Wind turbine generator state monitoring and fault tracing method and system

Abstract

The invention discloses a wind turbine generator state monitoring and fault tracing method and system, and relates to the technical field of wind power equipment monitoring. The method comprises the following steps: based on a plurality of vibration-acoustic emission sensors, collecting state sample data of the large fan component; feature index extraction is carried out; performing normalization processing on the feature indexes to obtain corresponding feature values; establishing a wind turbine generator large component fuzzy fault tree model; mapping the wind turbine generator large component fuzzy fault tree model into a Bayesian network model; optimizing the Bayesian network model based on the characteristic values; inputting the running state characteristic values of the large components of the fan transmission chain system under various working conditions into the optimized Bayesian network model for fault identification, and generating a running state evaluation result of the large components of the wind turbine generator; and if the large part of the wind turbine generator runs abnormally, the transfer entropy is introduced, and fault traceability diagnosis is carried out. According to the invention, the health state process monitoring of the large fan component and the fault tracing of the large fan component can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power equipment monitoring, and in particular to a method and system for wind turbine status monitoring and fault tracing. Background Art

[0002] Wind power is one of the cleanest energy resources in the world. The generation of wind power needs to rely on wind turbines. Wind turbines convert natural wind energy into mechanical energy, and then convert mechanical energy into electrical energy through the transmission chain system to achieve the conversion of wind energy into electrical energy. Wind turbines have a large structure and high requirements for wind resources. They are generally installed in places with abundant wind resources but sparse population, such as deserts and hills. During the long-term operation of the power generation process, the units are affected by complex environmental factors, their own structure and process parameters. The rotating parts in the transmission chain system are prone to various types of damage. Due to the relatively scattered distribution of the units, there are often serious consequences such as downtime or even unit damage caused by untimely operation and maintenance. Online status monitoring and fault diagnosis research on key components of its transmission chain can timely identify fault signs, grasp the degree of gradual development of faults in real time, and save troubleshooting time. It has important academic significance and engineering practical value for optimizing the operation and maintenance strategy and thus improving the operation reliability of the whole machine.

[0003] At present, the health status assessment of wind turbines is mainly achieved through vibration monitoring and signal analysis. Vibration sensors are installed at key parts of the wind turbine transmission chain to collect vibration data. The collected data is analyzed in the time domain, spectrum, and envelope demodulation to determine the vibration development trend, waveform energy, and main spectrum components of the transmission chain components. The health status of the wind turbine transmission chain is determined by combining expert experience to guide on-site operation and maintenance work. When determining a fault, the above method determines the fault type by spectrum components and preliminarily evaluates the fault level by vibration energy. Since the data collected by this monitoring method is a single type of vibration data and the evaluation model is relatively independent, it has certain applicability for determining faults such as bearing inner ring, outer ring, rolling element, cage fault, and gear wear fault, which have a high degree of consistency between the spectrum components and the fault type, and the high-speed end rotating component fault and health status assessment in the transmission chain system that is less affected by external factors and has obvious fault characteristics.

[0004] However, in actual engineering, due to the large number of components in the wind turbine transmission chain system and the many complex coupling relationships between the components, an early minor fault in any component during the operation of the unit may propagate to multiple fault-related variables, and even evolve complex fault forms in multiple components. The existing technology has the problems of single health status assessment signal of wind turbine transmission chain system, low coupling degree of assessment model, and poor assessment effect caused by unclear fault evolution path, which makes it impossible to achieve effective wind turbine status monitoring and fault tracing for actual engineering conditions. Summary of the invention

[0005] In order to overcome the above problems or at least partially solve the above problems, the present invention provides a method and system for wind turbine status monitoring and fault tracing, which can realize health status process monitoring of major wind turbine components and fault tracing of major wind turbine components.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: In a first aspect, the present invention provides a method for monitoring the state of a wind turbine and tracing a fault, comprising the following steps: Collect status sample data of large components of the fan; Extract characteristic indicators from the state sample data of the large components of the wind turbine to obtain corresponding characteristic indicators; Normalize the characteristic index to obtain the corresponding characteristic value; Establish fuzzy fault tree model of large components of wind turbine; Map the fuzzy fault tree model of large components of wind turbines into a Bayesian network model; Optimize the Bayesian network model based on eigenvalues; The operating state characteristic values ​​of the major components of the wind turbine transmission chain system under various working conditions are input into the optimized Bayesian network model for fault identification, and the operating state evaluation results of the major components of the wind turbine are generated; If the evaluation results of the operating status of major components of a wind turbine show that the major components of the wind turbine are operating abnormally, transfer entropy is introduced to perform fault tracing diagnosis.

