blade anomaly.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for detecting abnormalities in wind turbine blades. Background Technology

[0002] The blades of wind turbines are the core energy capture and conversion components of large wind turbine generators. They are usually formed in one piece by processes such as vacuum injection and autoclave, using glass fiber, carbon fiber or hybrid composite materials. They adopt a high lift-to-drag ratio airfoil family, aerodynamic design with gradually changing twist angle and thickness distribution along the span, and are combined with root pitch bearings and blade tip spoilers to achieve optimal angle of attack control in different wind speed ranges, maximizing wind energy absorption and suppressing stall.

[0003] Currently, the methods for detecting abnormalities in wind turbine blades typically involve a combination of regular manual inspections and intelligent, online methods. Based on the analysis of data from the SCADA system, the blade condition is indirectly inferred from anomalies in parameters such as the monitored power curve and rotational speed. Maintenance is then carried out based on the condition of the blades.

[0004] However, existing methods for detecting abnormalities in wind turbine blades suffer from low data quality due to direct analysis of data from SCADA systems. This makes it difficult to distinguish between fatigue cracks, deformation, and environmental disturbances, leading to missed or false alarms. Consequently, it becomes challenging to accurately diagnose abnormalities in wind turbine blades. Summary of the Invention

[0005] In view of the above-mentioned problems existing in the prior art, the purpose of the present invention is to provide a method and system for detecting abnormalities in wind turbine blades.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for detecting abnormalities in wind turbine blades includes: S1, collect vibration data from the wind turbine blades and operating parameters from the wind turbine as the initial dataset; S2, the moving average filtering method and linear interpolation algorithm are used to preprocess the data in the initial dataset to obtain the target dataset; S3 uses multimodal feature extraction to extract features from the target dataset; S4 employs a three-layer process of residual cleaning, temporal smoothing, and probability output to perform hierarchical processing on the data after feature extraction, enabling anomaly detection under abnormal operating conditions. S5 introduces a two-factor confirmation mechanism and an anomaly confidence score system to determine the abnormality of wind turbine blades.

[0007] A further improvement of this invention is that, in S2, a moving average filtering method and a linear interpolation algorithm are used to preprocess the data in the initial dataset to obtain the target dataset, which includes... Step S21: Use the moving average filtering method to identify outliers in the vibration parameters collected in the initial dataset. Step S22: Use a linear interpolation algorithm to complete the historical operation data of the data acquisition and monitoring control system in the initial dataset.

[0008] A further improvement of this invention is that, in step S21, the acquisition of vibration parameters includes: symmetrically installing high-precision triaxial accelerometers in the key stress concentration area at the blade root; setting an ultra-high frequency sampling frequency; and capturing three-dimensional vibration signals in the radial (X), tangential (Y), and axial (Z) directions in real time. The vibration signals are... During the data collection process, a temperature threshold is set to exclude sensor data below -40℃. The formula for judging low-temperature environment data is as follows:

[0009] in, For real-time temperature, The filtered vibration signal This is vibration signal data under normal conditions. The temperature requirement is not met.

[0010] A further improvement of this invention is that, in step S22, the linear interpolation algorithm includes: acquiring a data sequence containing core operating parameters such as wind speed, generator speed, pitch angle, and active power in real time from the wind turbine control system, and then completing the data using a linear interpolation algorithm. The formula for the linear interpolation algorithm is as follows:

[0011] in, The target value is obtained after linear interpolation calculation. and Given the dependent variable values ​​of two known data points, For the target value of the independent variable that needs to be interpolated, and To and and The corresponding known values ​​of the independent variable.

[0012] A further improvement of this invention is that, in S3, the multimodal feature extraction method is used to extract features from the data in the target dataset, including... Step S31: An improved adaptive noise algorithm for complete set empirical mode decomposition is introduced to suppress mode aliasing during data processing; the formula for the adaptive noise algorithm is as follows:

[0013] in, After the first The output result of this time; Empirical mode decomposition; It is a function of time t; For the first Gaussian white noise in the next iteration; This is the noise amplitude coefficient; For the first Energy operator of the first-order eigenmode function; Step S32: Construct a low-complexity feature extraction model based on the historical operation data supplemented by the data acquisition and monitoring control system.

