CN-122014527-A - Real-time diagnosis method, system, processing equipment and storage medium for blade damage of wind generating set

CN122014527ACN 122014527 ACN122014527 ACN 122014527ACN-122014527-A

Abstract

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

Inventors

WANG WENLIANG
PAN WENBIAO
ZHANG YUNXIN
XU YULONG
DONG XIAODONG

Assignees

中节能风力发电股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260210

Claims (10)

1. A real-time diagnosis method for blade damage of a wind generating set is characterized by comprising the following steps: Synchronously sampling vibration measuring points and stress measuring points of three blades of the wind generating set to be tested, and obtaining vibration signals and stress signals of all measuring points of the three blades of the wind generating set to be tested; Preprocessing and initial phase compensation are carried out on vibration signals and stress signals of all measuring points of three blades of the wind generating set to be measured; performing secondary processing on the vibration signals and stress signals after initial phase compensation of each measuring point of three blades of the wind generating set to be tested to obtain vibration signals and stress signals of each measuring point of each blade at equal rotation angle intervals; based on vibration signals and stress signals of rotation angle intervals of each measuring point of each blade, calculating vibration-stress coherence coefficients of each blade of the wind generating set to be measured and Pearson correlation coefficients, frequency domain correlations and dispersion of measuring point signals of the same position of each blade; And calculating the health score or damage probability of each blade of the wind generating set to be tested based on the calculated vibration-stress coherence coefficient, pearson correlation coefficient, frequency domain correlation and dispersion, and giving an alarm when the calculated health score or damage probability is lower than or exceeds a preset health score threshold value.
2. The method for diagnosing blade damage of a wind generating set according to claim 1, wherein the step of synchronously sampling vibration measuring points and stress measuring points of three blades of the wind generating set to be tested, and obtaining vibration signals and stress signals of each measuring point of the three blades of the wind generating set to be tested, further comprises: and calculating various characteristic parameters of vibration signals and stress signals of each measuring point of three blades of the wind generating set to be measured, and sending an alarm signal when the calculated various characteristic parameters exceed a preset alarm threshold value to realize the overrun alarm of the first stage.
3. The method for diagnosing blade damage of a wind generating set in real time according to claim 1, wherein the preprocessing and initial phase compensation of vibration signals and stress signals of each measuring point of three blades of the wind generating set to be tested comprise: Preprocessing vibration signals and stress signals of all measuring points of three blades of the wind generating set to be measured; The method comprises the steps of taking a reference blade of a wind generating set as a first blade, and defining two other blades of the wind generating set clockwise as a second blade and a third blade; Taking the phase of the vibration signal and the stress signal after the pretreatment of each measuring point of the first blade as a reference 0 degree, respectively lagging the same type signals of the second blade and the third blade by 120 degrees and 240 degrees, and calculating the actual phase difference of the vibration signal and the stress signal of each measuring point of the second blade and the third blade relative to the vibration signal and the stress signal of each measuring point of the reference blade; Based on the calculated actual phase difference, a time shift correction method is adopted to compensate the phases of vibration signals and stress signals of each measuring point of the second blade and the third blade respectively until the phases are consistent with the theoretical mechanical installation angle.
4. The method for diagnosing blade damage of a wind generating set according to claim 1, wherein the secondary processing is performed on the vibration signal and the stress signal after initial phase compensation of each measuring point of three blades of the wind generating set to be tested, to obtain vibration signals and stress signals of rotation angle intervals of each measuring point of each blade, and the method comprises the following steps: Performing short-time Fourier transform on the vibration signals and stress signals after initial phase compensation of each measuring point of the three blades to obtain corresponding time frequency spectrums; extracting the main rotation frequency of each time spectrum, and obtaining a curve corresponding to the change of the main rotation frequency component along with time as a fundamental frequency change spectrum; Performing time integration on each fundamental frequency variation spectrum to obtain the instantaneous rotation phase of each blade; According to the angle spectrum of each blade, the vibration signals and stress signals of the three blades sampled at equal time intervals are resampled into vibration signals and stress signals at equal rotation angle intervals through spline interpolation.
