Online diagnosis system of wind turbine vibration fault based on feature extraction

Abstract

The present application relates to the field of wind turbine fault detection, and specifically to an online diagnosis system for wind turbine vibration faults based on feature extraction, the system comprising: a wind turbine vibration signal acquisition module, which acquires the rolling bearing vibration signal of the wind turbine during wind power generation; a wind turbine vibration signal processing module, which decomposes the rolling bearing vibration signal, and obtains a noise interference signal component and a part of the signal component to be diagnosed based on the local distribution and similar characteristics of the decomposed signal; further reconstructs the noise interference signal component to obtain another part of the signal component to be diagnosed; a wind turbine vibration signal fault diagnosis module, which reconstructs the signal component to be diagnosed to obtain the wind turbine vibration signal fault diagnosis result. The present application aims to remove the noise interference signal in the wind turbine vibration signal, so that the vibration fault diagnosis result is more accurate.

Description

Technical Field

[0001] The present application relates to the field of wind turbine fault detection, and in particular to an online wind turbine vibration fault diagnosis system based on feature extraction. Background Art

[0002] Wind turbines are important equipment in the process of wind energy utilization. Vibration failure is one of the common problems of wind turbines. If it is not discovered and handled in time, it may lead to unstable operation of wind turbine equipment and even cause serious accidents. Through online diagnosis, vibration problems that may cause safety hazards can be discovered and eliminated in time to ensure the safety of equipment and personnel.

[0003] However, the natural environmental interference and complex internal structure faced by wind turbines during operation lead to the complexity of vibration fault signals. Due to the complexity of the signals, removing or reducing the noise interference components in the signals becomes extremely difficult and complicated, which increases the risk of misunderstanding or error in the diagnosis process, affecting the accuracy of wind turbine fault diagnosis. Summary of the invention

[0004] In view of the above, it is necessary to provide an online diagnosis system for wind turbine vibration faults based on feature extraction to more accurately remove the interfering noise signals of the vibration signals during the operation of the wind turbines and improve the accuracy of wind turbine vibration fault diagnosis.

[0005] An embodiment of the present application provides a wind turbine vibration fault online diagnosis system based on feature extraction, the system comprising:

[0006] Wind turbine vibration signal acquisition module: used to obtain the rolling bearing vibration signal of the wind turbine in a preset time period during wind power generation;

[0007] Wind turbine vibration signal processing module: used to decompose the rolling bearing vibration signal, mark all rolling bearing signal components based on the local distribution and similarity characteristics of the rolling bearing signal components, and obtain the noise interference signal components and the signal components to be diagnosed; based on the local and overall change characteristics of the noise interference signal components, reconstruct the noise interference signal components and continue to obtain the signal components to be diagnosed;

[0008] Wind turbine vibration signal fault diagnosis module: used to reconstruct the signal components to be diagnosed and obtain the wind turbine vibration signal fault diagnosis results.

[0009] The specific steps of obtaining the noise interference signal component and the signal component to be diagnosed include:

[0010] For each rolling bearing vibration signal component, a preset length is taken as a time series segment;

[0011] Based on the data difference between any two time series segments of the vibration signal components of each rolling bearing, a wind turbine generator set difference coefficient of the any two time series segments is obtained;

[0012] Based on the distribution of the difference coefficients of all wind turbine generator sets of the vibration signal components of each rolling bearing, a period similarity index of the vibration signal components of each rolling bearing is obtained;

[0013] The average level of the period similarity index of all rolling bearing vibration signal components is used as the noise separation threshold of the rolling bearing vibration signal components;

[0014] All rolling bearing vibration signal components whose period similarity index is less than or equal to the noise component threshold are recorded as noise interference signal components; and the remaining rolling bearing vibration signal components are recorded as signal components to be diagnosed.

[0015] The process of obtaining the wind turbine difference coefficients of any two time series segments is as follows:

[0016] For each rolling bearing vibration signal component, the distance metric between any two time series segments is obtained; the difference in the degree of discreteness of the arbitrary two time series segments is calculated, which is recorded as the disturbance difference between the arbitrary two time series segments; the distance metric and the disturbance difference between the arbitrary two time series segments are merged to obtain the wind turbine group difference coefficient of the arbitrary two time series segments.

