Method, device and system for monitoring icing state of blade of wind driven generator

Abstract

The invention discloses a method, a device and a system for monitoring the icing state of a wind driven generator blade, and belongs to the technical field of wind driven generator blade monitoring. The method is based on the acoustic emission technology, acoustic emission sensors are arranged on the inner wall of a blade of the wind driven generator, acoustic emission signals of the blade in the operation process are collected, the signals are processed through a symplectic geometric mode decomposition method, and state recognition is conducted through a deep learning model. The method overcomes the limitations of response lag, susceptibility to environmental interference, influence on pneumatic performance and the like existing in a traditional icing monitoring method, and can realize accurate identification and timely early warning of the icing state of the blade. Experimental verification shows that the recognition accuracy rates of the method on the normal operation state, the icing state and the raining state of the blade of the wind driven generator all exceed 99.5%, and the method has good engineering application value.

Description

Technical Field

[0001] The present invention relates to the technical field of wind turbine blade monitoring, and in particular to a method, device and system for monitoring the icing status of wind turbine blades. Background Art

[0002] As a clean and renewable energy source, wind power generation plays an important role in the transformation of energy structure. As wind farms expand to plateaus, mountains, oceans and other regions, the problem of wind turbine blade icing has become increasingly prominent and is no longer limited to traditional cold regions. Blade icing will change the aerodynamic shape of the blades and seriously reduce the power generation efficiency; at the same time, due to uneven icing, it will increase the unbalanced load, causing accelerated fatigue of wind turbine components, and in severe cases may cause the wind turbine to collapse; the falling ice may endanger the safety of surrounding personnel and property. According to statistics, blade icing causes billions of dollars in economic losses to the global wind power industry each year, and has become a key issue that needs to be urgently addressed in wind power operation and maintenance.

[0003] At present, wind turbine blade ice monitoring mainly adopts visual observation method, power curve method, vibration analysis method, ultrasonic attenuation method and capacitive sensor method. The visual observation method directly observes the ice coverage on the blade surface through manual or camera observation. Although it can achieve higher recognition accuracy when combined with deep learning technology, it is easily affected by factors such as light and weather, especially at night or when visibility is poor, it is difficult to monitor effectively. The power curve method is based on the power changes before and after the blade is frozen and combined with SCADA data for judgment. Although a more reliable recognition result can be obtained through the optimization algorithm, it is easily disturbed by wind conditions, and in the initial stage of ice coverage, it may be misjudged due to the increase in power caused by the increase in surface roughness. The vibration analysis method identifies the ice coverage state by measuring the change in vibration signal. It has high sensitivity, but it is easily affected by wind conditions and unit operation, and the arrangement of sensors on the blade surface will affect the aerodynamic performance. The ultrasonic attenuation method uses the change in sound wave propagation energy to judge the ice coverage state. It can effectively detect ice coverage at different locations, but acoustic devices need to be arranged at both ends of the monitoring position, which also affects the performance of the blade. The capacitive sensor method can monitor multiple parameters such as the presence, type, and thickness of ice coverage, but it is easily affected by temperature and humidity, and there are problems such as poor long-term reliability and increased risk of lightning strikes. In practical applications, these methods all have limitations to varying degrees, such as response lag, poor anti-interference ability, or impact on blade performance.

[0004] In summary, although the existing wind turbine blade ice monitoring methods can reflect the ice condition to a certain extent, they generally have problems such as delayed ice recognition response, low accuracy, and susceptibility to environmental interference. At the same time, the reliability of sensors in long-term use is difficult to guarantee, and the layout often affects the aerodynamic performance of the blades. These problems seriously restrict the actual application effect of ice monitoring systems in wind farms. Summary of the invention The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and to provide a method, device and system for monitoring the icing status of wind turbine blades, so as to solve the problems in the prior art of wind turbine blade icing monitoring, such as delayed recognition response, low accuracy and susceptibility to environmental interference during the icing monitoring process.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: A method for monitoring the ice coating status of wind turbine blades comprises the following steps: S1, collecting acoustic emission signals; the acoustic emission signals are generated during the operation of the blades, and the acoustic emission sensor collects the acoustic emission signals of the fan blades; S2, perform symplectic geometric mode decomposition on the acoustic emission signal to obtain the main symplectic geometric components; S3, reconstruction of the acoustic emission signal through the main symplectic geometric components; S4, judging the blade state through a deep neural network based on the reconstructed acoustic emission signal; the blade state includes a normal operation state, an ice-covered state and a rainy state.

