Real-time diagnosis method and system for blade damage of wind generating set, processing equipment and storage medium

Abstract

The invention relates to a wind generating set blade damage real-time diagnosis method and system, processing equipment and a storage medium. The method comprises the steps that vibration signals and stress signals of all measuring points of three blades of a to-be-measured wind generating set are acquired; carrying out preprocessing and initial phase compensation on vibration signals and stress signals of each measuring point of three blades of the wind generating set to be measured; obtaining a vibration signal and a stress signal of each measuring point of each blade at equal rotation angle intervals; calculating the vibration-stress coherence coefficient of each blade of the wind generating set to be measured and the Pearson correlation coefficient, the frequency domain correlation and the dispersion of the measuring point signal of the same position of each blade; the health score or damage probability of each blade of the wind generating set to be detected is calculated, when the calculated health score or damage probability is lower than a preset health score threshold value or exceeds a preset damage probability threshold value, the corresponding blade is damaged, and an alarm is given. The method can be widely applied to the field of wind generating set state monitoring.

Description

Technical Field

[0001] This invention relates to the field of wind turbine condition monitoring, and in particular to a method, system, processing equipment, and storage medium for real-time diagnosis of wind turbine blade damage. Background Technology

[0002] As the core component of energy conversion, wind turbine blades are in a complex and ever-changing working environment for a long time, facing the combined effects of various mechanical and environmental loads. The structural health of the blades directly affects power generation efficiency and operational safety.

[0003] Currently, blade monitoring methods mainly suffer from the following technical bottlenecks: 1) Insufficient single-modal monitoring. Vibration monitoring: Traditional vibration monitoring analyzes blade vibration by installing a biaxial vibration sensor at a specific location. While it can detect the overall blade mode, it cannot diagnose early micro-damage and cannot effectively distinguish between environmental load fluctuations and actual structural damage. Stress monitoring: Although stress sensors such as fiber optic gratings can capture local strain, current technology has not solved the signal conversion problem between dynamic blades and static towers, and temperature drift leads to a measurement error of ±0.5με / ℃. 2) Defects in signal processing methods. Currently, commonly used signal processing methods are divided into two categories. The first category is based on fault mechanisms, judging blade conditions by setting alarm thresholds. This method does not effectively distinguish between different operating conditions, requires a large amount of manual processing, and has low diagnostic accuracy. The second category uses machine learning to process vibration signals, but it relies on a large amount of labeled data for training, and the model's generalization ability is insufficient in the small sample scenarios of actual wind farms.

[0004] In addition, current blade monitoring methods suffer from the following industry pain points: 1) Insufficient detection sensitivity. The smallest identifiable crack size in existing technologies is 5cm, which cannot meet the early warning requirements for early micro-damage (<1cm). The detection rate of composite material delamination defects is less than 60%, and the positioning error exceeds ±1.5m. 2) Poor detection accuracy. The false alarm rate of vibration monitoring methods reaches 35%, and the false alarm rate reaches 25%. For micro-cracks or early damage, the false alarm rate can reach 40%, making it impossible to meet practical application requirements. Summary of the Invention

[0005] To address the aforementioned problems, the purpose of this invention is to provide a method, system, processing equipment, and storage medium for real-time diagnosis of wind turbine blade damage, which offers high diagnostic accuracy, high detection sensitivity, and strong detection precision.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: Firstly, it provides a method for real-time diagnosis of wind turbine blade damage, comprising:

[0007] Simultaneously sample the vibration and stress points of the three blades of the wind turbine generator under test to obtain the vibration and stress signals of each measuring point of the three blades of the wind turbine generator under test. The vibration and stress signals at each measuring point of the three blades of the wind turbine generator under test are preprocessed and initially phase compensated. The vibration and stress signals of each measuring point on each of the three blades of the wind turbine generator under test are processed in a secondary manner after initial phase compensation to obtain vibration and stress signals of each measuring point on each blade at equal rotation angle intervals. Based on the vibration and stress signals of each measuring point on each blade at equal rotation angle intervals, the vibration-stress coherence coefficient of each blade of the wind turbine generator under test, as well as the Pearson correlation coefficient, frequency domain correlation, and dispersion of the signals of measuring points at the same position on each blade are calculated. Based on the calculated vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation, and dispersion, the health score or damage probability of each blade of the wind turbine generator under test is calculated. When the calculated health score or damage probability is lower than the preset health score threshold or exceeds the preset damage probability threshold, the corresponding blade is damaged and an alarm is triggered.

[0008] Furthermore, the synchronous sampling of vibration and stress measurement points of the three blades of the wind turbine under test to obtain vibration and stress signals at each measurement point of the three blades of the wind turbine under test also includes: The system calculates various characteristic parameters of vibration and stress signals at each measuring point of the three blades of the wind turbine generator under test. When the calculated characteristic parameters exceed the preset alarm threshold, an alarm signal is sent to achieve the first level of over-limit alarm.

[0009] Furthermore, the preprocessing and initial phase compensation of the vibration and stress signals at each measuring point of the three blades of the wind turbine generator under test includes: The vibration and stress signals of each measuring point on the three blades of the wind turbine generator under test are preprocessed. The reference blade of the wind turbine generator is taken as the first blade, and the other two blades of the wind turbine generator are defined clockwise as the second blade and the third blade. Using the phase of the pre-processed vibration and stress signals at each measuring point of the first blade as the reference 0°, and the same type of signals of the second and third blades lagging by 120° and 240° respectively, the actual phase difference between the vibration and stress signals at each measuring point of the second and third blades and the vibration and stress signals at each measuring point of the reference blade is calculated. Based on the calculated actual phase difference, a time-shift correction method is used to compensate the phase of the vibration and stress signals at each measuring point of the second and third blades to be consistent with the theoretical mechanical installation angle.