[0007] The present invention adopts vibration-acoustic emission sensors to simultaneously collect vibration signals and acoustic emission signals of large components of the transmission chain system, utilizes multi-sensor data fusion technology to fuse the two types of signals, and establishes a causal association model of multiple process variables of the complex system of the fan transmission chain through Bayesian network data-driven technology, so as to realize the process monitoring of the health status of large components of the fan, and then combines the transfer entropy to perform time series causal analysis of the signal to realize the fault tracing research of large components of the fan.

[0008] Based on the first aspect, further, the method for extracting characteristic indicators from the state sample data of large components of a wind turbine comprises the following steps: Extract and calculate effective value and energy based on the operating data in the status sample data of large components of the wind turbine; According to the status sample data of large components of the wind turbine, the high-frequency acoustic emission power spectrum is obtained through FFT analysis, and the frequency center moment and segment spectrum energy are extracted in real time; Based on the status sample data of large wind turbine components, the gear and bearing defects hidden in the noise are identified through time-frequency analysis; Based on the status sample data of large wind turbine components, envelope detection analysis is used to identify the vibration form and damage characteristics caused by the impact of gears and bearings.

[0009] Based on the first aspect, further, the formula for calculating the effective value is: , the formula for calculating energy is: , where: N is the number of samples, is the sample data, is the effective value, i is a natural number from 1 to N, T is the sampling period, is the time domain function of the sample data, Is the energy value.

[0010] Based on the first aspect, further, the method for mapping the fuzzy fault tree model of a large component of a wind turbine generator set into a Bayesian network model comprises the following steps: Mapping the underlying faults of large components of the fuzzy fault tree model of large components of wind turbines to Bayesian root nodes; The intermediate faults of the fuzzy fault tree model of the large components of the wind turbine are mapped to the intermediate nodes of the Bayesian network; The top-level faults of the fuzzy fault tree model of large components of wind turbines are mapped to leaf nodes of the Bayesian network; The logic gates of the fuzzy fault tree model of large components of wind turbines are mapped to directed edges of the Bayesian network.

[0011] Based on the first aspect, further, the above method for optimizing the Bayesian network model based on eigenvalues ​​includes the following steps: The characteristic value is used as the input parameter of the Bayesian network model, and the parameter value of each node of the Bayesian network model is updated by parameter learning to obtain the probability distribution table of each node; The Bayesian network model is optimized based on the probability distribution table of each node to obtain the best matching unit.

[0012] Based on the first aspect, further, the method for introducing transfer entropy to perform fault tracing diagnosis includes the following steps: According to the evaluation results of the operating status of the major components of the wind turbine, the corresponding fault sample data is obtained, and the corresponding information entropy is calculated; The working condition variable corresponding to the calculated minimum information entropy is used as the benchmark variable, the mutual trust entropy between the remaining variables and the benchmark variable is calculated, and the redundant variables are removed based on the mutual trust entropy; Calculate the transfer entropy between the remaining variables after removing the redundant variables, and determine the transmission relationship of the fault between the variables based on the transfer entropy; Based on the calculation of the corresponding variance and standard deviation of the transfer relationship, the variables are screened to obtain the main variables with obvious transfer relationship in the fault transmission process; A transfer relationship diagram is constructed based on the transfer relationship between the main variables to form a complete fault chain and realize fault tracing.

[0013] Based on the first aspect, further, the method for obtaining corresponding fault sample data according to the evaluation results of the operating status of major components of the wind turbine generator set and calculating the corresponding information entropy includes the following steps: Obtain corresponding fault sample data based on the evaluation results of the operating status of major components of the wind turbine; Normalize the fault sample data to obtain a new sample set , calculate the information entropy of the new sample, the calculation formula of information entropy is as follows: in, For variables The probability of is the information entropy.

[0014] Based on the first aspect, further, the method for removing redundant variables based on mutual trust entropy includes the following steps: The correlation coefficient of each mutual trust entropy is calculated, and the variables with correlation coefficients greater than the preset correlation coefficient threshold are regarded as redundant variables, and the corresponding redundant variables are removed.