5. The method for diagnosing blade damage of a wind generating set in real time according to claim 3, wherein calculating the dispersion of the measurement point signals of the same positions of three blades of the wind generating set to be tested comprises: Respectively carrying out continuous wavelet transformation on signals of measuring points at the same positions of three blades of the wind generating set to be measured, and adopting Morlet wavelet basis functions to obtain corresponding time-frequency distribution; calculating wavelet cross common of signals of measuring points at the same position of every two blades of the wind generating set to be measured; Dividing the square of the calculated amplitude of the wavelet cross spectrum by the product of the two smoothed self-spectrums to obtain three wavelet coherent spectrum matrixes; Calculating the sample entropy of each wavelet coherent spectrum matrix to measure the complexity of the coherent mode; based on the sample entropy of each wavelet coherent spectrum matrix, the dispersion of three blades is calculated.
6. The method for diagnosing blade damage of wind generating set according to claim 1, wherein the calculating the health score or damage probability of each blade of the wind generating set to be tested based on the calculated vibration-stress coherence coefficient, pearson correlation coefficient, frequency domain correlation and dispersion, and when the calculated health score or damage probability is lower than or exceeds a preset health score threshold value, the corresponding blade is damaged, and the method comprises the steps of: Inputting the calculated vibration-stress coherence coefficient, pearson correlation coefficient, frequency domain correlation and dispersion into a pre-constructed deep learning model to obtain health scores or damage probabilities of all blades of the wind generating set to be tested; And when the calculated health score or damage probability is lower than or exceeds a preset health score threshold, the corresponding blade is damaged, and an alarm is given.
7. The method for diagnosing blade damage of a wind generating set in real time as recited in claim 6, wherein the deep learning model comprises: the input layer is used for inputting the calculated characteristic values, including vibration-stress coherence coefficient, pearson correlation coefficient, frequency domain correlation and dispersion, and mapping the normalized values to the [0,1] interval; The attention fusion layer is used for enabling the model to automatically learn the importance of different feature dimensions to the health state judgment, carrying out weighted fusion and enhancing the contribution of the discriminant features; The triplet loss function is used for constructing a triplet, and the distance between the feature vector Anchor and the feature vector Positive in the feature space is far smaller than the distance between the feature vector Anchor and the feature vector Negative through training, so that the model can learn the difference between health and damage states better, wherein the Anchor is the feature vector of one health sample, the Positive is the feature vector of another health sample, and the Negative is the feature vector of one damage sample; An output layer for outputting a health score or damage probability; and the alarm layer is used for alarming when the output health score or damage probability is lower than or exceeds a preset health score threshold value.
8. The real-time diagnosis system for the damage of the wind generating set blade is characterized by comprising a multi-dimensional sensing end, an edge calculation collector, a network transmission system and a cloud system, wherein the multi-dimensional sensing end comprises a double-shaft optical fiber temperature vibration integrated sensor and an optical fiber strain sensor; The blade root, the position 1/3L away from the blade root and the blade tip of each blade of the wind generating set are provided with one double-shaft optical fiber temperature vibration integrated sensor, wherein L is the length of the blade; the double-shaft optical fiber temperature and vibration integrated sensor is used for collecting temperature signals and vibration signals of corresponding positions of each blade of the wind generating set in real time; The optical fiber strain sensor is used for collecting stress signals of the corresponding positions of each blade of the wind generating set in real time; The edge calculation collector is used for carrying out analog-to-digital conversion, pretreatment and compression packaging on vibration signals and stress signals of each measuring point of three blades of the wind generating set to be measured and then sending the vibration signals and the stress signals to the network transmission system; The network transmission system is used for transmitting the data transmitted by the edge calculation collectors to the appointed server of the booster station through the wireless module, and transmitting the data transmitted by the edge calculation collectors of different wind generating sets to other wireless modules step by step through the wireless module and then to the appointed server of the corresponding booster station; The cloud system is used for receiving data uploaded by a designated server of the booster station, calculating the vibration-stress coherence coefficient of each blade of the wind generating set to be tested and the Pearson correlation coefficient, the frequency domain correlation and the dispersion of measuring point signals at the same position of each blade, further calculating the health score or damage probability of each blade of the wind generating set to be tested, and giving an alarm when the calculated health score or damage probability is lower than a preset health score threshold or exceeds a preset damage probability threshold.
9. A processing device, characterized by comprising a computer program, wherein the computer program, when executed by the processing device, is adapted to carry out the steps corresponding to the real-time diagnosis method of blade damage of a wind park according to any one of claims 1-7.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program, wherein the computer program, when executed by a processor, is adapted to implement the steps corresponding to the method for real-time diagnosis of blade damage of a wind turbine generator set according to any of claims 1-7.