[0010] Furthermore, the vibration and stress signals of each measuring point on the three blades of the wind turbine under test, after initial phase compensation, are processed a second time to obtain vibration and stress signals at equal rotation angle intervals at each measuring point on each blade, including: Short-time Fourier transforms were performed on the vibration and stress signals at each measuring point of the three blades after initial phase compensation to obtain the corresponding time spectrum; Extract the principal rotation frequency of the spectrum at each time point, and obtain the curve of the corresponding principal rotation frequency component changing with time, as the fundamental frequency variation spectrum; By integrating the fundamental frequency variation spectrum over time, the instantaneous rotation phase of each blade can be obtained. Based on the angle spectrum of each blade, the vibration and stress signals of the three blades sampled at equal time intervals are resampled into vibration and stress signals with equal rotation angle intervals through spline interpolation.

[0011] Furthermore, the calculation of the dispersion of the measurement points at the same position on the three blades of the wind turbine generator under test includes: The signals from the same measurement points on the three blades of the wind turbine generator under test are subjected to continuous wavelet transform, and the Morlet wavelet basis function is used to obtain the corresponding time-frequency distribution. Calculate the wavelet cross spectrum of the signal at the same position of every two blades of the wind turbine generator under test; Dividing the squared amplitude of the calculated wavelet cross spectrum by the product of the two smoothed self-spectrums yields three wavelet coherence spectrum matrices. Calculate the sample entropy of each wavelet coherence spectrum matrix to measure the complexity of the coherent mode; The dispersion of the three blades is calculated based on the sample entropy of each wavelet coherence spectrum matrix.

[0012] Furthermore, based on the calculated vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation, and dispersion, the health score or damage probability of each blade of the wind turbine generator under test is calculated. When the calculated health score or damage probability is lower than a preset health score threshold or exceeds a preset damage probability threshold, the corresponding blade is damaged, and an alarm is triggered, including: The calculated vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation, and dispersion are input into a pre-built deep learning model to obtain the health score or damage probability of each blade of the wind turbine generator under test. When the calculated health score or damage probability is lower than the preset health score threshold or exceeds the preset damage probability threshold, the corresponding blade is damaged and an alarm is triggered.

[0013] Furthermore, the deep learning model includes: The input layer is used to input the calculated feature values, including the vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation and dispersion, and after standardization, they are mapped to the [0, 1] interval; The attention fusion layer is used to enable the model to automatically learn the importance of different feature dimensions in health status determination and perform weighted fusion to enhance the contribution of discriminative features. The triplet loss function is used to construct triplets and train the model so that the distance between the feature vector Anchor and the feature vector Positive in the feature space is much smaller than the distance between the feature vector Anchor and the feature vector Negative. This allows the model to better learn the difference between healthy and damaged states. Here, Anchor is the feature vector of a healthy sample, Positive is the feature vector of another healthy sample, and Negative is the feature vector of a damaged sample. The output layer is used to output a health score or damage probability. The alarm layer is used to issue an alarm when the output health score or damage probability is lower than a preset health score threshold or exceeds a preset damage probability threshold.

[0014] Secondly, a real-time diagnostic system for wind turbine blade damage is provided, including a multi-dimensional sensing end, an edge computing acquisition unit, a network transmission system and a cloud system, wherein the multi-dimensional sensing end includes a dual-axis fiber optic temperature and vibration integrated sensor and a fiber optic strain sensor. Each blade of the wind turbine generator set is equipped with a dual-axis fiber optic temperature and vibration integrated sensor at the blade root, 1 / 3L from the blade root, and the blade tip, where L is the blade length; each blade of the wind turbine generator set is equipped with four fiber optic strain sensors in the main beam cap area, and two fiber optic strain sensors in the leading and trailing edge areas of each blade of the wind turbine generator set. The dual-axis fiber optic temperature and vibration integrated sensor is used to collect temperature and vibration signals at corresponding positions of each blade of the wind turbine generator in real time. The fiber optic strain sensor is used to collect stress signals at corresponding positions of each blade of the wind turbine generator in real time. The edge computing data acquisition unit is used to perform analog-to-digital conversion, preprocessing, and compression of vibration and stress signals from each measuring point on the three blades of the wind turbine generator under test, and then send them to the network transmission system. The network transmission system is used to transmit the data sent by the edge computing collector to the designated server of the booster station via the wireless module; and to transmit the data sent by the edge computing collectors of different wind turbine generators to other wireless modules in a step-by-step manner via the wireless module, and then to the designated server of the corresponding booster station. The cloud system is used to receive data uploaded by the designated server of the booster station, calculate the vibration-stress coherence coefficient of each blade of the wind turbine generator under test, as well as the Pearson correlation coefficient, frequency domain correlation, and dispersion of the signals at the same location of each blade, and then calculate the health score or damage probability of each blade of the wind turbine generator under test. When the calculated health score or damage probability is lower than the preset health score threshold or exceeds the preset damage probability threshold, the corresponding blade is damaged and an alarm is triggered.

[0015] Thirdly, a processing device is provided, including a computer program, wherein when the computer program is executed by the processing device, it is used to implement the steps corresponding to the above-mentioned real-time diagnosis method for wind turbine blade damage.