[0014] A further improvement of this invention is that, in S4, a three-layer process of residual cleaning, temporal smoothing, and probability output is used to perform hierarchical processing on the data after feature extraction, and anomaly detection under abnormal operating conditions includes: Step S41: The kernel density estimation method is used to model the residual distribution under normal conditions; Step S42: Use a hidden Markov model to capture the temporal correlation of the residual sequence to obtain the current blade state.

[0015] A further improvement of this invention is that, in S5, a two-factor confirmation mechanism and an anomaly confidence score system are introduced to determine the abnormal situation of the wind turbine blades. This includes: making a preliminary anomaly judgment based on the real-time residual and the dynamic threshold; if the absolute value of the real-time residual is greater than the dynamic threshold, an instantaneous anomaly flag is triggered; otherwise, it is not triggered. Then, by judging the probability of the time-series state and the threshold, a secondary anomaly state assessment is performed to obtain a secondary time-series anomaly flag. If both the instantaneous anomaly flag and the secondary time-series anomaly flag are satisfied, it is determined to be a state anomaly. At this time, the wind turbine blade anomaly alarm is triggered, and the anomaly confidence score is calculated to classify the anomaly state into three levels for judgment. Then, different maintenance schemes are adopted according to different anomaly situations.

[0016] A wind turbine blade anomaly detection system includes: The data acquisition unit collects vibration data from the wind turbine blades and operating parameters of the wind turbine as the initial dataset; The data preprocessing unit uses the moving average filtering method and the linear interpolation algorithm to preprocess the data in the initial dataset to obtain the target dataset; The feature extraction unit uses a multimodal feature extraction method to extract features from the data in the target dataset; The anomaly detection unit employs a three-layer process of residual cleaning, temporal smoothing, and probability output to perform hierarchical processing on the feature-extracted data, enabling anomaly detection under abnormal operating conditions. The anomaly detection unit introduces a two-factor confirmation mechanism and an anomaly confidence score system to determine the anomaly status of the wind turbine blades.

[0017] A network-side server includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute the wind turbine blade anomaly detection method.

[0018] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for detecting abnormal blades in wind turbine generators.

[0019] Compared with the prior art, the present invention has at least the following beneficial technical effects: This invention provides a method and system for detecting anomalous blades in wind turbines. It collects vibration data from the wind turbine blades and operating parameters of the wind turbine as an initial dataset. The initial dataset is preprocessed using a moving average filtering method and a linear interpolation algorithm to obtain a target dataset. Multimodal feature extraction is used to extract features from the target dataset. A three-layer process—residual cleaning, temporal smoothing, and probability output—is employed to perform hierarchical processing on the feature-extracted data, enabling anomaly detection under abnormal operating conditions. A two-factor confirmation mechanism and an anomaly confidence scoring system are introduced to determine the anomalous status of the wind turbine blades. By using a moving average filtering method and a linear interpolation algorithm for data preprocessing, feature extraction, and hierarchical data processing, a more accurate detection of blade anomalies is achieved. Attached Figure Description

[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a flowchart of a method for detecting abnormalities in the blades of a wind turbine in this invention.

[0022] Figure 2 This is a structural block diagram of a wind turbine blade anomaly detection system according to the present invention.