[0006] A further improvement of the present invention is: Preferably, in S1, the acoustic emission signal is the stress change caused by ice covering the wind turbine blades during the process of ice accumulation, ice stress and ice melting.

[0007] Preferably, the main symplectic geometric components are obtained by analyzing test signals, and the analysis process is: Perform symplectic geometric modal decomposition on the acoustic emission signal obtained from the experiment to obtain several symplectic geometric components, and establish scatter plots, signal superposition plots and phase space plots based on the symplectic geometric components and the original acoustic emission signal; judge the correlation between each symplectic geometric component and the original acoustic emission signal through the scatter plots, signal superposition plots and phase space plots to obtain the main symplectic geometric components; The acoustic emission signals are reconstructed through the symplectic geometric components using the optimal reconstruction scheme.

[0008] Preferably, in the step, the correlation between each symplectic geometric component and the original acoustic emission signal is determined by a scatter plot, and if the correlation is greater than a set threshold value of one, the symplectic geometric component is the main symplectic geometric component.

[0009] Preferably, in the step, the retention rate of each symplectic geometric component to the original signal amplitude feature is determined by the signal superposition diagram. If the retention rate is greater than a set threshold value of two, the symplectic geometric component is the main symplectic geometric component.

[0010] Preferably, in the step, the retention rate of each symplectic geometric component to the dynamic characteristics of the original signal is determined by the phase space diagram. If the retention rate is greater than a set threshold value of three, the symplectic geometric component is the main symplectic geometric component.

[0011] Preferably, in S4, the deep neural network model can be any one of a convolutional neural network, a long short-term memory network, a gated recurrent unit or a Transformer neural network.

[0012] Preferably, the deep neural network model is trained with training set data and then tested with test set data; the training set data and test set data are derived from wind turbine blade tests.

[0013] A device for monitoring the ice coating status of wind turbine blades, comprising: A collection unit, used for collecting acoustic emission signals; the acoustic emission signals are generated during the operation of the blades, and the acoustic emission sensor collects the acoustic emission signals of the fan blades; A decomposition unit, used for performing symplectic geometric mode decomposition on the acoustic emission signal to obtain main symplectic geometric components; A reconstruction unit, used for reconstructing the acoustic emission signal through the main symplectic geometric components; A judgment unit is used to judge the blade state through a deep neural network based on the reconstructed acoustic emission signal; the blade state includes a normal operation state, an ice-covered state and a rainy state.

[0014] A wind turbine blade icing status monitoring system for implementing the above monitoring method comprises an acoustic emission sensor and an information processor; the acoustic emission sensor is arranged in the wind turbine blade, and the acoustic emission sensor and the information processor are connected by electrical signals.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses a method for monitoring the icing status of wind turbine blades. The method is based on acoustic emission technology. An acoustic emission sensor is arranged on the inner wall of the wind turbine blade to collect the acoustic emission signal of the blade during operation. The signal is processed using the symplectic geometry mode decomposition method, and a deep learning model is used for state recognition. The method overcomes the limitations of traditional icing monitoring methods, such as response lag, susceptibility to environmental interference, and influence on aerodynamic performance, and can achieve accurate recognition of the blade icing status and timely warning. Experimental verification shows that the method has an accuracy rate of over 99.5% for the three states of normal operation, icing and raining of wind turbine blades, and has good engineering application value.