[0017] The period similarity index of each rolling bearing vibration signal component is specifically an average value of wind turbine group difference coefficients of all pairwise combined time series segments of each rolling bearing vibration signal component.

[0018] The specific process of continuing to obtain the signal component to be diagnosed is as follows:

[0019] Dividing the noise interference signal component into a preset number of segments;

[0020] Compare the data distribution characteristics of each segment of each noise interference signal component with the overall data to obtain the interference degree of each segment;

[0021] Based on the interference degree of all segments of each noise interference signal component, a noise interference matrix of each noise interference signal component is obtained;

[0022] Based on the eigenvalues ​​of each noise interference matrix, obtaining the interference characteristic threshold of each noise interference matrix;

[0023] All eigenvalues ​​and corresponding eigenvectors in each noise interference matrix that are greater than or equal to the corresponding interference feature threshold are used to reconstruct each noise interference signal component, and the reconstructed signal is recorded as the signal component to be diagnosed.

[0024] The interference degree of each segment is obtained as follows:

[0025] Calculate the local disorder degree of each fragment of any noise interference signal component; calculate the overall disorder degree of any noise interference signal component; and use the inverse proportional mapping result of the local disorder degree of each fragment of any noise interference signal component and the corresponding overall disorder degree as the interference degree of each fragment of any noise interference signal component.

[0026] The specific steps of obtaining the noise interference matrix of each noise interference signal component include:

[0027] The interference degree of each segment of each noise interference signal component is fused with the corresponding segment to obtain each interference segment, and the matrix composed of all interference segments is used as the noise interference matrix of each noise interference signal component.

[0028] The interference characteristic threshold of each noise interference matrix is ​​specifically the average value of all characteristic values ​​in each noise interference matrix.

[0029] The specific steps of obtaining the wind turbine vibration signal fault diagnosis result include:

[0030] Reconstruct all components of the signal to be diagnosed to obtain the operation signal of the wind turbine to be diagnosed;

[0031] Using the twin network, the correlation between the wind turbine operating signal to be diagnosed and the standard signal is obtained;

[0032] If the correlation is less than a preset judgment threshold, it is determined that there is a fault in the current power generation operation of the wind turbine set; otherwise, there is no fault in the current power generation operation of the wind turbine set.

[0033] The standard signal is specifically a rolling bearing vibration signal in a preset time period under normal power generation operation of the wind turbine.

[0034] This application has at least the following beneficial effects:

[0035] The present application decomposes the collected vibration signals of the rolling bearings of the wind turbine set to obtain the vibration signal components of the rolling bearings of the wind turbine set, which helps to extract, analyze and understand the important information in the vibration signals, and can also provide an analysis basis for the subsequent elimination of noise signals; all rolling bearing signal components are marked to obtain noise interference signal components and signal components to be diagnosed, and the noise signal is effectively distinguished from the normal signal based on the random change characteristics of the noise interference signal; the local and overall change characteristics of different vibration signal components are extracted to more accurately reflect the change characteristics of the vibration signal of the rolling bearing of the wind turbine set; since misjudgment may occur only by distinguishing based on the change characteristics of the noise signal, the change characteristics of the vibration signal of the rolling bearing of the wind turbine set are combined to more accurately characterize the abnormal vibration signal of the wind turbine set compared with the one-dimensional time series signal, and the noise interference signal in the vibration fault signal of the wind turbine set is further analyzed to more accurately remove the interference noise signal of the vibration signal during the operation of the wind turbine set, and based on the comparison of the signal to be diagnosed with the standard signal, the accurate diagnosis of the vibration fault during the operation of the wind turbine set is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A block diagram of the online diagnosis system for wind turbine vibration faults based on feature extraction provided in this application;

[0037] Figure 2 A specific flow chart for online diagnosis of wind turbine vibration faults provided in this application. DETAILED DESCRIPTION

[0038] In the description of the embodiments of the present application, words such as "exemplary", "or", "for example" and the like are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "or", "for example" and the like is intended to present related concepts in a concrete manner.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the present application. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application.