[0023] Figure 3 This is a schematic diagram of the network-side server provided by the present invention. Detailed Implementation

[0024] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0025] In the description of this invention, it should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0026] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0027] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0028] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0029] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0030] Example 1 This invention provides a method for detecting anomalous blades in wind turbines. It collects vibration data from the wind turbine blades and operating parameters of the wind turbine as an initial dataset. The initial dataset is preprocessed using a moving average filtering method and a linear interpolation algorithm to obtain a target dataset. Multimodal feature extraction is then used to extract features from the target dataset. A three-layer process—residual cleaning, temporal smoothing, and probability output—is employed to perform hierarchical processing on the feature-extracted data, enabling anomaly detection under abnormal operating conditions. A two-factor confirmation mechanism and an anomaly confidence scoring system are introduced to determine the anomalous status of the wind turbine blades. By using a moving average filtering method and a linear interpolation algorithm for data preprocessing, feature extraction, and hierarchical data processing, a more accurate detection of blade anomalies is achieved.

[0031] The following is a detailed description of the implementation details of the wind turbine blade anomaly detection method of the present invention. The following content is only for ease of understanding and is not necessary for implementing this solution. like Figure 1 S1, collecting vibration data from the wind turbine blades and operating parameters from the wind turbine as the initial dataset.

[0032] Specifically: Data is collected from wind turbine blades to ensure the validity and accuracy of the collected data. A multi-source heterogeneous data fusion acquisition strategy is adopted: strain gauge sensors installed on the blade surface are used to collect structural stress data in real time to capture stress abrupt changes caused by fatigue cracks or abnormal deformation; laser displacement sensors are used to monitor the blade geometry and vibration modes with sub-millimeter precision to detect minute shape deviations; meteorological stations are deployed to synchronously acquire environmental parameters such as wind speed, wind direction, and temperature to eliminate interference from external factors; and the data acquisition and monitoring control system is connected to obtain historical operating data of the data acquisition and monitoring control system, using vibration data and historical operating data as the initial dataset.

[0033] S2, the moving average filtering method and linear interpolation algorithm are used to preprocess the data in the initial dataset to obtain the target dataset; Step S21: Use the moving average filtering method to identify outliers in the vibration parameters collected in the initial dataset.

[0034] Specifically, vibration parameter acquisition involves symmetrically installing high-precision triaxial accelerometers in the critical stress concentration area at the blade root, setting an ultra-high frequency sampling frequency, and capturing three-dimensional vibration signals in the radial (X), tangential (Y), and axial (Z) directions in real time. The vibration signals are... During the data acquisition process, a temperature threshold is set to exclude sensor data below -40℃. Low temperatures can cause sensor performance to degrade, leading to inaccurate data. The formula for judging low-temperature data is as follows:

[0035] in, For real-time temperature, The filtered vibration signal This is vibration signal data under normal conditions. The temperature requirement is not met.

[0036] The temperature-filtered data is then smoothed using a moving average filtering algorithm to suppress the influence of random noise and make the vibration data curve smoother. The moving average filtering algorithm formula is as follows:

[0037] in, To smooth the window size, The index value of the current time or data sequence. Used to determine the boundaries of the smoothing window. The value of the original data at time i after filtering is given. The average value of the filtered data.

[0038] Step S22: Use a linear interpolation algorithm to complete the historical operating data of the centralized data acquisition and monitoring control (SCADA) system in the initial dataset.

[0039] Specifically, the historical operating data of the data acquisition and monitoring control system is a data sequence acquired in real time from the wind turbine control system, containing core operating parameters such as wind speed, generator speed, pitch angle, and active power. However, data loss can occur during data transmission, leading to errors in data integrity and accuracy. A linear interpolation algorithm is used to complete the data. The linear interpolation algorithm formula is as follows:

[0040] in, The target value is obtained after linear interpolation calculation. and Given the dependent variable values ​​of two known data points, For the target value of the independent variable that needs to be interpolated, and To and and The corresponding known values ​​of the independent variable; Based on the historical operating data supplemented by the data acquisition and monitoring control system and the processed vibration data, a multi-dimensional data association model is constructed to obtain the target dataset.