[0015] Based on the first aspect, further, the method for screening variables by calculating the corresponding variance and standard deviation based on the transfer relationship includes the following steps: The corresponding variance and standard deviation are calculated based on the transfer relationship, and the variables are screened using Chebyshev's inequality.

[0016] In a second aspect, the present invention provides a wind turbine state monitoring and fault tracing system, comprising a sample acquisition module, a feature extraction module, a normalization processing module, a fault tree model building module, a mapping module, a model optimization module, a fault identification module and a fault tracing module, wherein: Sample collection module, used to collect status sample data of large components of the fan; The feature extraction module is used to extract feature indicators from the state sample data of the large components of the wind turbine to obtain corresponding feature indicators; A normalization processing module is used to normalize the characteristic index to obtain the corresponding characteristic value; Fault tree model building module, used to build fuzzy fault tree models of large components of wind turbines; A mapping module, used for mapping the fuzzy fault tree model of the large components of the wind turbine to a Bayesian network model; Model optimization module, used to optimize the Bayesian network model based on eigenvalues; The fault identification module is used to input the operating status characteristic values ​​of the major components of the wind turbine transmission chain system under various working conditions into the optimized Bayesian network model for fault identification and generate the operating status evaluation results of the major components of the wind turbine set; The fault tracing module is used to introduce transfer entropy and perform fault tracing diagnosis when the operating status assessment results of the major components of the wind turbine set show that the major components of the wind turbine set are operating abnormally.

[0017] This system uses the cooperation of multiple modules such as sample acquisition module, feature extraction module, normalization processing module, fault tree model construction module, mapping module, model optimization module, fault identification module and fault tracing module. It uses vibration-acoustic emission sensors to simultaneously collect vibration signals and acoustic emission signals of large components of the transmission chain system, and uses multi-sensor data fusion technology to fuse the two types of signals. Through the Bayesian network data-driven technology, a causal association model of multiple process variables of the complex system of the fan transmission chain is established to realize the process monitoring of the health status of large components of the fan. Then, combined with the transfer entropy, the signal is subjected to time series causal analysis to realize the fault tracing research of large components of the fan.

[0018] The present invention has at least the following advantages or beneficial effects: 1. Using vibration-acoustic emission sensors to collect signals from the wind turbine drive chain system for status assessment and analysis can make up for the shortcomings of single vibration signal monitoring and improve the reliability of monitoring results.

[0019] 2. The wind load changes unpredictably during the operation of wind turbines. The Bayesian network has a strong ability to handle uncertainty problems and can effectively express and integrate the multi-source signals of wind turbine vibration and acoustic emission. When the Bayesian network is used for fault diagnosis, the fault symptoms and fault causes are used as nodes. When the fault symptoms appear, the probability of various fault causes can be inferred based on the causal relationship and probability value between the nodes, thereby obtaining a diagnostic conclusion. This has great advantages in monitoring faults caused by uncertain factors in the complex system of the wind turbine transmission chain.

[0020] 3. After a wind turbine fails, the fault needs to be eliminated in a timely manner. The introduction of transfer entropy can further determine the cause of the fault on the basis of monitoring the existence of the fault, realize fault tracing, and help operation and maintenance personnel to eliminate the fault in time, restore the normal operation of the wind turbine, and effectively ensure the power generation efficiency of the unit.

[0021] 4. Based on the vibration-acoustic emission signals of wind turbines, the fusion of Bayesian network and transfer entropy is applied to the health status monitoring of large components of wind turbine transmission chain. It can not only make up for the shortcomings of traditional model diagnosis technology in dealing with complex fault problems and uncertain fault problems, but also trace the occurrence and development of faults from the time and space dimensions. This method has the ability of parallel reasoning, global update and information tracking, and can effectively realize the health status monitoring and fault tracing diagnosis of large components of wind turbines. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 This is a flow chart of a method for wind turbine status monitoring and fault tracing according to an embodiment of the present invention; Figure 2 A fault tree model of a wind turbine transmission chain system in an embodiment of the present invention; Figure 3 A schematic diagram of a fuzzy fault tree and a Bayesian network mapping in an embodiment of the present invention; Figure 4 This is a schematic diagram of the installation position of the spindle sampling point sensor in an embodiment of the present invention; Figure 5 This is a schematic diagram of the time domain waveform of the main shaft of a normal unit; Figure 6 This is a schematic diagram of the time domain waveform of the main shaft of the abnormal unit; Figure 7 This is a schematic diagram of the kurtosis index of a normal unit; Figure 8 It is a schematic diagram of the kurtosis index of abnormal units; Fig. 9 Schematic diagram of the faulty transmission chain. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0025] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0027] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0028] In the description of the embodiments of the present invention, "plurality" means at least 2.