[0016] The present invention also discloses a wind turbine blade icing status monitoring system, which includes an acoustic emission sensor arranged on the inner wall of the wind turbine blade, and a signal processor connected to the electrical signal of the acoustic emission sensor. The data collected or identified by the acoustic emission sensor is combined with signal processing and deep learning technology to accurately identify the blade icing status, thereby improving the operational safety and economic benefits of the wind farm. The system arranges the acoustic emission sensor inside the wind turbine blade. Based on the rapid response characteristics of the acoustic emission signal, it can detect early icing in time, avoid affecting the aerodynamic performance of the blade, effectively protect the acoustic emission sensor, and avoid the impact of the harsh environment on the equipment. The system has a simple structure, low maintenance cost, and is easy to implement in engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is the principle diagram of wind turbine blade ice acoustic emission signal monitoring; Among them: 1. Wind turbine blades; 2. Acoustic emission sensor; 3. Acoustic emission signal; 4. Ice layer; F1, wind load; F2, blade support force; F3, centrifugal force; F4, adhesion force; F5, gravity; Figure 2 is a flow chart of the monitoring method of the present invention; Figure 3 It is a flow chart of the monitoring method in a specific embodiment of the present invention; Figure 4 This is the recognition effect diagram of the deep neural network model; Figure 5 This is a comparative analysis diagram of the SGC1 component in the ice-covered state and the original signal; Among them, (a) is a scatter plot; (b) is a signal superposition plot; (c) is a phase space plot; Figure 6 This is a comparative analysis diagram of the SGC2 component in the ice-covered state and the original signal; Among them, (a) is a scatter plot; (b) is a signal overlay plot; (c) is a phase space plot; Figure 7 This is a comparative analysis diagram of the SGC3 component in ice-covered state and the original signal; Among them, (a) is a scatter plot; (b) is a signal superposition plot; (c) is a phase space plot; Figure 8 This is a comparative analysis diagram of the SGC4 component in ice-covered state and the original signal; Among them, (a) is a scatter plot; (b) is a signal superposition plot; (c) is a phase space plot; Fig. 9 This is a comparison and analysis diagram of the reconstruction of SGC1+SGC2 components in the ice-covered state and the original signal; Among them, (a) is a scatter plot; (b) is a signal superposition plot; (c) is a phase space plot; Fig.10 This is a comparison and analysis diagram of the reconstruction of SGC1+SGC2 components in the raining state and the original signal; Among them, (a) is a scatter plot; (b) is a signal overlay plot; (c) is a phase space plot. DETAILED DESCRIPTION

[0018] The present invention is further described in detail below in conjunction with the accompanying drawings: In the description of the present invention, it should be noted that the orientations or positional relationships indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. are based on the orientations or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention; the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance; in addition, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection or a detachable connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal communication of two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0019] See also Figure 1 The first aspect of the present invention discloses a system for monitoring the ice coating status of wind turbine blades, including an acoustic emission sensor 2 and an information processor, wherein the acoustic emission sensor 2 is arranged inside the outer wall of the fan blade 1. The outer wall here refers to the wall surface of the fan blade 1 away from the rotating shaft side. Setting the acoustic emission sensor 2 on the side wall outside the fan blade 1 can prevent the acoustic emission sensor 2 from affecting the aerodynamic performance of the fan blade 1, and setting the acoustic emission sensor 2 inside the fan blade 1 can prevent the sensor from being affected by the external environment and improve the service life. The information processor can be a computer or other peripheral industrial equipment. The information processor and the acoustic emission sensor 2 are electrically connected and used to analyze the signal sent back by the acoustic emission sensor 2.

[0020] For the fan blade 1, the parameters that can affect the propagation characteristics of the acoustic emission signal are the interface propagation parameters, including the acoustic impedance of the ice layer, the acoustic impedance of the blade material, and the transmission coefficient. The specific acoustic impedance of the ice layer is about 3.5×10 6 kg / (m²·s); the acoustic impedance of the blade material is about 6.0×10 6kg / (m²·s), the transmission coefficient is close to 1, and the above interface propagation parameters can ensure the effective transmission of the signal; and the fan blade itself has the characteristics of low damping, low dispersion effect and small energy attenuation of composite materials. The above interface propagation parameters and the material properties of the fan blade itself enable the acoustic emission sensor 2 to sensitively collect changes in the ice state on the fan blade.