[0041] S3 uses multimodal feature extraction to extract features from the target dataset; Specifically: In the anomaly detection of wind turbine blades, multimodal feature extraction is carried out by fusing data from different types of sensors to construct a comprehensive characterization of blade state information; vibration sensors capture high-frequency vibration signals during blade operation, which are converted into time-frequency maps through short-time Fourier transform (STFT) to identify abnormal vibration modes; infrared thermal imagers collect the surface temperature distribution of the blades and use deep learning algorithms to extract thermal features such as temperature gradients and hot spots; visual sensors acquire images of the blade surface and detect structural damage such as cracks and wear through convolutional neural networks (CNN); the temporal-domain vibration signals, thermal infrared spectra, and visual image features are spatiotemporally aligned, and a feature-level fusion strategy is adopted. Principal component analysis (PCA) is used for dimensionality reduction and noise reduction to form a high-dimensional feature vector containing structural, thermodynamic, and dynamic information, thus forming multidimensional data.

[0042] Step S31 introduces an improved complete set empirical mode decomposition adaptive noise algorithm to suppress mode aliasing during data processing.

[0043] Specifically, firstly, multiple sets of Gaussian white noise with different amplitudes are added to the original signal to form a ensemble signal; secondly, intrinsic mode functions are extracted layer by layer through empirical mode decomposition; finally, the average of the multiple decomposition results is calculated to eliminate noise interference and obtain a clean signal component, as shown in the following formula:

[0044] in, The original signal, For the first eigenmode functions of order 1 This represents the final remaining trend term; The adaptive noise algorithm formula is as follows:

[0045] in, After the first The output result of this time; Empirical mode decomposition; It is a function of time t; For the first Gaussian white noise in the next iteration; This is the noise amplitude coefficient; For the first Energy operator for the first-order intrinsic mode function.

[0046] In the signal decomposition process for wind turbine blade anomaly detection, after performing mode decomposition on the original signal, the six most representative components are selected from the mode components. The specific formula is as follows:

[0047] in, For component values; After the first The output result of this time; These are modal correction coefficients; for iteration For the first Gaussian white noise for each iteration; (k=1,...,6); By constructing a time-frequency attention model, spectral analysis is performed on multimodal data such as vibration and stress to calculate the energy distribution of each frequency band. Based on an energy weight allocation strategy, dynamic weights are assigned to the spectral characteristics of different modes to highlight the characteristic information of fault-sensitive frequency bands. While suppressing background noise interference, the identification of fault characteristics is enhanced, thereby improving the accuracy and reliability of blade anomaly detection. The attention weight calculation formula is as follows:

[0048] in, For the first Attention weights corresponding to each element; Temperature coefficient; To sum the exponent terms under all index j, normalization is performed.

[0049] Based on the calculated weights, a weighted principal component analysis method is used to assign different weights to the contribution of features to blade anomaly detection. While retaining key information, the original high-dimensional data is reduced to 16 dimensions, which reduces computational complexity and improves anomaly detection efficiency.

[0050] Step S32: Construct a low-complexity feature extraction model based on the historical operation data supplemented by the data acquisition and monitoring control system; Parameter relationship diagram constructed based on wind turbine operation mechanism Node set Core indicators include wind speed, power, rotational speed, propeller pitch angle, temperature, and vibration frequency; the weight of E represents the physical correlation between parameters; a graph convolutional neural network (GCN) is used to mine the complex relationships contained in the graph, and the core calculation formula is:

[0051] in, For the first +1 hidden layer feature matrix; =A+I is an adjacency matrix with self-loops; It is a degree matrix; For the first Layer features; These are learnable weights; For activation functions; By employing channel-wise convolution to perform convolution operations independently for each input channel, the focus is on local feature extraction; while point-wise convolution achieves cross-channel information fusion through 1×1 convolution, optimizing the network hierarchical structure and outputting 32-dimensional features.

[0052] The 16-dimensional blade time-domain vibration characteristics are deeply fused with the 32-dimensional operating status characteristics recorded in real time by the wind turbine data acquisition and monitoring control system. The fused 48-dimensional feature vector is normalized to standardize the input distribution of each layer, which can alleviate the gradient vanishing problem and accelerate the model convergence speed.