[0029] Example: like Figure 1 As shown, in a first aspect, an embodiment of the present invention provides a method for monitoring the state of a wind turbine and tracing a fault, comprising the following steps: S1. Collect state sample data of large components of the wind turbine; configure multiple vibration-acoustic emission sensors in the transmission chain system of the wind turbine to collect state data samples of large components of the wind turbine. The installation positions of the vibration-acoustic emission sensors can be axial / radial of the main shaft bearing seat, axial / radial of the low-speed bearing end of the gearbox, axial / radial of the medium-speed bearing end of the gearbox, axial / radial of the high-speed bearing end of the gearbox, axial / radial of the generator input end bearing seat, and axial / radial of the generator output end bearing seat. First, the end surface of the measured position is smoothed and polished, and then the vibration-acoustic emission sensor is magnetically attracted to the smoothed position through the magnetic seat, and the measuring points are numbered. The sensor data line is connected to the corresponding data interface of the acquisition device in the order of the numbers, and the data acquisition system is built and the sampling setting is performed to obtain the original state data samples. Large components of wind turbines refer to large components of wind power direct-drive units and large components of wind power double-fed units. The large components of wind power direct-drive units include: main shaft and generator; the large components of wind power double-fed units include: main shaft, gearbox and generator.

[0030] S2. Extract characteristic indicators from the state sample data of the large components of the wind turbine to obtain corresponding characteristic indicators; the characteristic indicators include displacement, speed, effective value, peak-to-peak value, skewness, pulse, kurtosis, etc.

[0031] Furthermore, it includes: extracting and calculating effective value and energy based on operating data in the status sample data of large components of the fan; obtaining high-frequency acoustic emission power spectrum through FFT analysis based on the status sample data of large components of the fan and extracting frequency center moment and segmented spectrum energy in real time; identifying gear and bearing defects hidden in noise through time-frequency analysis based on the status sample data of large components of the fan; identifying vibration form and damage characteristics caused by gear and bearing impact through envelope detection analysis based on the status sample data of large components of the fan.

[0032] Further, the formula for calculating the effective value is: , the formula for calculating energy is: , where: N is the number of samples, is the sample data, is the effective value, i is a natural number from 1 to N, T is the sampling period, is the time domain function of the sample data, is the energy value. The energy signal average power calculation formula is: , P is the average power of the energy signal.

[0033] In some embodiments of the present invention, different characteristic parameters are selected as characteristic values ​​to characterize the operating status of the fan according to the monitored parts and structural characteristics. For example, characteristic parameters such as effective value and energy can be used to monitor cracks and low-speed bearings. The high-frequency acoustic emission power spectrum obtained by FFT analysis can be used to extract the frequency center moment and segmented spectrum energy for classification and identification of cracks and wear defects. Time-frequency analysis can effectively identify gear and bearing defects hidden in the noise. Envelope detection analysis can identify vibration forms and damage characteristics caused by gear and bearing impact.

[0034] S3, normalizing the characteristic index to obtain the corresponding characteristic value; In some embodiments of the present invention, the above-mentioned characteristic index is normalized to obtain a characteristic value, and the characteristic value is used as the input of the Bayesian network model to perform Bayesian network model learning. The normalization calculation formula is: ,in, is the characteristic value, is the minimum eigenvalue, is the maximum eigenvalue, is the normalized eigenvalue.

[0035] S4. A fuzzy fault tree model of large components is established based on the differences in structural characteristics and environmental factors of large components of wind turbines. The structural characteristics can be divided into doubly fed units (including gearboxes) and direct drive units (excluding gearboxes). Environmental factors include altitude, temperature, humidity, wind speed, wind direction and other factors. A targeted fuzzy fault tree model is constructed based on different structures combined with corresponding environmental factors.