[0021] Once the wind turbine blade 1 has an ice layer 4 , the ice layer 4 will generate an acoustic emission signal source 3 due to various situations during operation, which will then be received by the acoustic emission sensor 2 to monitor the ice status of the wind turbine blade 1 .

[0022] The above-mentioned acoustic emission signal source 3 generates the following situations: (1) The ice accumulation process, including the stress waves generated when supercooled water droplets hit the blades, the structural rearrangement during the formation and growth of ice crystals, and the stress changes at the interface between the ice layer and the blades.

[0023] (2) Changes in the stress on the ice layer, including: periodic stress changes caused by blade rotation, ice layer deformation caused by wind loads, and crack formation under the action of centrifugal force.

[0024] (3) Ice melting process, including thermal stress caused by temperature gradient, structural changes caused by reduced ice strength, and energy release during ice shedding.

[0025] It should be understood that the number and location of the acoustic emission sensors 2 can be adjusted according to actual conditions.

[0026] In some embodiments of the present invention, the acoustic emission sensor 2 is preferably arranged on the inner wall of the blade tip at about 1 / 3 of the blade length. This position is the area where ice is most serious and melts last, and is subject to large centrifugal force, so the ice acoustic emission signal is more significant. By monitoring the ice state here, the ice state of the wind turbine blade 1 can be quickly and accurately obtained.

[0027] In some embodiments of the present invention, the acoustic emission sensor 2 is a narrowband acoustic emission sensor with a frequency range of 10 KHz-400 KHz, including a data acquisition unit with a sampling rate of 2 MHz and a preamplifier with a gain of 40 dB.

[0028] Among them, after the data acquisition unit receives the signal generated by the above-mentioned acoustic emission signal source 3, it converts and amplifies the collected signal in the preamplifier and outputs the signal to the signal processor. The signal processor is a hardware device used in the wind turbine blade icing status monitoring device.

[0029] The monitoring principle of the present invention is as follows Figure 1As shown. During the operation of the wind turbine blade, the ice layer is subjected to multiple effects such as centrifugal force (F3), wind load (F1), gravity (F5) and adhesion force (F4), generating acoustic emission signals. Since the acoustic impedance of ice and blade material (such as glass fiber / epoxy resin composite material) is close, the acoustic wave transmission coefficient is close to 1, so that the acoustic emission signal can be effectively transmitted to the acoustic emission sensor 2 set inside the wind turbine blade 2.

[0030] See also Figure 2 The second aspect of the present invention discloses a method for monitoring the ice coating status of wind turbine blades, comprising the following steps: S1, collecting acoustic emission signals; the acoustic emission signals 3 are generated during the operation of the blades, and the acoustic emission sensor 2 collects the acoustic emission signals of the fan blades 1; S2, performing symplectic geometric mode decomposition (SGMD) processing on the collected acoustic emission signal, decomposing the original signal into multiple symplectic geometric components (SGC), and selecting the symplectic geometric components containing the main characteristic information of the signal as the main symplectic geometric components; S3, reconstruction of the acoustic emission signal through the main symplectic geometric components; S4, judging the blade state through a deep neural network based on the reconstructed acoustic emission signal; the blade state includes a normal operation state, an ice-covered state and a rainy state.

[0031] In some embodiments of the present invention, in S2, the original signal is decomposed into n symplectic geometric components (SGC1-SGC n ) and a residual term.

[0032] In some embodiments of the present invention, the main symplectic geometric components are obtained by analyzing experimental data, and the analysis process is: (1) Perform symplectic geometric modal decomposition on the experimental acoustic emission signal to obtain several symplectic geometric components, reconstruct the acoustic emission signal through each symplectic geometric component to obtain the symplectic geometric component, and establish a scatter plot, signal superposition plot and phase space plot based on the symplectic geometric components and the original acoustic emission signal; judge the correlation between each symplectic geometric component and the original acoustic emission signal through the scatter plot, signal superposition plot and phase space plot to obtain the main symplectic geometric components; (2) Reconstruct the acoustic emission signal based on the main symplectic geometric components.