[0053] S4 employs a three-layer process of residual cleaning, temporal smoothing, and probability output to perform hierarchical processing on the data after feature extraction, enabling anomaly detection under abnormal operating conditions. Step S41: The kernel density estimation method is used to model the residual distribution under normal conditions; Kernel density estimation is a non-parametric estimation method that does not require prior assumptions about the data distribution and can adaptively fit the true distribution of the data. Kernel density is calculated for the 48-dimensional feature vector recorded by the fused vibration and wind turbine data acquisition and monitoring control system. The kernel density calculation formula is as follows:

[0054] in, Real-time residual The probability density estimate at point ; The number of historical residual samples used for modeling; For real-time residuals; It is the first in the historical residual sample One value; For adaptive bandwidth; K(·) is a Gaussian function; Based on the calculated kernel density Dynamic thresholds are used to detect and judge anomalies in wind turbine blades, which facilitates the identification of anomalies under different operating conditions and avoids false positives or missed detections caused by fixed thresholds. The dynamic threshold adjustment formula is as follows:

[0055] in, For dynamic thresholds; For real-time residuals; Real-time residual The probability density estimate at point ; Based on the normal fluctuation range of the blade under different wind speed and temperature environments, the upper limit of the residual of 99% of normal samples is set as the anomaly judgment threshold. This effectively filters out small fluctuations during normal operation and accurately captures residual mutations caused by abnormal states such as blade structural damage and surface cracks.

[0056] Step S42: Use a hidden Markov model to capture the temporal correlation of the residual sequence to obtain the current blade state.

[0057] In wind turbine blade anomaly detection, a Hidden Markov Model (HMM) is used for analysis to effectively capture the temporal correlation of the residual sequence; the system state space is defined as... The state transition matrix is ​​initialized as follows:

[0058] in, The probability of a wind turbine blade transitioning from a normal state to an abnormal state is 5%, reflecting the potential for fault evolution during normal operation. The probability of the equipment recovering from an abnormal state to a normal state is 3%, which reflects the possibility that the equipment can recover naturally or be repaired after a failure.

[0059] Anomalies are detected by outputting residual sequences and state probabilities. The residual sequence records the difference between actual monitoring data and theoretical model predictions, capturing subtle deviations in the operation of wind turbine blades. The state probability dynamically describes the probability that the blades are in normal or abnormal states at time t, enabling anomaly detection to find potential risk signals in uncertainty.

[0060] S5 introduces a two-factor confirmation mechanism and an anomaly confidence score system to determine the abnormality of wind turbine blades.

[0061] Specifically, the two-factor confirmation mechanism includes: making a preliminary anomaly judgment on the real-time residual and the dynamic threshold; if the absolute value of the real-time residual is greater than the dynamic threshold, an instantaneous anomaly flag is triggered; otherwise, it is not triggered. Then, by judging the probability of the time series state and the threshold, a secondary anomaly state assessment is performed to obtain a secondary time series anomaly flag. If both the instantaneous anomaly flag and the secondary time series anomaly flag are satisfied, it is determined to be a state anomaly, and at this time, the wind turbine blade anomaly alarm is triggered.

[0062] The anomaly confidence scoring system includes: calculating the anomaly confidence score simultaneously when an alarm for an anomaly in the wind turbine blades is triggered. The formula for the anomaly confidence score is as follows:

[0063] in, Scoring for abnormal confidence levels; This represents all observed data from the start time to time t. Under the given conditions, the probability that the blade is in an abnormal state at time t; For dynamic thresholds; For real-time residuals; When set A value >1.0 indicates a high-risk condition, suggesting severe damage to the blade structure or a rapid deterioration in performance, requiring immediate shutdown and emergency maintenance. At this point, the risk level is considered medium, indicating that the blades have potential defects or a detectable decline in performance, requiring maintenance. When the risk is low, the fault trend is monitored to ensure that potential problems are detected in time before the risk escalates. The abnormal state is divided into three levels for judgment, and different maintenance plans are adopted according to different abnormal situations to achieve effective management of the entire abnormal process.