[0040] It should also be noted that the terms "first" and "second" in this application and its drawings are used to distinguish similar objects, rather than to describe a specific order or sequence. The method disclosed in the embodiments of the present application or the method shown in the flow chart includes one or more steps for implementing the method. Without departing from the scope of protection of this application, the execution order of multiple steps can be interchanged with each other, and some steps can also be deleted.

[0041] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0042] The specific scheme of the online diagnosis system for wind turbine vibration faults based on feature extraction provided by the present application is described in detail below with reference to the accompanying drawings.

[0043] See also Figure 1 , which shows a block diagram of a wind turbine vibration fault online diagnosis system based on feature extraction provided by an embodiment of the present application. The system includes: a wind turbine vibration signal acquisition module 101, a wind turbine vibration signal processing module 102, and a wind turbine vibration signal fault diagnosis module 103.

[0044] The embodiment of the present application first proposes an online diagnosis system for wind turbine vibration faults based on feature extraction, which is applied to the field of wind turbine fault detection. The system includes:

[0045] The wind turbine vibration signal acquisition module 101 is used to acquire the rolling bearing vibration signal of the wind turbine in a preset time period during the wind power generation process.

[0046] Wind turbines include many rotating parts, which bear the core task of dynamic rotation during the power generation process of wind turbines and are in a long-term friction rotation process. Generally speaking, due to the harsh working environment of wind turbines, rolling bearings are very likely to be damaged after long-term operation. The resulting chain reaction of various parts systems of wind turbines seriously affects the normal operation of wind turbines.

[0047] According to the actual working state of the wind turbine during operation, the vibration signal of the bearing during operation of the wind turbine is obtained through a vibration sensor. Commonly used vibration sensors include displacement sensors, velocity sensors and acceleration sensors. Since the acceleration sensor has the excellent characteristics of high flexibility, small size and easy installation, some embodiments of the present application install acceleration sensors at the horizontal position and axial position of the main shaft of the wind turbine respectively to obtain the vibration signal of the rolling bearing in a preset time period during the power generation operation of the wind turbine.

[0048] It should be noted that the preset time period is a set of all historical moments whose time interval with the current moment is less than or equal to the preset duration. The preset duration is 1 hour and the sampling interval is 1 second. The implementer can adjust it at his own discretion and this application does not impose any restrictions on this.

[0049] Wind turbine vibration signal processing module 102: used to decompose the rolling bearing vibration signal, mark all rolling bearing signal components based on the local distribution and similar characteristics of the rolling bearing signal components, and obtain noise interference signal components and signal components to be diagnosed; reconstruct the noise interference signal components based on the local and overall change characteristics of the noise interference signal components, and continue to obtain the signal components to be diagnosed.

[0050] During the actual operation of wind turbines, since the natural environment in which they are located is usually harsh, the vibration signal will be accompanied by greater natural noise interference. At the same time, since there are many internal parts structures of wind turbines, the interference signals generated during the actual operation of different parts structures will also have a greater interference effect on the accurate identification and diagnosis of fault signals during the operation of the wind turbine.

[0051] Based on this, it is necessary to decompose the collected rolling bearing vibration signal. In some embodiments of the present application, the rolling bearing vibration signal is used as the input of the variational mode decomposition (VMD) algorithm to obtain the vibration signal components of each rolling bearing. Since the main meanings of the different rolling bearing vibration signal components obtained by decomposition may be different, it is necessary to further calculate and analyze the rolling bearing vibration signal components, and the number of rolling bearing vibration signal components after decomposition is assumed to be N.

[0052] It should be noted that the VMD algorithm is a known technology, and this application will not elaborate on it. In addition, the implementer may choose other signal decomposition algorithms to process the signal, and this application does not limit this.

[0053] If the wind turbine is in normal operation, the collected rolling bearing vibration signal and the normal operation of the wind turbine during power generation should have certain periodic vibration change characteristics.

[0054] When a wind turbine fails, the collected fault signal and natural environment noise interference signal will destroy the original periodic variation characteristics. However, unlike the environmental noise under natural noise interference, when the rolling bearing of the wind turbine fails, the internal machinery will produce abnormal extrusion and wear, resulting in abnormal changes in the signal vibration amplitude. At this time, the vibration signal will have a more obvious amplitude change relative to the rolling bearing vibration signal under normal operation. At the same time, since the abnormal interference signal generated by the harsh working environment during the operation of the wind turbine has a strong randomness, the signal of this environmental interference will have a more obvious random change characteristic.