[0016] Fourthly, a computer-readable storage medium is provided, on which a computer program is stored, wherein the computer program, when executed by a processor, is used to implement the steps corresponding to the above-described real-time diagnosis method for wind turbine blade damage.

[0017] The present invention has the following advantages due to the adoption of the above technical solutions: 1. This invention, through synchronous monitoring of vibration and stress sensors and combined with a multi-dimensional similarity assessment system, can effectively warn of early micro-damage (<1cm) to wind turbine blades, significantly improving detection sensitivity.

[0018] 2. By constructing a three-level diagnostic system, this invention utilizes various techniques such as short-time Fourier transform, time-domain similarity analysis, frequency-domain similarity analysis, and time-frequency joint analysis to more accurately determine the health status of blades, significantly reducing the false alarm rate (35%) and false alarm rate (25%) of traditional vibration monitoring technology, especially for the false alarm rate of microcracks or early damage.

[0019] 3. This invention employs a nonlinear sampling rate adjustment and energy optimization strategy, dynamically adjusting the sampling rate based on blade vibration differences and temperature conditions. This ensures data accuracy while effectively reducing system power consumption and enhancing system adaptability and stability.

[0020] 4. The application of the edge computing collector of the present invention enables real-time data acquisition and processing, avoids data loss, and at the same time, cloud data storage and analysis ensures long-term data preservation and efficient processing, thereby improving the overall robustness of the system.

[0021] 5. This invention constructs an efficient deep learning model by introducing a triplet loss function and an attention fusion mechanism, which can automatically learn the feature representation of the blade health status, thereby improving the accuracy and efficiency of fault diagnosis.

[0022] 6. This invention combines multi-source data fusion of vibration, stress and environmental signals to achieve a comprehensive assessment of the health status of the blades and improve the level of intelligent diagnosis.

[0023] 7. Based on the similarity difference localization algorithm, the present invention can quickly locate the faulty blade and its specific location, reducing unnecessary maintenance and repair work and optimizing resource utilization.

[0024] 8. By improving detection sensitivity and accuracy, and reducing false alarms and missed alarms, this invention effectively reduces the operation and maintenance costs of wind turbine generator sets and improves economic efficiency.

[0025] In summary, this invention brings significant benefits in terms of significantly improving detection sensitivity and accuracy, enhancing system robustness and stability, improving diagnostic efficiency and intelligence, and optimizing resource utilization and reducing costs. It can be widely applied in the field of wind turbine generator condition monitoring. Attached Figure Description

[0026] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. In the drawings: Figure 1 This is a schematic diagram of the system structure provided in an embodiment of the present invention. Detailed Implementation

[0027] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey the scope of the invention to those skilled in the art.

[0028] It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” as used herein may also include the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a particular order described or illustrated unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.

[0029] Although terms such as first, second, third, etc., may be used in this document to describe multiple elements, components, regions, layers, and / or segments, these elements, components, regions, layers, and / or segments should not be limited by these terms. These terms may be used only to distinguish one element, component, region, layer, or segment from another. Unless the context clearly indicates otherwise, terms such as "first," "second," and other numerical terms used herein do not imply order or sequence. Therefore, the first element, component, region, layer, or segment discussed below may be referred to as the second element, component, region, layer, or segment without departing from the teachings of the exemplary embodiments.

[0030] Currently, blade monitoring methods suffer from technical bottlenecks due to limitations in single-modal monitoring and deficiencies in signal processing methods. Furthermore, current blade monitoring methods suffer from insufficient detection sensitivity and poor detection accuracy, which are common pain points in the industry. This invention provides a real-time damage diagnosis method for wind turbine blades, comprising: simultaneously sampling vibration and stress measurement points of three blades of the wind turbine under test to obtain vibration and stress signals at each measurement point of the three blades; preprocessing and performing initial phase compensation on the vibration and stress signals at each measurement point of the three blades; performing secondary processing on the initially phase-compensated vibration and stress signals at each measurement point of the three blades to obtain vibration and stress signals at equal rotational angle intervals for each blade; and based on each blade… Vibration and stress signals at equal rotational angle intervals at each measuring point are used to calculate the vibration-stress coherence coefficient of each blade of the wind turbine under test, as well as the Pearson correlation coefficient, frequency domain correlation, and dispersion of the signals from measuring points at the same location on each blade. Based on the calculated vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation, and dispersion, the health score or damage probability of each blade of the wind turbine under test is calculated. When the calculated health score or damage probability is lower than a preset health score threshold or exceeds a preset damage probability threshold, the corresponding blade is damaged, and an alarm is triggered. This invention uses techniques such as time-domain similarity, frequency-domain similarity, and time-frequency joint analysis to achieve real-time diagnosis of blade structural damage, and is particularly suitable for detecting cracks and delamination defects in composite blades; dynamic signal decoupling analysis under rotating conditions; and multi-blade correlation comparison diagnosis.

[0031] Example 1 This embodiment provides a method for real-time diagnosis of wind turbine blade damage, including the following steps: 1) Simultaneously sample the vibration and stress measurement points of the three blades of the wind turbine generator under test, obtain the vibration and stress signals of each measurement point of the three blades of the wind turbine generator under test, and calculate the corresponding peak-to-peak value, effective value and kurtosis and other characteristic parameters. When the calculated characteristic parameters exceed the preset alarm threshold, send an alarm signal to realize the first level of over-limit alarm.