[0064] This invention provides a method for detecting anomalous blades in wind turbines. It collects vibration data from the wind turbine blades and operating parameters of the wind turbine as an initial dataset. The initial dataset is preprocessed using a moving average filtering method and a linear interpolation algorithm to obtain a target dataset. Multimodal feature extraction is then used to extract features from the target dataset. A three-layer process—residual cleaning, temporal smoothing, and probability output—is employed to perform hierarchical processing on the feature-extracted data, enabling anomaly detection under abnormal operating conditions. A two-factor confirmation mechanism and an anomaly confidence scoring system are introduced to determine the anomalous status of the wind turbine blades. By using a moving average filtering method and a linear interpolation algorithm for data preprocessing, feature extraction, and hierarchical data processing, a more accurate detection of blade anomalies is achieved.

[0065] Example 2 like Figure 2 As shown, the present invention provides a blade anomaly detection system for wind turbines, comprising: The data acquisition unit collects vibration data from the wind turbine blades and operating parameters of the wind turbine as the initial dataset; The data preprocessing unit uses the moving average filtering method and the linear interpolation algorithm to preprocess the data in the initial dataset to obtain the target dataset; The feature extraction unit uses a multimodal feature extraction method to extract features from the data in the target dataset; The anomaly detection unit employs a three-layer process of residual cleaning, temporal smoothing, and probability output to perform hierarchical processing on the feature-extracted data, enabling anomaly detection under abnormal operating conditions. The anomaly detection unit introduces a two-factor confirmation mechanism and an anomaly confidence score system to determine the anomaly status of the wind turbine blades.

[0066] Example 3 like Figure 3 As shown, the present invention provides a network-side server including at least one processor 302; and a memory 301 communicatively connected to at least one processor 302; wherein the memory 301 stores instructions that can be executed by at least one processor 302, and the instructions are executed by at least one processor 302 to enable at least one processor 302 to perform the above-described data processing method. The memory 301 and processor 302 are connected via a bus, which may include any number of interconnecting buses and bridges. The bus connects various circuits of one or more processors 302 and memory 301 together. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by processor 302 is transmitted over a wireless medium via an antenna, which further receives data and transmits it to processor 302.

[0067] Processor 302 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory 301 can be used to store data used by processor 302 during operation.

[0068] Example 4 This invention provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements a method for detecting abnormalities in wind turbine blades.

[0069] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0070] This application is described with reference to flowchart illustrations and / or block diagrams of methods, systems, and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0071] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0072] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0073] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the scope of the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0074] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can be appropriately combined to form other embodiments that can be understood by those skilled in the art. The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for detecting abnormalities in wind turbine blades, characterized in that, include: S1, collect vibration data from the wind turbine blades and operating parameters from the wind turbine as the initial dataset; S2, the moving average filtering method and linear interpolation algorithm are used to preprocess the data in the initial dataset to obtain the target dataset; S3 uses multimodal feature extraction to extract features from the target dataset; S4 employs a three-layer process of residual cleaning, temporal smoothing, and probability output to perform hierarchical processing on the data after feature extraction, enabling anomaly detection under abnormal operating conditions. S5 introduces a two-factor confirmation mechanism and an anomaly confidence score system to determine the abnormality of wind turbine blades.

2. The method for detecting abnormalities in wind turbine blades according to claim 1, characterized in that, In S2, the moving average filtering method and linear interpolation algorithm are used to preprocess the data in the initial dataset to obtain the target dataset, which includes... Step S21: Use the moving average filtering method to identify outliers in the vibration parameters collected in the initial dataset. Step S22: Use a linear interpolation algorithm to complete the historical operation data of the data acquisition and monitoring control system in the initial dataset.