[0055] On the other hand, the local numerical variation characteristics of the vibration signal components of the rolling bearings of wind turbines are analyzed:

[0056] For each rolling bearing vibration signal component, a time sequence segment of a preset length is taken and sliding is performed with a preset step length.

[0057] For N different rolling bearing vibration signal components, the length is The time sequence fragment of Slide with step length; where, The value of is 15. Indicates rounding down. Calculate the timing correlation between data information in different time series segments:

[0058] Based on the discrete degree of data of each time series segment of the rolling bearing vibration signal component, the local discrete degree of each time series segment is obtained.

[0059] It should be noted that the degree of dispersion reflects the dispersion or fluctuation of each data point in the data set relative to the overall mean. A higher degree of dispersion means that the differences between the data points are larger, and a lower degree of dispersion means that the data points are closer to the mean. The degree of dispersion of data can be described by variance, standard deviation, coefficient of variation, range, etc.

[0060] In one embodiment of the present application, the variance of each time series segment is used as the local discreteness of each time series segment.

[0061] In another embodiment of the present application, the coefficient of variation of each time series segment is used as the local dispersion of each time series segment.

[0062] In another embodiment of the present application, the product of the variance and the coefficient of variation of each time series segment is used as the local dispersion of each time series segment.

[0063] It should be understood that if an unstable change state of an abnormal data signal appears in the time series segment of the rolling bearing vibration signal component, the data in the corresponding time series segment of the rolling bearing vibration signal component of the wind turbine will be more discrete, so the calculated local discreteness will also be relatively large.

[0064] For each rolling bearing vibration signal component, the distance metric between any two time series segments is obtained; the difference in local discreteness between the any two time series segments is calculated, which is recorded as the disturbance difference between the any two time series segments; the distance metric between the any two time series segments and the disturbance difference are merged to obtain the wind turbine group difference coefficient between the any two time series segments.

[0065] The average value of the wind turbine group difference coefficients of all pairwise combined time series segments of each rolling bearing vibration signal component is taken as the period similarity index of each rolling bearing vibration signal component.

[0066] It should be noted that the distance measurement between sequences is a method of measuring the difference between sequences, including: Euclidean distance, Manhattan distance, DTW distance, etc.; usually, the difference, absolute value of the difference, ratio, etc. are used to measure the numerical difference between variables. The implementer can choose the appropriate method at his own discretion, and this application does not impose any restrictions on this.

[0067] In some embodiments of the present application, the DTW distance between any two time series segments is calculated. If the DTW distance between any two time series segments is smaller, it means that the changes between the two time series segments are more similar. At the same time, the absolute value of the difference between the local discreteness of any two different time series segments is calculated, and the DTW distance between any two time series segments is multiplied by the absolute value of the difference, and the product result is recorded as the wind turbine difference coefficient of any two time series segments.

[0068] It should be understood that if the signal changes between any two time segments in the rolling bearing vibration signal component are relatively similar, the DTW distance of the wind turbine time segment calculated at this time will be smaller, and the absolute value of the local discreteness difference will also be relatively small, so the value of the wind turbine difference coefficient calculated at this time will also be relatively small, indicating that the changes in the values ​​between the corresponding two time segments are relatively similar.

[0069] Furthermore, if the rolling bearing vibration signal component has the characteristic of periodic variation, then any two time series segments of the rolling bearing vibration signal component will have certain similar variation characteristics.

[0070] It should be understood that for the rolling bearing vibration signal component during the operation of the wind turbine, if the signal component represents the normal operating status of the wind turbine, then the signal component should have certain periodic variation characteristics. At this time, the value of the calculated periodic similarity index will also be relatively small, indicating that the rolling bearing vibration signal component is more likely to represent the wind turbine signal under normal operating conditions.

[0071] The N different rolling bearing vibration signal components obtained after decomposing the rolling bearing vibration signal may respectively represent the wind turbine signal, noise interference signal and fault signal under normal operating conditions. At this time, it is necessary to separate the noise interference signal from the N different rolling bearing vibration signal components.