[0032] 2) Preprocess and perform initial phase compensation on the vibration and stress signals at each measuring point of the three blades of the wind turbine generator under test, specifically as follows: 2.1) Preprocess the vibration and stress signals of each measuring point on the three blades of the wind turbine generator under test, including filtering and preliminary amplitude calibration, to improve data quality.

[0033] 2.2) Since the three blades of the wind turbine are installed at 120° intervals and the mechanical phase difference is fixed, the reference blade of the wind turbine is taken as the first blade, and the other two blades of the wind turbine are defined as the second and third blades in clockwise direction.

[0034] 2.3) Using the phase of the pre-processed vibration and stress signals at each measuring point of the first blade as the reference (0°), the signals of the same type of the second and third blades are lagging by 120° and 240° respectively. Calculate the actual phase difference between the vibration and stress signals at each measuring point of the second and third blades and the vibration and stress signals at each measuring point of the reference blade.

[0035] 2.4) Based on the calculated actual phase difference, the time shift correction method is used to compensate the phase of the vibration signal and stress signal at each measuring point of the second and third blades to be consistent with the theoretical mechanical installation angle (120°, 240°).

[0036] 3) After initial phase compensation, the vibration and stress signals at each measuring point on the three blades of the wind turbine generator under test are subjected to short-time Fourier transform, fundamental frequency spectrum calculation, time integration, and resampling processing to obtain vibration and stress signals at equal rotation angle intervals at each measuring point on each blade, in order to eliminate the influence of speed fluctuations on the waveform. Specifically: 3.1) Perform short-time Fourier transform on the vibration and stress signals at each measuring point of the three blades after initial phase compensation to obtain the corresponding time spectrum. ,in, For time, For frequency.

[0037] 3.2) Extract the principal rotation frequency of the spectrum at each time interval. Frequency, get the corresponding The curve of frequency component changing with time is denoted as the fundamental frequency variation spectrum. It directly reflects the fluctuation of the rotational speed.

[0038] 3.3) Spectra of each fundamental frequency variation By performing time integration, the instantaneous rotational phase, i.e., the angle spectrum, of each blade is obtained. This indicates the relationship between the blade rotation angle and time: (1) in, This is the initial phase.

[0039] 3.4) Based on the angle spectrum of each blade The vibration and stress signals of the three blades, originally sampled at equal time intervals, are resampled using spline interpolation to obtain vibration and stress signals at equal rotational angle intervals. In this way, each data point corresponds to a fixed rotational angle, eliminating the stretching or compression effects on the signal waveform caused by uneven rotational speeds. This ensures a strict correspondence between signal characteristics and rotational position, facilitating synchronous comparison between blades.

[0040] 4) Based on the vibration and stress signals at equal rotational angle intervals at each measuring point on each blade, calculate the vibration-stress coherence coefficient of each blade of the wind turbine generator under test. When a blade is damaged, its impeller rotation frequency coherence coefficient is... It will decrease the impeller's frequency coherence coefficient. The coefficient will increase.

[0041] Specifically, the formula for calculating the vibration-stress coherence coefficient is as follows: (2) in, The vibration-stress coherence coefficient is used to quantify the degree of linear coherence between the vibration response and stress response on the same blade. For healthy blades, the vibration and stress coupling modes are stable at the impeller rotation frequency, and this coefficient is also relatively stable. For vibration ( ) and stress ( The cross-power spectrum of ) For vibration ( ) and stress ( The self-power spectrum of ) These are the leaf identifiers, corresponding to the first leaf, second leaf, and third leaf.

[0042] 5) Based on the vibration and stress signals of each measuring point on each blade at equal rotation angle intervals, calculate the Pearson correlation coefficient of the signals of the same measuring point on each blade of the wind turbine generator under test.

[0043] Specifically, regarding the blade rotation characteristics, a dynamic weighting coefficient is applied. When a blade of the wind turbine under test is damaged, the temporal similarity of that blade will decrease. (First blade) With the second blade The formula for calculating the Pearson correlation coefficient is: (3) in, For temporal similarity; These are dynamic weighting coefficients, and , This represents the number of sampling points; For the first blade After equal-angle resampling, the first Signal amplitude at each measuring point; For the second blade After equal-angle resampling, the first Signal amplitude at each measuring point; For the first blade The weighted average of the signal, and ; For the second blade The weighted average of the signal. Second blade. With the third blade Pearson correlation coefficient and the first blade With the third blade The Pearson correlation coefficient is similar to that of formula (2), so I will not elaborate further here.

[0044] Calculate the first blade according to the above formula. With the second blade Pearson correlation coefficient, second leaf With the third blade Pearson correlation coefficient and the first blade With the third blade By comparing and analyzing the three sets of Pearson correlation coefficients, it is possible to determine the specific blade that has failed.

[0045] 6) Based on the vibration and stress signals of each measuring point on each blade at equal rotation angle intervals, calculate the frequency domain correlation of the measuring points at the same position on the three blades of the wind turbine generator under test.

[0046] Specifically, when a blade of the wind turbine generator under test fails, the frequency domain similarity will decrease. The formula for calculating the frequency domain similarity is: (4) in, Frequency domain similarity; For the first Fourier transform (FFT) of the blade signal at frequency Complex values ​​at the location.