3. The method for detecting abnormalities in wind turbine blades according to claim 2, characterized in that, In step S21, the acquisition of vibration parameters includes symmetrically installing high-precision triaxial accelerometers in the key stress concentration area at the blade root, setting an ultra-high frequency sampling frequency, and capturing three-dimensional vibration signals in the radial (X), tangential (Y), and axial (Z) directions in real time. The vibration signals are... During the data collection process, a temperature threshold is set to exclude sensor data below -40℃. The formula for judging low-temperature environment data is as follows: in, For real-time temperature, The filtered vibration signal This is vibration signal data under normal conditions. The temperature requirement is not met.

4. The method for detecting abnormalities in wind turbine blades according to claim 3, characterized in that, In step S22, the linear interpolation algorithm includes acquiring a data sequence containing core operating parameters such as wind speed, generator speed, pitch angle, and active power in real time from the wind turbine control system, and then using a linear interpolation algorithm to complete the data. The formula for the linear interpolation algorithm is as follows: in, The target value is obtained after linear interpolation calculation. and Given the dependent variable values ​​of two known data points, For the target value of the independent variable that needs to be interpolated, and To and and The corresponding known values ​​of the independent variable.

5. The method for detecting abnormalities in wind turbine blades according to claim 1, characterized in that, In S3, multimodal feature extraction is used to extract features from the target dataset, including... Step S31: An improved complete set empirical mode decomposition adaptive noise algorithm is introduced to suppress mode aliasing during data processing. The formula for the adaptive noise algorithm is as follows: in, After the first The output result of this time; Empirical mode decomposition; It is a function of time t; For the first Gaussian white noise in the next iteration; This is the noise amplitude coefficient; For the first Energy operator of the first-order eigenmode function; Step S32: Construct a low-complexity feature extraction model based on the historical operation data supplemented by the data acquisition and monitoring control system.

6. The method for detecting abnormalities in wind turbine blades according to claim 1, characterized in that, In S4, a three-layer process of residual cleaning, temporal smoothing, and probability output is used to process the data after feature extraction. Anomaly detection under abnormal operating conditions includes: Step S41: The kernel density estimation method is used to model the residual distribution under normal conditions; Step S42: Use a hidden Markov model to capture the temporal correlation of the residual sequence to obtain the current blade state.

7. The method for detecting abnormalities in wind turbine blades according to claim 1, characterized in that, In S5, a two-factor confirmation mechanism and an anomaly confidence score system are introduced to determine the abnormal situation of wind turbine blades. This includes: making a preliminary anomaly judgment based on real-time residuals and dynamic thresholds; if the absolute value of the real-time residuals is greater than the dynamic threshold, an instantaneous anomaly flag is triggered; otherwise, it is not triggered. Then, a secondary anomaly state assessment is performed by judging the probability of the time series state and the threshold, and a secondary time series anomaly flag is obtained. If both the instantaneous anomaly flag and the secondary time series anomaly flag are satisfied, it is determined to be a state anomaly. At this time, the wind turbine blade anomaly alarm is triggered, and the anomaly confidence score is calculated to classify the anomaly state into three levels for judgment. Then, different maintenance schemes are adopted according to different anomaly situations.

8. A blade anomaly detection system for wind turbine generators, characterized in that, include: The data acquisition unit collects vibration data from the wind turbine blades and operating parameters of the wind turbine as the initial dataset; The data preprocessing unit uses the moving average filtering method and the linear interpolation algorithm to preprocess the data in the initial dataset to obtain the target dataset; The feature extraction unit uses a multimodal feature extraction method to extract features from the data in the target dataset; The anomaly detection unit employs a three-layer process of residual cleaning, temporal smoothing, and probability output to perform hierarchical processing on the feature-extracted data, enabling anomaly detection under abnormal operating conditions. The anomaly detection unit introduces a two-factor confirmation mechanism and an anomaly confidence score system to determine the anomaly status of the wind turbine blades.

9. A network-side server, characterized in that, include: At least one processor; The system includes a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the blade anomaly detection method for wind turbines as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for detecting abnormal blades of wind turbines as described in any one of claims 1 to 7.

Images