[0036] In some embodiments of the present invention, a fuzzy fault tree model of large components of a wind turbine is established. Combined with empirical analysis, it can be known that the abnormal conditions that often occur in the transmission chain system during the operation of a wind turbine are roughly divided into: main shaft abnormality, gear box abnormality, and generator abnormality; among them, main shaft abnormality is often caused by main shaft bearing failure, temperature abnormality, shaft abnormality, etc., gear box abnormality is often caused by gear box bearing failure, gear failure, temperature abnormality and shaft abnormality, and generator abnormality is often caused by bearing failure, temperature abnormality, and winding abnormality at both ends of the generator; further subdivided, bearing failure may be inner ring failure, outer ring failure, rolling element failure or retaining cage failure, gear failure may be gear wear, gear pitting, gear breakage, etc., and shaft abnormality may be caused by shaft imbalance or misalignment. Based on this, a common fault model tree of large components of a wind turbine is established as follows: Figure 2 shown.

[0037] S5, mapping the fuzzy fault tree model of the large components of the wind turbine to a Bayesian network model; Furthermore, it includes: mapping the bottom-level faults of large components of the fuzzy fault tree model of large components of the wind turbine generator set to Bayesian root nodes; mapping the middle-level faults of the fuzzy fault tree model of large components of the wind turbine generator set to Bayesian network intermediate nodes; mapping the top-level faults of the fuzzy fault tree model of large components of the wind turbine generator set to Bayesian network leaf nodes; and mapping the logic gates of the fuzzy fault tree model of large components of the wind turbine generator set to Bayesian network directed edges.

[0038] In some embodiments of the present invention, a Bayesian network structure is determined. The fuzzy fault tree model of the large components of the wind turbine is mapped to a Bayesian network model, such as Figure 3 As shown, the bottom-level faults of large components are mapped to the root nodes of the Bayesian network, the middle-level faults are mapped to the intermediate nodes of the Bayesian network, the top-level faults are mapped to the leaf nodes of the Bayesian network, and the logic gates are mapped to the directed edges of the Bayesian network.

[0039] S6. Optimizing the Bayesian network model based on the eigenvalues; Furthermore, it includes: taking the characteristic value as the input parameter of the Bayesian network model, updating the parameter value of each node of the Bayesian network model by parameter learning, and obtaining the probability distribution table of each node; optimizing the Bayesian network model based on the probability distribution table of each node to obtain the best matching unit.

[0040] In some embodiments of the present invention, after the Bayesian network structure is determined, the Bayesian network is subjected to machine learning, and the specific steps are: the characteristic values ​​of the above steps are used as the input parameters of the Bayesian network, and the parameter values ​​of each node are continuously updated by parameter learning to obtain a probability distribution table suitable for each node, thereby optimizing the model and obtaining the best matching unit. Here, the Bayesian estimation method is used for parameter learning, and the posterior distribution of the Bayesian estimation and the mathematical expression of the Bayesian parameter are respectively: , in, θ is the Bayesian parameter, for θ The prior distribution of D For sample , P ( D | θ ) is a sample D The joint distribution of θ is the Bayesian parameter θ The differential of .

[0041] S7. Input the operating status characteristic values ​​of the major components of the wind turbine transmission chain system under various working conditions into the optimized Bayesian network model for fault identification, and generate the operating status evaluation results of the major components of the wind turbine set; input the operating status characteristic values ​​of the major components of the wind turbine transmission chain system under the working conditions of concern into the Bayesian network model to obtain the posterior probability of the fault occurrence. The difference between the posterior probability and the posterior probability of the best matching unit represents the possibility of the fault occurrence. When the difference is less than the threshold, it is determined that the operating status of the major components of the wind turbine set is normal. Conversely, when the difference is greater than the threshold, it is determined that the operating status of the major components of the wind turbine set is abnormal.

[0042] S8. If the evaluation results of the operating status of the major components of the wind turbine generator set show that the major components of the wind turbine generator set are operating abnormally, transfer entropy is introduced to perform fault tracing diagnosis.