[0047] 7) Based on the vibration and stress signals at equal rotation angle intervals at each measuring point on each blade, calculate the dispersion of the signals at the same position on the three blades of the wind turbine generator under test, specifically: 7.1) A three-channel wavelet coherence spectrum matrix is ​​constructed using Morlet wavelet basis function decomposition, and the relative deviation of the sample entropy (SampEn) of each channel is calculated: 7.1.1) Perform continuous wavelet transform (CWT) on the signals from the same measurement points on the three blades of the wind turbine generator under test, and use Morlet wavelet basis functions to obtain the corresponding time-frequency distribution. ,in, For phase, The parameter controlling the scaling of the mother wavelet function determines the width of the wavelet window, which is adjusted... It can analyze the vibration characteristics of blades at different frequencies, thereby identifying the abnormal frequency components corresponding to the fault.

[0048] 7.1.2) Calculate the wavelet cross spectrum of the signal at the same position of each pair of blades of the wind turbine generator under test.

[0049] 7.1.3) Divide the squared amplitude of the calculated wavelet cross spectrum by the product of the two smoothed autospectral values ​​to obtain three wavelet coherence spectrum matrices (corresponding to...). - blade, - blade, - blade).

[0050] 7.1.4) Calculate the sample entropy of each wavelet coherence spectral matrix to measure the complexity of the coherent mode. The sample entropy is denoted as... , , (corresponding to respectively) - blade, - blade, - blade).

[0051] 7.2) Sample entropy based on each wavelet coherence spectrum matrix , , The dispersion of the three blades is calculated, where the dispersion increases when a problem occurs in one of the blades.

[0052] The formula for calculating the dispersion is: (5) in, This represents the dispersion; a larger value indicates a greater time-frequency difference among the three blades. Entropy of three samples , , The average value.

[0053] 8) Based on the calculated vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation, and dispersion, calculate the health score or damage probability of each blade of the wind turbine generator under test. When the calculated health score or damage probability is lower than a preset health score threshold or exceeds a preset damage probability threshold, the corresponding blade is damaged, and an alarm is triggered. Specifically: 8.1) Construct a deep learning model.

[0054] Specifically, a deep learning model includes an input layer, an attention fusion layer, a triplet loss function, an output layer, and an alarm layer.

[0055] The input layer is used to input the calculated feature values, including the vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation, and dispersion. After standardization, these values ​​are mapped to the [0, 1] interval. The standardization formula is as follows: (6) in, These are the standardized eigenvalues; These are the original eigenvalues. , These represent the maximum eigenvalue of the healthy sample and the minimum eigenvalue of the damaged sample, respectively.

[0056] The attention fusion layer is used to enable the model to automatically learn the importance of different feature dimensions in health status determination and perform weighted fusion to enhance the contribution of discriminative features.

[0057] The triplet loss function is used to construct triplet (Anchor, Positive, Negative). Through training, the distance between the feature vector Anchor and the feature vector Positive in the feature space is made much smaller than the distance between the feature vector Anchor and the feature vector Negative. This allows the model to better learn the differences between healthy and damaged states, improving discrimination and generalization ability. Here, Anchor is the feature vector of a healthy sample, Positive is the feature vector of another healthy sample (of the same class), and Negative is the feature vector of a damaged sample (of a different class).

[0058] The output layer is used to output a health score or damage probability.

[0059] The alarm layer is used to trigger an alarm when the output health score or damage probability is lower than a preset health score threshold or exceeds a preset damage probability threshold, prompting manual review of the alarm status.

[0060] Specifically, health score The calculation formula is: (7) in, These are the fusion features after training.

[0061] Specifically, damage probability The fused features after training are obtained by using the Sigmoid function. The probability value is converted to the interval [0,1]. The higher the probability value, the greater the risk of damage. The calculation formula is: (8) in, This is the sensitivity coefficient, with a preset value of 5. This is the critical value, with a default value of 0.5.

[0062] 8.2) Input the calculated vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation and dispersion into the constructed deep learning model to obtain the health score or damage probability of each blade of the wind turbine generator under test.

[0063] 8.3) When the calculated health score or damage probability is lower than the preset health score threshold or exceeds the preset damage probability threshold, the corresponding blade is damaged and an alarm is triggered.

[0064] Example 2 like Figure 1 As shown, this embodiment provides a real-time diagnostic system for wind turbine blade damage, including a multi-dimensional sensing end, an edge computing data acquisition unit, a network transmission system, and a cloud system. The multi-dimensional sensing end includes a dual-axis fiber optic temperature and vibration integrated sensor and a fiber optic strain sensor.

[0065] Each blade of the wind turbine generator is equipped with a biaxial fiber optic temperature and vibration sensor at its root, 1 / 3L from the root, and tip, where L is the blade length. Each blade's main beam cap area has four fiber optic strain sensors, and each blade's leading and trailing edge areas have two fiber optic strain sensors.

[0066] The dual-axis fiber optic temperature and vibration integrated sensor is used to collect temperature and vibration signals at corresponding positions of each blade of a wind turbine generator in real time.

[0067] Fiber optic strain sensors are used to acquire stress signals at corresponding positions on each blade of a wind turbine generator in real time.

[0068] The edge computing data acquisition unit is installed in the nacelle of the wind turbine generator set. It is used to perform analog-to-digital conversion, preprocessing and compression of vibration and stress signals from each measuring point on the three blades of the wind turbine generator set under test, and then send them to the network transmission system. It also calculates various characteristic parameters such as peak-to-peak value, RMS value and kurtosis.

[0069] The network transmission system is used to transmit data sent by the edge computing collector to the receiving end in the nacelle via a wireless module, and then transmit it to the designated server in the substation via the wind turbine ring network; and to transmit data sent by the edge computing collectors of different wind turbine units to the wireless modules closer to the wind farm in stages via wireless modules, and finally transmit it to the designated server in the corresponding substation wirelessly.