[0072] Based on this, the noise interference signal in the rolling bearing vibration signal component is eliminated:

[0073] The average level of the period similarity index of all rolling bearing vibration signal components is used as the noise separation threshold of the rolling bearing vibration signal components;

[0074] All rolling bearing vibration signal components whose period similarity index is less than or equal to the noise component threshold are recorded as noise interference signal components; and the remaining rolling bearing vibration signal components are recorded as signal components to be diagnosed.

[0075] In some embodiments of the present application, the average value of the period similarity index of N rolling bearing vibration signal components is first calculated and recorded as the rolling bearing vibration signal component noise separation threshold.

[0076] Different from the fault signal generated by the failure of wind turbine components, the noise interference signal has a strong random change characteristic, so the periodic similarity index calculated in the corresponding signal component will also be relatively small.

[0077] Since the signal change characteristic information contained in the rolling bearing vibration signal component is relatively complex, only the noise separation threshold is used to divide the rolling bearing vibration signal component. At this time, some rolling bearing vibration signal components may be misjudged. Therefore, it is necessary to further calculate and analyze the signal state of the noise interference signal component:

[0078] First, construct the noise interference matrix of the noise interference signal component:

[0079] Divide any noise interference signal component into a preset number of segments, and calculate the local disorder degree of each segment; calculate the overall disorder degree of any noise interference signal component; and use the inverse proportional mapping result of the local disorder degree of each segment of any noise interference signal component and the corresponding overall disorder degree as the interference degree of each segment of any noise interference signal component;

[0080] The interference degree of each segment of each noise interference signal component is merged with the corresponding segment to obtain each interference segment, and a matrix composed of all interference segments is used as the noise interference matrix of any noise interference signal component.

[0081] It should be noted that the degree of chaos reflects the irregularity of the overall data distribution, including: Gini impurity, information entropy, variance, etc.; implementers can choose appropriate indicators to characterize the degree of chaos of a set of signals, and this application does not impose any restrictions on this.

[0082] In some embodiments of the present application, the noise interference signal component is divided into m different segments, and the mean filling method or downsampling method is used to make the signal length of each segment also m, where m is 25 and can be adjusted by the implementer. The information entropy of each noise interference signal component is calculated and recorded as the overall chaos degree; the information entropy of each segment is calculated and recorded as the local chaos degree; the ratio of the local chaos degree of each segment of each rolling bearing vibration signal to the corresponding overall chaos degree is used as the interference degree of the corresponding segment; the interference degree of each segment is multiplied by the corresponding segment, and the product result is used as the interference segment; the interference segment is used as the noise interference identification matrix of each rolling bearing vibration signal.

[0083] It should be noted that each noise interference signal component corresponds to a noise interference matrix, wherein each column of the noise interference matrix represents a local change of a rolling bearing vibration signal component.

[0084] It should be understood that if the interference information in the noise interference signal component segment is relatively rich, the corresponding information entropy value calculated for the segment will also be relatively large, and the signal value amplitude in the corresponding column of the constructed noise interference identification matrix will also be relatively large.

[0085] Obtain the eigenvalues ​​and corresponding eigenvectors of each noise interference matrix; take the average value of the eigenvalues ​​of each noise interference matrix as the interference feature threshold; retain all eigenvalues ​​and corresponding eigenvectors greater than or equal to the corresponding interference feature threshold, reconstruct each noise interference signal component, and record the reconstructed signal as the signal component to be diagnosed.

[0086] It should be noted that a noise interference matrix has a corresponding interference eigenvalue.

[0087] In some embodiments of the present application, a singular value decomposition (SVD) algorithm is used as input for each noise interference identification matrix to obtain the eigenvalues ​​and eigenvectors of the corresponding noise interference identification matrix. Since noise has a certain randomness during the operation of the wind turbine, the eigenvalue of the noise information after decomposition will be relatively small compared to the normal signal information.

[0088] Therefore, the average value of the eigenvalues ​​of each noise interference identification matrix is ​​recorded as the interference feature threshold, and the eigenvalues ​​smaller than the interference feature threshold in the corresponding noise interference identification matrix are eliminated, and the eigenvalues ​​greater than or equal to the interference feature threshold are retained; the corresponding noise interference signal components are reconstructed based on the retained eigenvalues ​​and the corresponding eigenvectors, and the reconstructed signals are added to the signal components to be diagnosed.