[0043] Furthermore, it includes: obtaining corresponding fault sample data according to the evaluation results of the operating status of major components of the wind turbine generator set, and calculating the corresponding information entropy; using the operating condition variable corresponding to the calculated minimum information entropy as the benchmark variable, calculating the mutual trust entropy between the remaining variables and the benchmark variable, and removing redundant variables based on the mutual trust entropy; calculating the transfer entropy between the remaining variables after removing the redundant variables, and determining the transfer relationship of the fault between each variable based on the transfer entropy; calculating the corresponding variance and standard deviation based on the transfer relationship, screening the variables to obtain the main variables with obvious transfer relationship in the fault transmission process; constructing a transfer relationship diagram according to the transfer relationship between the main variables, forming a complete fault chain, and realizing fault tracing.

[0044] Further, it includes: obtaining corresponding fault sample data according to the operating status evaluation results of the major components of the wind turbine; normalizing the fault sample data to obtain a new sample set , calculate the information entropy of the new sample, the calculation formula of information entropy is as follows: ,in, For variables The probability of is the information entropy.

[0045] Furthermore, the above method for removing redundant variables based on mutual trust entropy includes the following steps: The correlation coefficient of each mutual trust entropy is calculated, and the variables with correlation coefficients greater than the preset correlation coefficient threshold are regarded as redundant variables, and the corresponding redundant variables are removed.

[0046] Furthermore, the above method for calculating the corresponding variance and standard deviation based on the transfer relationship and screening the variables includes the following steps: The corresponding variance and standard deviation are calculated based on the transfer relationship, and the variables are screened using Chebyshev's inequality.

[0047] In some embodiments of the present invention, the specific steps of performing fault tracing are as follows: Determine information entropy: Normalize the above fault condition samples to obtain a new sample set , calculate the information entropy of the new sample, the calculation formula of information entropy is as follows: ,in, For variables The probability of is the information entropy.

[0048] Remove redundant variables in the new sample: Take the operating condition variable corresponding to the minimum information entropy calculated from the new sample as the benchmark variable , calculate the sample set X The remaining variables and the benchmark variables The mutual trust entropy , and then calculate the correlation coefficient of mutual trust entropy , take the confidence interval as [-0.95, +0.95] correlation coefficient threshold , remove Redundant variables.

[0049] Mutual Trust Entropy The calculation formula is as follows: ,in, p ( x,y ) is a variable x and y The joint distribution probability of p ( x ), p ( y ) are variables x and y The boundary distribution probability of and For variables x and y The differential of .

[0050] The calculation formula of the correlation coefficient of mutual trust entropy is as follows: ,in, For variablesX and Y The mutual trust entropy, and Variables X and Y The information entropy of is the correlation coefficient of mutual trust entropy.

[0051] Determine the transfer entropy: Calculate the transfer entropy between the remaining variables after removing the redundant variables. The transfer entropy calculation formula is as follows: ,in, for , , The joint distribution probability of for , The conditional distribution probability of for , The conditional distribution probability of .

[0052] When the transfer entropy of X to Y is greater than the transfer entropy of Y to X, X is called the cause and Y is called the effect, and the causal relationship between the two variables is established to determine the transfer relationship of the fault between the variables. , , t ( Y|X )for X right Y The transfer entropy of .

[0053] calculate The variance and standard deviation of the variables are further screened using Chebyshev inequality to obtain the main variables with obvious transmission relationships during the fault occurrence process. A transmission relationship diagram is constructed based on the transmission relationship between the main variables to form a complete fault chain and realize fault tracing.

[0054] The present invention adopts vibration-acoustic emission sensors to simultaneously collect vibration signals and acoustic emission signals of large components of the transmission chain system, utilizes multi-sensor data fusion technology to fuse the two types of signals, and establishes a causal association model of multiple process variables of the complex system of the fan transmission chain through Bayesian network data-driven technology, so as to realize the process monitoring of the health status of large components of the fan, and then combines the transfer entropy to perform time series causal analysis of the signal to realize the fault tracing research of large components of the fan.