[0070] The cloud system receives data uploaded by the designated server of the booster station through centralized control or production operation and maintenance systems via isolation devices and firewalls. It calculates the vibration-stress coherence coefficient of each blade of the wind turbine generator under test, as well as the Pearson correlation coefficient, frequency domain correlation, and dispersion of the signals at the same location on each blade. Then, it calculates the health score or damage probability of each blade of the wind turbine generator under test. When the calculated health score or damage probability is lower than the preset health score threshold or exceeds the preset damage probability threshold, the corresponding blade is damaged and an alarm is triggered.

[0071] In a preferred embodiment, it should be noted that the nacelle, wind turbine ring network, booster station, centralized control or production operation and maintenance system, isolation device and firewall mentioned above are all components of the wind farm, wherein: The nacelle is the cabin on top of the wind turbine generator set, housing the gearbox, generator, and other major components. It is the source and primary aggregation point for data generation.

[0072] The wind turbine ring network is a local area network that connects all the equipment inside all the wind turbine generators in the wind farm. The edge computing data collector of this system is connected to the internal network of the generator through the wind turbine ring network, thereby transmitting the data back to the designated server.

[0073] A booster station is a facility within a wind farm that collects and steps up the electrical energy generated by each wind turbine to a higher voltage level before sending it to the power grid. These components together form the data transmission path from the individual turbine (nacelle) -> the internal network of the wind farm (wind turbine ring network) -> the data center of the wind farm (the designated server of the booster station), which is the essential channel for diagnostic data to flow from sensors to the cloud analysis platform.

[0074] The centralized control or production operation and maintenance system is the core software platform for wind farm operation and management. It is responsible for monitoring the real-time operating status of all wind turbine generators (SCADA data such as power, speed, and temperature), performing start-up and shutdown control, performance analysis, and operation and maintenance management. As a professional condition monitoring and fault diagnosis (CMS / PHM) subsystem, its data (diagnostic results, characteristic values, and alarms) needs to be integrated into the centralized control or production operation and maintenance system, or interact and link with the centralized control or production operation and maintenance system.

[0075] Isolation devices and firewalls are critical equipment for ensuring the network security of wind farm industrial control systems. Secure isolation must be maintained between the wind farm's production control area (including centralized control or production operation and maintenance systems, wind turbine controllers, and designated servers at the booster station) and the management information area (such as office networks and cloud interfaces). Isolation devices (such as network gateways) are used to physically or logically disconnect the direct connection between the two areas, performing protocol stripping and data transfer. Firewalls are used to implement access control policies. Data transmitted from the production control area to the management information area or the public network (cloud) must pass through these security devices to ensure the production system's security is protected from external network threats.

[0076] In a preferred embodiment, the four fiber optic strain sensors in the main beam cap region are positioned as follows: two fiber optic strain sensors are symmetrically arranged on the upper and lower surfaces of the wind turbine blade at a distance of 0.3L from the blade root, and two fiber optic strain sensors are symmetrically arranged on the upper and lower surfaces of the wind turbine blade at a distance of 0.6L from the blade root.

[0077] In a preferred embodiment, the two fiber optic strain sensors in the leading and trailing edge regions are positioned as follows: one fiber optic strain sensor is positioned at 15% chord length of the leading edge of the wind turbine blade, and another fiber optic strain sensor is positioned at the adhesive seam of the trailing edge of the wind turbine blade.

[0078] In a preferred embodiment, the edge computing data acquisition unit is a specialized data acquisition device designed for wind turbine blade condition monitoring and fault diagnosis. It is the core hardware of the entire system, connecting the front-end sensing sensors and the back-end platform system in the complex operating environment of the wind turbine. It features high-speed parallel sampling across all channels, strong anti-interference capabilities and electromagnetic compatibility, real-time edge computing without data loss at any given moment, and system self-recovery, network status self-checking, sensor status self-checking functions, and a variable data transmission strategy to ensure system stability.

[0079] Specifically, the edge computing data collector is equipped with an analog-to-digital conversion module, a preprocessing module, a signal processing module, an over-limit alarm module, and a compression and packaging module.

[0080] The analog-to-digital conversion module is used to convert the vibration and stress signals of each measuring point on the three blades of the wind turbine under test into analog and digital data, so as to obtain the vibration and stress data of each measuring point on the three blades of the wind turbine under test.

[0081] The preprocessing module is used to preprocess the vibration and stress data after analog-to-digital conversion, including filtering, preliminary amplitude calibration, short-time Fourier transform, fundamental frequency variation spectrum calculation, time integration, and resampling, in order to improve data quality.

[0082] The signal processing module is used to calculate various characteristic parameters such as peak-to-peak value, RMS value, and kurtosis of the preprocessed analog-to-digital converted vibration and stress data.

[0083] The over-limit alarm module is used to perform simple threshold judgment. When multiple calculated characteristic parameters exceed the preset alarm threshold, an alarm signal is sent to realize the first level of over-limit alarm. The alarm signal is transmitted to the network transmission system through the network module of the wind turbine generator.

[0084] The compression and packaging module is used to compress and package the pre-processed vibration and stress data, and add metadata such as timestamps, unit ID, and sensor status, which is then transmitted to the network transmission system through the network module of the wind turbine generator.

[0085] In a preferred embodiment, the cloud system includes a data storage module, a computing module, a deep learning module, and an alarm module.