[0089] It should be noted that signal reconstruction is a known technology and will not be described in detail in this application.

[0090] Wind turbine vibration signal fault diagnosis module 103: reconstructs the signal components to be diagnosed to obtain wind turbine vibration signal fault diagnosis results.

[0091] The rolling bearing vibration signal of the wind turbine in the preset time period under the normal power generation operation state is used as the standard signal, and the signal is used as the diagnosis template to perform online diagnosis of vibration faults during the operation of the wind turbine:

[0092] Specifically, all components of the signal to be diagnosed are reconstructed, and the reconstructed signal is recorded as the wind turbine operating signal to be diagnosed.

[0093] Specifically, the standard rolling bearing vibration signal and the wind turbine operating signal to be diagnosed are respectively used as inputs of the twin network to obtain the correlation between the standard rolling bearing vibration signal and the wind turbine operating signal to be diagnosed.

[0094] The twin network used in some embodiments of the present application is a convolutional twin network, which has three convolutional layers and one fully connected layer, the loss function is cross entropy loss, the optimizer is Adam, the batch size is 64, and the initial learning rate is set to 0.001. The implementer can adjust it according to the actual situation, and this application does not limit it.

[0095] If the correlation between the two signals is less than the preset judgment threshold, there is a large difference between the current wind turbine vibration signal and the wind turbine vibration signal under normal operation, and it is determined that there is a fault in the current wind turbine power generation operation; otherwise, there is no fault in the wind turbine power generation operation.

[0096] In some embodiments of the present application, the judgment threshold is set to 0.92; the implementer can adjust the judgment threshold according to actual conditions.

[0097] Among them, the specific flow chart of online diagnosis of wind turbine vibration fault is as follows: Figure 2 shown.

[0098] It should be understood that the correlation between the wind turbine operating signal to be diagnosed and the standard rolling bearing vibration signal can be obtained through the twin network. If both signals indicate the normal operation of the wind turbine, there should be a strong similarity correlation between the two signals. Conversely, if the current wind turbine is in an abnormal working state, the output value of the twin network should be relatively small, indicating that there is a large difference between the two signals, that is, the current wind turbine operation has a fault.

[0099] The embodiment of the present application decomposes the collected vibration signal of the rolling bearing of the wind turbine set to obtain the vibration signal component of the rolling bearing of the wind turbine set, which is helpful to extract, analyze and understand the important information in the vibration signal, and can also provide an analysis basis for the subsequent elimination of noise signals; all rolling bearing signal components are marked to obtain noise interference signal components and signal components to be diagnosed, and the noise signal is effectively distinguished from the normal signal based on the random change characteristics of the noise interference signal; the local and overall change characteristics of different vibration signal components are extracted to more accurately reflect the change characteristics of the vibration signal of the rolling bearing of the wind turbine set; because the distinction is made only based on the change characteristics of the noise signal, misjudgment may occur, so the change characteristics of the vibration signal of the rolling bearing of the wind turbine set are combined to more accurately characterize the abnormal vibration signal of the wind turbine set compared with the one-dimensional time series signal, and the noise interference signal in the vibration fault signal of the wind turbine set is further analyzed to more accurately remove the interference noise signal of the vibration signal during the operation of the wind turbine set, and based on the comparison of the signal to be diagnosed with the standard signal, accurate diagnosis of vibration faults during the operation of the wind turbine set is achieved.

[0100] The flowchart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to the embodiment of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two continuous boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. In the description corresponding to the flowchart and the block diagram in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in a different order from the order disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two continuous operations or steps can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.