[0033] Specifically, in step (1), the scatter plot is used to evaluate the correlation between the symplectic geometric component and the acoustic emission signal; the specific process is to determine the correlation between each symplectic geometric component and the original acoustic emission signal through the scatter plot, and if the correlation is greater than a set threshold value one, the symplectic geometric component is the main symplectic geometric component.

[0034] Specifically, in step (1), the signal overlay diagram is used to evaluate and analyze the retention rate of each symplectic geometric component to the original signal amplitude feature. If the retention rate is greater than a set threshold value of two, the symplectic geometric component is the main symplectic geometric component.

[0035] Specifically, in step (1), the phase space diagram is used to evaluate and analyze the retention rate of each symplectic geometric component to the dynamic characteristics of the original signal to verify the retention of the dynamic characteristics. If the retention rate is greater than a set threshold value of three, the symplectic geometric component is the main symplectic geometric component.

[0036] It should be noted that the above-mentioned threshold value 1, threshold value 2 and threshold value 3 can be adjusted according to the actual accuracy requirements of the method in the application process. The higher the accuracy, the higher the number of threshold values ​​required, and the higher the repeatability requirements of the symplectic geometric component and the original acoustic emission signal.

[0037] It should be noted that when judging the main symplectic geometric components through scatter plots, signal overlay plots and phase space plots, if the three judgment results are different, the main symplectic geometric components need to be determined according to the specific application scenario and research purpose. In some cases, if the similarity of the signal is the most important, the scatter plot is more critical; if the amplitude information of the signal is important for analysis, the signal overlay plot is more important; if the dynamic characteristics of the signal are the main research object, the phase space diagram is more critical. Therefore, it needs to be determined according to the actual situation.

[0038] In some embodiments of the present invention, a deep learning model is used to classify and identify the reconstructed signal to determine the state of the blade. In this step, the deep learning model can be a neural network model such as a convolutional neural network (CNN), a long short-term memory network (LSTM), a gated recurrent unit (GRU) or a Transformer neural network; it should be understood that the deep learning model is not limited to the above four models and can be adjusted according to actual conditions.

[0039] In some embodiments of the present invention, the deep learning model training set and test set are constructed using laboratory data.

[0040] A third aspect of the present invention discloses a device for monitoring the ice coating status of wind turbine blades, comprising: A collection unit, used for collecting acoustic emission signals; the acoustic emission signals are generated during the operation of the blades, and the acoustic emission sensor collects the acoustic emission signals of the fan blades; A decomposition unit, used for performing symplectic geometric mode decomposition on the acoustic emission signal to obtain main symplectic geometric components; A reconstruction unit, used for reconstructing the acoustic emission signal through the main symplectic geometric components; A judgment unit is used to judge the blade state through a deep neural network based on the reconstructed acoustic emission signal; the blade state includes a normal operation state, an ice-covered state and a rainy state.

[0041] The following is a further description in conjunction with a specific embodiment.

[0042] Example A small fan blade test bench was built in the laboratory to simulate the operating state of the fan blades. The motor speed was controlled by a frequency converter. The three fan blades on the test bench were 1.6m long glass fiber / epoxy composite blades; the speed of the adjustable speed motor was 0-280 rpm. An acoustic emission sensor was set in each simulated fan blade to collect the corresponding acoustic emission signal.

[0043] The experimental working conditions of the small fan blade test bench are: the fan blade speed is: 0-140 rpm, with 8 gears; the environment includes two states: no wind and windy; the blade state includes three states: normal, ice-covered and rainy.

[0044] Through the test bench constructed above, a series of training sets and test sets are obtained to train and build the deep neural network model; the input of the training set and the test set is the acoustic emission signal, and the output is the blade status.