[0086] The data storage module is used to store long-term historical monitoring data, characteristic data, and diagnostic records for all wind turbine blades.

[0087] The computation module is used to calculate the vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation, and dispersion of each blade of the wind turbine generator under test.

[0088] The deep learning module is used to build, train, and optimize deep learning models, and calculates the health score or damage probability of each blade of the wind turbine under test based on the calculated vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation, and dispersion.

[0089] The alarm module is used to send an alarm signal when the calculated health score or damage probability is lower than the preset health score threshold or exceeds the preset damage probability threshold, thus realizing the second-level over-limit alarm; and to manage all alarm events, generate diagnostic reports, and push them to maintenance personnel.

[0090] Example 3 This embodiment provides a processing device corresponding to the real-time diagnosis method for wind turbine blade damage provided in Embodiment 1. The processing device can be applied to client processing devices, such as mobile phones, laptops, tablets, desktop computers, etc., to execute the method of Embodiment 1.

[0091] The processing device includes a processor, a memory, a communication interface, and a bus. The processor, memory, and communication interface are connected via the bus to enable communication between them. The memory stores a computer program that can run on the processing device. When the processing device runs the computer program, it executes the real-time wind turbine blade damage diagnosis method provided in Embodiment 1.

[0092] In some implementations, the memory may be high-speed random access memory (RAM), and may also include non-volatile memory, such as at least one disk storage device.

[0093] In other implementations, the processor can be any type of general-purpose processor, such as a central processing unit (CPU) or a digital signal processor (DSP), and there is no limitation here.

[0094] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0095] Those skilled in the art will understand that the structure of the above-described computing device is only a partial structure related to the present invention and does not constitute a limitation on the computing device to which the present invention is applied. A specific computing device may include more or fewer components, or combine certain components, or have different component arrangements.

[0096] Example 4 This embodiment provides a computer program product corresponding to the real-time diagnosis method for wind turbine blade damage provided in Embodiment 1. The computer program product may include a computer-readable storage medium on which computer-readable program instructions for executing the real-time diagnosis method for wind turbine blade damage described in Embodiment 1 are loaded.

[0097] A computer-readable storage medium can be a tangible device that holds and stores instructions for use by an instruction execution device. A computer-readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof.

[0098] The computer-readable storage medium provided in the above embodiments has a similar implementation principle and technical effect to the above method embodiments, and will not be described again here.

[0099] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by a computer program. These computer programs 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 instructions executable by 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 device that provides the functions specified in one or more boxes.

[0100] These computer programs 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.

[0101] These computer programs 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.

[0102] The above embodiments are only used to illustrate the present invention. The structure, connection method and manufacturing process of each component can be varied. All equivalent transformations and improvements made on the basis of the technical solution of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A method for real-time diagnosis of damage to wind turbine blades, characterized in that, include: Simultaneously sample the vibration and stress points of the three blades of the wind turbine generator under test to obtain the vibration and stress signals of each measuring point of the three blades of the wind turbine generator under test. The vibration and stress signals of each measuring point on the three blades of the wind turbine generator under test are preprocessed and initially phase compensated. The vibration and stress signals of each measuring point on each of the three blades of the wind turbine generator under test are processed in a secondary manner after initial phase compensation to obtain vibration and stress signals of each measuring point on each blade at equal rotation angle intervals. Based on the vibration and stress signals of each measuring point on each blade at equal rotation angle intervals, the vibration-stress coherence coefficient of each blade of the wind turbine generator under test, as well as the Pearson correlation coefficient, frequency domain correlation, and dispersion of the signals of measuring points at the same position on each blade are calculated. Based on the calculated vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation, and dispersion, the health score or damage probability of each blade of the wind turbine generator under test is calculated. When the calculated health score or damage probability is lower than the preset health score threshold or exceeds the preset damage probability threshold, the corresponding blade is damaged and an alarm is triggered.

2. The method for real-time diagnosis of wind turbine blade damage as described in claim 1, characterized in that, The method of synchronously sampling vibration and stress measurement points of the three blades of the wind turbine generator under test to obtain vibration and stress signals at each measurement point of the three blades of the wind turbine generator under test also includes: The system calculates various characteristic parameters of vibration and stress signals at each measuring point of the three blades of the wind turbine generator under test. When the calculated characteristic parameters exceed the preset alarm threshold, an alarm signal is sent to achieve the first level of over-limit alarm.

3. The method for real-time diagnosis of wind turbine blade damage as described in claim 1, characterized in that, The preprocessing and initial phase compensation of the vibration and stress signals at each measuring point of the three blades of the wind turbine generator under test includes: The vibration and stress signals of each measuring point on the three blades of the wind turbine generator under test are preprocessed. The reference blade of the wind turbine generator is taken as the first blade, and the other two blades of the wind turbine generator are defined clockwise as the second blade and the third blade. Using the phase of the pre-processed vibration and stress signals at each measuring point of the first blade as the reference 0°, and the same type of signals of the second and third blades lagging by 120° and 240° respectively, the actual phase difference between the vibration and stress signals at each measuring point of the second and third blades and the vibration and stress signals at each measuring point of the reference blade is calculated. Based on the calculated actual phase difference, a time-shift correction method is used to compensate the phase of the vibration and stress signals at each measuring point of the second and third blades to be consistent with the theoretical mechanical installation angle.