[0101] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. The online diagnosis system of wind turbine vibration fault based on feature extraction is characterized by: The system includes: Wind turbine vibration signal acquisition module: used to obtain the rolling bearing vibration signal of the wind turbine in a preset time period during wind power generation; Wind turbine vibration signal processing module: used to decompose the rolling bearing vibration signal, mark all rolling bearing signal components based on the local distribution and similarity characteristics of the rolling bearing signal components, and obtain the noise interference signal components and the signal components to be diagnosed; based on the local and overall change characteristics of the noise interference signal components, reconstruct the noise interference signal components and continue to obtain the signal components to be diagnosed; Wind turbine vibration signal fault diagnosis module: used to reconstruct the signal components to be diagnosed and obtain the wind turbine vibration signal fault diagnosis results; The steps of obtaining the noise interference signal component and the signal component to be diagnosed include: For each rolling bearing vibration signal component, a preset length is taken as a time series segment; Based on the data difference between any two time series segments of the vibration signal components of each rolling bearing, a wind turbine generator set difference coefficient of the any two time series segments is obtained; Based on the distribution of the difference coefficients of all wind turbine generator sets of the vibration signal components of the rolling bearings, a period similarity index of the vibration signal components of the rolling bearings is obtained; The average level of the period similarity index of all rolling bearing vibration signal components is used as the noise component threshold of the rolling bearing vibration signal component; Record all rolling bearing vibration signal components with a period similarity index less than or equal to the noise component threshold as noise interference signal components; and record the remaining rolling bearing vibration signal components as signal components to be diagnosed; The process of obtaining the wind turbine group difference coefficient of any two time series segments is as follows: For each rolling bearing vibration signal component, obtain the distance metric between any two time series segments; calculate the difference in the degree of discreteness of the any two time series segments, recorded as the disturbance difference of the any two time series segments; fuse the distance metric of the any two time series segments with the disturbance difference to obtain the wind turbine group difference coefficient of the any two time series segments; The period similarity index of each rolling bearing vibration signal component is specifically an average value of wind turbine group difference coefficients of all pairwise combined time series segments of each rolling bearing vibration signal component.

2. The wind turbine vibration fault online diagnosis system based on feature extraction according to claim 1, characterized in that: The specific process of continuing to obtain the signal component to be diagnosed is: Dividing the noise interference signal component into a preset number of segments; Compare the data distribution characteristics of each segment of each noise interference signal component with the overall data to obtain the interference degree of each segment; Based on the interference degree of all segments of each noise interference signal component, a noise interference matrix of each noise interference signal component is obtained; Based on the eigenvalues ​​of each noise interference matrix, obtaining the interference characteristic threshold of each noise interference matrix; All eigenvalues ​​and corresponding eigenvectors in each noise interference matrix that are greater than or equal to the corresponding interference feature threshold are used to reconstruct each noise interference signal component, and the reconstructed signal is recorded as the signal component to be diagnosed.

3. The wind turbine vibration fault online diagnosis system based on feature extraction according to claim 2, characterized in that: The interference degree of each fragment is obtained as follows: Calculate the local disorder degree of each fragment of any noise interference signal component; calculate the overall disorder degree of any noise interference signal component; and use the inverse proportional mapping result of the local disorder degree of each fragment of any noise interference signal component and the corresponding overall disorder degree as the interference degree of each fragment of any noise interference signal component.

4. The wind turbine vibration fault online diagnosis system based on feature extraction according to claim 2, characterized in that: The specific steps of obtaining the noise interference matrix of each noise interference signal component include: The interference degree of each segment of each noise interference signal component is fused with the corresponding segment to obtain each interference segment, and the matrix composed of all interference segments is used as the noise interference matrix of each noise interference signal component.

5. The wind turbine vibration fault online diagnosis system based on feature extraction according to claim 2, characterized in that: The interference characteristic threshold of each noise interference matrix is ​​specifically the average value of all characteristic values ​​in each noise interference matrix.

6. The wind turbine vibration fault online diagnosis system based on feature extraction according to claim 1, characterized in that: The specific steps of obtaining the wind turbine vibration signal fault diagnosis result include: Reconstruct all components of the signal to be diagnosed to obtain the operation signal of the wind turbine to be diagnosed; Using the twin network, the correlation between the wind turbine operating signal to be diagnosed and the standard signal is obtained; If the correlation is less than a preset judgment threshold, it is determined that there is a fault in the current power generation operation of the wind turbine set; otherwise, there is no fault in the current power generation operation of the wind turbine set.

7. The wind turbine vibration fault online diagnosis system based on feature extraction according to claim 6, characterized in that: The standard signal is specifically a rolling bearing vibration signal of a preset time period under normal power generation operation of the wind turbine.

Images