[0045] Four deep neural network models were built: CNN model, LSTM model, GRU model and Transformer model. The parameters of each model are as follows: (1) CNN model -Input layer: 128-dimensional time series -Convolutional layer: 3 feature extraction blocks -Pooling layer: Average pooling with stride 2 -Fully connected layer: 3 types of output (2) LSTM model -Input layer: same as CNN -LSTM layer: 128 hidden units -Fully connected layer: 3 types of output (3) GRU model - Input layer: same as CNN - GRU layer: 512 hidden units - Fully connected layer: 3 types of output (4) Transformer model -Input layer: same as CNN -Encoder: 6-layer attention mechanism -Decoder: Positional encoding -Output layer: 3-class classification.

[0046] The training of the above model takes the Transformer model as an example. Its architecture includes input layer, encoder layer and classification layer. The input layer first normalizes the acquired data and slices the signal into 128-point length signals. The encoding layer encodes the data through the multi-head self-attention mechanism, feedforward neural network and normalization layer, and passes the encoding result to the classification layer. The classification layer classifies the encoding result through the fully connected layer and Softmax activation function to obtain the final classification result. During the training of the Transformer model, the batch size in the training parameters is 64, the learning rate is 0.001, the optimizer is Adam, and the loss function is cross entropy.

[0047] See also Figure 4 This is a visual comparison of the recognition effects of four deep learning models, including the confusion matrix and t-SNE feature visualization results of each model. The confusion matrix shows the recognition results of each model for normal operation, ice-covered state and rainy state. The CNN model has a 100% recognition accuracy for normal state and ice-covered state, with only 1.4% of rainy state samples misjudged as normal state; the recognition accuracy of the LSTM model and GRU model are both over 99.6%, and the recognition of ice-covered state is particularly accurate; the Transformer model performs best, with recognition accuracy of all three states reaching over 99.9%.

[0048] The t-SNE feature visualization diagram shows the feature extraction and classification effects of the four models on samples in the three states. The blue dots represent normal operation, the green dots represent ice-covered states, and the orange dots represent rainy states. From the visualization results, it can be seen that all models have achieved effective separation of samples in the three states and formed a clear clustering structure. In particular, the samples in the ice-covered state (green dots) maintain a clear boundary interval with the other two states, which verifies that the acoustic emission-based monitoring method can reliably identify the ice-covered state of the blades. The Transformer model has the most compact feature distribution and the clearest inter-class boundaries, which corresponds to its optimal recognition accuracy.

[0049] Four neural network models were verified by experimental data, and the average recognition accuracy of the four deep learning models exceeded 99.5%; among them, the Transformer model performed best, with an accuracy of 99.9% and an accuracy standard deviation of less than 0.001, showing good stability; and each round of training only takes 0.09 seconds, with high computational efficiency. Through the comparative analysis of confusion matrix and t-SNE visualization, the excellent performance of each deep learning model in the task of wind turbine blade state recognition is intuitively demonstrated, verifying the effectiveness and reliability of the method of the present invention.

[0050] The above neural network model is applied to a specific ice monitoring system, which includes an acoustic emission sensor and a signal processor; the acoustic emission sensor is a narrowband sensor with a size of Φ19×15mm, which is fixed inside the wind turbine blade by a magnetic clamp. During the signal transmission process, the acoustic coupling is ensured by a special coupling agent. The acoustic emission sensor is equipped with a data acquisition unit with a dynamic range of 70dB, a sampling accuracy of 16 bits, and a storage capacity of 64G.

[0051] The acoustic emission sensor is connected to the information processor through a signal transmission line for electrical signal connection. The signal transmission line is a coaxial cable with an impedance of 75Ω. A shielding layer is provided in the coaxial cable to suppress electromagnetic interference. The coaxial cable is connected to the acoustic emission sensor through a waterproof connector.

[0052] The method for monitoring the ice coating status of wind turbine blades by using the above device comprises the following steps: S1, collect acoustic emission signals, and after the signals are collected, they need to go through signal noise reduction and data normalization in sequence; S2, performing symplectic geometric mode decomposition (SGMD) processing on the collected acoustic emission signals.

[0053] Through symplectic geometric mode decomposition, the original acoustic emission signals in raining and icing states are decomposed into six symplectic geometric components (SGC1-SGC6) and a residual term.