Description

Real-time diagnosis method, system, processing equipment and storage medium for blade damage of wind generating set Technical Field The invention relates to the field of wind generating set state monitoring, in particular to a method, a system, processing equipment and a storage medium for diagnosing damage of a wind generating set blade in real time. Background The wind generating set blade is used as a core component for energy conversion, is in a complex and changeable working environment for a long time, faces the combined action of various mechanical and environmental loads, and the structural health state of the blade directly influences the power generation efficiency and the operation safety. At present, the blade monitoring method mainly has the following technical bottlenecks of 1) single-mode monitoring. Vibration monitoring traditional vibration monitoring is that a double-axial vibration sensor is arranged at a specific position to analyze the vibration condition of a blade, and the integral mode of the blade can be detected, but early micro-damage of the blade cannot be diagnosed, and environmental load fluctuation and real structural damage cannot be effectively distinguished. Stress monitoring, namely, a stress sensor such as a fiber grating and the like can capture local strain, but the prior art does not solve the problem of signal conversion between a dynamic blade and a static tower, and the temperature drift causes measurement errors of +/-0.5 mu epsilon/DEGC. 2) A drawback of the signal processing method. The current common signal processing methods are divided into two types, the first type is based on a fault mechanism, the condition of the blade is judged by setting an alarm threshold, the method does not effectively distinguish different working conditions, a large amount of manual compounding is needed, and the accuracy of a diagnosis result is low. The second type adopts a machine learning mode to process vibration signals, but depends on a large amount of labeling data for training, and has insufficient generalization capability for a model in a small sample scene of an actual wind power plant. In addition, the current blade monitoring method mainly comprises the following steps of 1) insufficient detection sensitivity. The minimum identifiable crack size in the prior art is 5cm, and early warning requirements of early micro-damage (< 1 cm) cannot be met. The detection rate of delamination defects of the composite material is less than 60 percent, and the positioning error exceeds +/-1.5 m. 2) The detection accuracy is poor. The false alarm of the vibration monitoring method reaches 35%, the false alarm rate reaches 25%, and the false alarm rate can reach 40% for microcracks or early damage, so that the actual use requirements cannot be met. Disclosure of Invention Aiming at the problems, the invention aims to provide a real-time diagnosis method, a real-time diagnosis system, a real-time diagnosis processing device and a real-time diagnosis storage medium for blade damage of a wind generating set, wherein the diagnosis result has high accuracy, high detection sensitivity and high detection accuracy. In order to achieve the purpose, the invention adopts the following technical scheme that in the first aspect, the invention provides a real-time diagnosis method for blade damage of a wind generating set, which comprises the following steps: Synchronously sampling vibration measuring points and stress measuring points of three blades of the wind generating set to be tested, and obtaining vibration signals and stress signals of all measuring points of the three blades of the wind generating set to be tested; Preprocessing and initial phase compensation are carried out on vibration signals and stress signals of all measuring points of three blades of the wind generating set to be measured; performing secondary processing on the vibration signals and stress signals after initial phase compensation of each measuring point of three blades of the wind generating set to be tested to obtain vibration signals and stress signals of each measuring point of each blade at equal rotation angle intervals; based on vibration signals and stress signals of rotation angle intervals of each measuring point of each blade, calculating vibration-stress coherence coefficients of each blade of the wind generating set to be measured and Pearson correlation coefficients, frequency domain correlations and dispersion of measuring point signals of the same position of each blade; And calculating the health score or damage probability of each blade of the wind generating set to be tested based on the calculated vibration-stress coherence coefficient, pearson correlation coefficient, frequency domain correlation and dispersion, and giving an alarm when the calculated health score or damage probability is lower than or exceeds a preset health score threshold value. Further, the vibration measuring point and the st