4. The method for real-time diagnosis of wind turbine blade damage as described in claim 1, characterized in that, The vibration and stress signals of each measuring point on each of the three blades of the wind turbine under test are subjected to secondary processing to obtain vibration and stress signals at equal rotation angle intervals at each measuring point on each blade, including: Short-time Fourier transforms were performed on the vibration and stress signals at each measuring point of the three blades after initial phase compensation to obtain the corresponding time spectrum; Extract the principal rotation frequency of the spectrum at each time point, and obtain the curve of the corresponding principal rotation frequency component changing with time, as the fundamental frequency variation spectrum; By integrating the fundamental frequency variation spectrum over time, the instantaneous rotation phase of each blade can be obtained. Based on the angle spectrum of each blade, the vibration and stress signals of the three blades sampled at equal time intervals are resampled into vibration and stress signals with equal rotation angle intervals through spline interpolation.

5. The method for real-time diagnosis of wind turbine blade damage as described in claim 3, characterized in that, The calculation of the dispersion of the measurement points at the same position on the three blades of the wind turbine generator under test includes: The signals from the same measurement points on the three blades of the wind turbine generator under test are subjected to continuous wavelet transform, and the Morlet wavelet basis function is used to obtain the corresponding time-frequency distribution. Calculate the wavelet cross spectrum of the signal at the same position of every two blades of the wind turbine generator under test; Dividing the squared amplitude of the calculated wavelet cross spectrum by the product of the two smoothed self-spectrums yields three wavelet coherence spectrum matrices. Calculate the sample entropy of each wavelet coherence spectrum matrix to measure the complexity of the coherent mode; The dispersion of the three blades is calculated based on the sample entropy of each wavelet coherence spectrum matrix.

6. The method for real-time diagnosis of wind turbine blade damage as described in claim 1, characterized in that, The calculated vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation, and dispersion are used to calculate the health score or damage probability of each blade of the wind turbine generator under test. When the calculated health score or damage probability is lower than a preset health score threshold or exceeds a preset damage probability threshold, the corresponding blade is damaged, and an alarm is triggered, including: The calculated vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation, and dispersion are input into a pre-built deep learning model to obtain the health score or damage probability of each blade of the wind turbine generator under test. When the calculated health score or damage probability is lower than the preset health score threshold or exceeds the preset damage probability threshold, the corresponding blade is damaged and an alarm is triggered.

7. The method for real-time diagnosis of wind turbine blade damage as described in claim 6, characterized in that, The deep learning model includes: The input layer is used to input the calculated feature values, including the vibration-stress coherence coefficient, Pearson correlation coefficient, frequency domain correlation and dispersion, and after standardization, they are mapped to the [0, 1] interval; The attention fusion layer is used to enable the model to automatically learn the importance of different feature dimensions in health status determination and perform weighted fusion to enhance the contribution of discriminative features. The triplet loss function is used to construct triplets and train the model so that the distance between the feature vector Anchor and the feature vector Positive in the feature space is much smaller than the distance between the feature vector Anchor and the feature vector Negative. This allows the model to better learn the difference between healthy and damaged states. Here, Anchor is the feature vector of a healthy sample, Positive is the feature vector of another healthy sample, and Negative is the feature vector of a damaged sample. The output layer is used to output a health score or damage probability. The alarm layer is used to issue an alarm when the output health score or damage probability is lower than a preset health score threshold or exceeds a preset damage probability threshold.

8. A real-time diagnostic system for wind turbine blade damage, characterized in that, It includes a multi-dimensional sensing end, an edge computing acquisition device, a network transmission system, and a cloud system. The multi-dimensional sensing end includes a dual-axis fiber optic temperature and vibration integrated sensor and a fiber optic strain sensor. Each blade of the wind turbine generator set is equipped with a dual-axis fiber optic temperature and vibration integrated sensor at the blade root, 1 / 3L from the blade root, and the blade tip, where L is the blade length; each blade of the wind turbine generator set is equipped with four fiber optic strain sensors in the main beam cap area, and two fiber optic strain sensors in the leading and trailing edge areas of each blade of the wind turbine generator set. The dual-axis fiber optic temperature and vibration integrated sensor is used to collect temperature and vibration signals at corresponding positions of each blade of the wind turbine generator in real time. The fiber optic strain sensor is used to collect stress signals at corresponding positions of each blade of the wind turbine generator in real time. The edge computing data acquisition unit is used to perform analog-to-digital conversion, preprocessing, and compression of vibration and stress signals from each measuring point on the three blades of the wind turbine generator under test, and then send them to the network transmission system. The network transmission system is used to transmit the data sent by the edge computing collector to the designated server of the booster station via the wireless module; and to transmit the data sent by the edge computing collectors of different wind turbine generators to other wireless modules in a step-by-step manner via the wireless module, and then to the designated server of the corresponding booster station. The cloud system is used to receive data uploaded by the designated server of the booster station, calculate the vibration-stress coherence coefficient of each blade of the wind turbine generator under test, as well as the Pearson correlation coefficient, frequency domain correlation, and dispersion of the signals at the same location of each blade, and then calculate the health score or damage probability of each blade of the wind turbine generator under test. When the calculated health score or damage probability is lower than the preset health score threshold or exceeds the preset damage probability threshold, the corresponding blade is damaged and an alarm is triggered.

9. A processing device, characterized in that, The method includes a computer program, wherein when executed by a processing device, the computer program is used to implement the steps corresponding to the real-time diagnosis method for wind turbine blade damage as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it is used to implement the steps corresponding to the real-time diagnosis method for wind turbine blade damage according to any one of claims 1-7.

Images