[0054] In order to verify the SGMD decomposition effect and determine the optimal reconstruction scheme, this study takes the acoustic emission signal under ice-covered state as the research object, and uses scatter plots, signal overlay plots and phase space Figure 3 In three dimensions, the characteristic relationship between each SGC component and the original signal was systematically analyzed. Figure 6-Figure 8 This is the demonstration process of the SGC1+SGC2 reconstruction scheme finally determined by the ice cover signal. Fig. 9 It is the performance of the SGC1+SGC2 reconstruction scheme finally determined by the ice cover signal. Fig.10 It is the performance of the SGC1+SGC2 reconstruction scheme finally determined by the rain signal.

[0055] Specifically, the correlation between the reconstructed signal of each SGC component and the original signal was first evaluated using a scatter plot. The results showed that the SGC1 and SGC2 components were highly correlated with the original signal (correlation coefficient R>0.90), while the correlation of subsequent components starting from SGC3 was significantly reduced (correlation coefficient R<0.45); secondly, the degree of preservation of the amplitude characteristics of the original signal by the reconstructed signal was analyzed by signal overlay diagram. It was found that the SGC1 and SGC2 components could maintain more than 85% of the amplitude characteristics of the original signal, while the amplitude retention rates of subsequent components starting from SGC3 were all less than 60%; finally, the phase space diagram was used to verify the degree of preservation of the dynamic characteristics of the original signal by the reconstructed signal. The results showed that the phase space trajectory structures of the SGC1 and SGC2 components were highly similar to the original signal (structural similarity>0.80), while the phase space trajectories of subsequent components starting from SGC3 were significantly deformed. Based on the above analysis results, the present invention selects SGC1+SGC2 as the optimal reconstruction scheme, which not only achieves a correlation coefficient of more than 95% and an amplitude retention rate of more than 90%, but also shows good applicability under different working conditions and can effectively reduce the impact of noise.

[0056] Specifically, the SGC1 component ( Figure 5 ) showed significant correlation characteristics: the scatter plot showed a highly consistent distribution pattern with the reference line (correlation coefficient R=0.92), the signal overlay diagram showed that the reconstructed signal effectively retained the main oscillation characteristics of the original signal (amplitude retention rate>85%), and the phase space reconstruction confirmed the retention of basic dynamic characteristics. Similarly, the SGC2 component ( Figure 6 ) also exhibited excellent signal preservation, which was particularly evident in the phase space reconstruction, where the trajectory structure was highly similar to the original signal (structural similarity index > 0.80). Figure 7 ), the characteristic performance of the reconstructed signal shows obvious degradation. The scatter plot shows that the deviation between the data points and the reference line increases (the correlation coefficient drops to 0.45), the matching accuracy in the signal overlay graph decreases (amplitude retention rate <60%), and the phase space trajectory structure changes significantly. This degradation phenomenon occurs in the SGC4 component ( Figure 8 ), the reconstructed signal can hardly reflect the main features of the original signal (correlation coefficient < 0.30), and the phase space trajectory shows essential morphological differences.

[0057] Based on the above comprehensive analysis, this embodiment selects a combination of two components, SGC1 and SGC2, for signal reconstruction. Fig. 9 Shown and Fig.10As shown in the figure, the SGC1+SGC2 reconstruction scheme achieved the best performance in all evaluation indicators: the data points in the scatter plot were closely distributed around the reference line (correlation coefficient R=0.95), the signal overlay diagram showed excellent waveform matching (amplitude retention rate>90%), and the phase space reconstruction verified the good preservation of the dynamic characteristics, including the morphological and structural characteristics of the trajectory (structural similarity index>0.85).

[0058] Based on the main symplectic geometric components, after superimposing and reconstructing the main symplectic geometric components, more than 90% of the amplitude characteristics of the original signal can be maintained, and the correlation coefficient exceeds 0.95, which effectively reduces the influence of noise.

[0059] The present invention successfully solves many problems existing in traditional monitoring methods through the above technical scheme, realizes accurate identification and timely warning of the icing status of wind turbine blades, and has important engineering application value.

[0060] Through the above detailed description, the technical solution of the present invention has been fully demonstrated. This solution not only solves the limitations of traditional monitoring methods, but also has good engineering practice value. The experimental results show that this method has high reliability, timely response, and is easy to implement, providing an innovative technical solution for wind turbine blade ice monitoring.

[0061] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for monitoring the ice coating status of wind turbine blades, characterized in that: The following steps are involved: S1, collecting an acoustic emission signal; the acoustic emission signal (3) is generated during the operation of the blade, and the acoustic emission sensor (2) collects the acoustic emission signal of the fan blade (1); S2, perform symplectic geometric mode decomposition on the acoustic emission signal to obtain the main symplectic geometric components; S3, reconstruction of the acoustic emission signal through the main symplectic geometric components; S4, judging the blade state through a deep neural network based on the reconstructed acoustic emission signal; the blade state includes a normal operation state, an ice-covered state and a rainy state.

2. A method for monitoring the icing status of wind turbine blades according to claim 1, characterized in that: In S1, the acoustic emission signal (3) is the stress change caused by ice covering the wind turbine blade (1) during the process of ice accumulation, ice stress and ice melting.

3. The method for monitoring the icing status of wind turbine blades according to claim 1, characterized in that: The main symplectic geometric components are obtained through the analysis of test signals, and the analysis process is as follows: (1) Perform symplectic geometric modal decomposition on the acoustic emission signal obtained from the experiment to obtain several symplectic geometric components. Scatter plots, signal overlay plots and phase space plots are established based on the symplectic geometric components and the original acoustic emission signal. The correlation between each symplectic geometric component and the original acoustic emission signal is judged through the scatter plots, signal overlay plots and phase space plots to obtain the main symplectic geometric components. (2) Reconstruct the acoustic emission signal through the optimal reconstruction scheme of symplectic geometric components.

4. A method for monitoring the icing status of wind turbine blades according to claim 3, characterized in that: In step (1), the correlation between each symplectic geometric component and the original acoustic emission signal is determined by a scatter plot. If the correlation is greater than a set threshold value, the symplectic geometric component is the main symplectic geometric component.

5. The method for monitoring the icing status of wind turbine blades according to claim 3, characterized in that: In step (1), the retention rate of each symplectic geometric component to the original signal amplitude feature is determined through the signal superposition diagram. If the retention rate is greater than the set threshold value 2, the symplectic geometric component is the main symplectic geometric component.

6. A method for monitoring the icing status of wind turbine blades according to claim 3, characterized in that: In step (1), the retention rate of each symplectic geometric component to the dynamic characteristics of the original signal is determined through the phase space diagram. If the retention rate is greater than the set threshold value three, the symplectic geometric component is the main symplectic geometric component.

7. The method for monitoring the ice coating status of wind turbine blades according to claim 1, characterized in that: In S4, the deep neural network model can be any one of a convolutional neural network, a long short-term memory network, a gated recurrent unit or a Transformer neural network.

8. A method for monitoring the icing status of wind turbine blades according to claim 7, characterized in that: The deep neural network model is trained with training set data and then tested with test set data; the training set data and test set data are derived from wind turbine blade tests.

9. A device for monitoring the ice coating status of wind turbine blades, characterized in that: include: A collection unit, used for collecting acoustic emission signals; The acoustic emission signal (3) is generated during the operation of the blade, and the acoustic emission sensor (2) collects the acoustic emission signal of the fan blade (1); A decomposition unit, used for performing symplectic geometric mode decomposition on the acoustic emission signal to obtain main symplectic geometric components; A reconstruction unit, used for reconstructing the acoustic emission signal through the main symplectic geometric components; A judgment unit is used to judge the blade state through a deep neural network based on the reconstructed acoustic emission signal; the blade state includes a normal operation state, an ice-covered state and a rainy state.

10. A wind turbine blade icing status monitoring system for implementing the monitoring method of claim 1, characterized in that: It comprises an acoustic emission sensor (2) and an information processor; the acoustic emission sensor (2) is arranged in a fan blade (1), and the acoustic emission sensor (2) and the information processor are connected by electrical signals.

Images