[0055] In a second aspect, an embodiment of the present invention provides a wind turbine state monitoring and fault tracing system, including a sample acquisition module, a feature extraction module, a normalization processing module, a fault tree model construction module, a mapping module, a model optimization module, a fault identification module and a fault tracing module, wherein: Sample collection module, used to collect status sample data of large components of the fan; The feature extraction module is used to extract feature indicators from the state sample data of the large components of the wind turbine to obtain corresponding feature indicators; A normalization processing module is used to normalize the characteristic index to obtain the corresponding characteristic value; Fault tree model building module, used to build fuzzy fault tree models of large components of wind turbines; A mapping module, used for mapping the fuzzy fault tree model of the large components of the wind turbine to a Bayesian network model; Model optimization module, used to optimize the Bayesian network model based on eigenvalues; The fault identification module is used to input the operating status characteristic values ​​of the major components of the wind turbine transmission chain system under various working conditions into the optimized Bayesian network model for fault identification and generate the operating status evaluation results of the major components of the wind turbine set; The fault tracing module is used to introduce transfer entropy and perform fault tracing diagnosis when the operating status assessment results of the major components of the wind turbine set show that the major components of the wind turbine set are operating abnormally.

[0056] This system uses the cooperation of multiple modules such as sample acquisition module, feature extraction module, normalization processing module, fault tree model construction module, mapping module, model optimization module, fault identification module and fault tracing module. It uses vibration-acoustic emission sensors to simultaneously collect vibration signals and acoustic emission signals of large components of the transmission chain system, and uses multi-sensor data fusion technology to fuse the two types of signals. Through the Bayesian network data-driven technology, a causal association model of multiple process variables of the complex system of the fan transmission chain is established to realize the process monitoring of the health status of large components of the fan. Then, combined with the transfer entropy, the signal is subjected to time series causal analysis to realize the fault tracing research of large components of the fan.

[0057] In order to further verify the effect of the present invention, the main bearing health status monitoring of a 2MW doubly-fed wind turbine generator set in a certain wind farm of a certain main engine factory is taken as an example. There are 33 wind turbines in the wind farm, and each wind turbine is numbered (for example: 1# wind turbine, 2# wind turbine...33# wind turbine). Three DFS-6000 vibration-acoustic emission dual-function sensors are selected and installed in the horizontal radial, vertical radial and axial appropriate positions of the front bearing seat of each wind turbine main shaft (such as Figure 4As shown), the three measuring points of each fan are numbered (for example: 1# fan_1# main bearing seat 1 radial, 1# fan_1# main bearing seat 2 radial, 1# fan_1# main bearing seat 1 axial), and the sampling parameters are set. Since the fan main shaft has a low rotation frequency and a weak characteristic signal, in order to improve the frequency resolution, the sampling frequency is set to 256Hz and the sampling time is greater than 10min. After the configuration is completed according to the above requirements, the sensor channels are numbered and connected to the data acquisition device to collect status data samples of normal units and abnormal units respectively. Then, the time domain waveforms of the data samples of the normal unit (unit number: 7# fan) and the abnormal unit (unit number: 30# fan) are compared, and it is found that the vibration amplitude of the abnormal unit is slightly larger than the vibration amplitude of the normal unit (such as Figure 5 , Figure 6 As shown), but the difference is not very obvious, and it is impossible to determine whether a fault has occurred or not. The radial measurement point sample signals when the main shaft is running normally are selected for Bayesian network model learning. In order to make the learning effect better, 100 sample data are selected, and the time domain features are extracted for each sample. Finally, the best effective value, peak-to-peak value, skewness, pulse, and kurtosis are determined as the characteristic values ​​of the five indicators with the best effect as the evaluation indicators of the fan main bearing operation status, and the prior probability distribution of the five indicators is calculated as the best matching value. The characteristic values ​​of the above five indicators of each sample of the abnormal unit are calculated and input into the Bayesian network model of the present invention, and the posterior probability distribution of the five indicators of each sample is calculated. The results show that the posterior probabilities of the above five indicators of the abnormal unit all exceed the threshold. Taking the kurtosis index as an example, the kurtosis index of the normal unit and the abnormal unit is extracted as shown in the figure. Figure 7 , Figure 8 As shown in the figure, the kurtosis index of the abnormal unit far exceeds that of the normal unit, which indicates that the main shaft bearing of the abnormal unit does have fault characteristics.

[0058] The transfer entropy is used to analyze the propagation evolution path of the above spindle bearing fault. Figure 2 From the fault tree model, we know that there are four variables that affect the running state of the spindle bearing (bearing inner ring fault, bearing outer ring fault, bearing rolling element fault, and bearing cage fault). The fault samples are normalized, and then the information entropy between the four variables is estimated by kernel density, as shown in Table 1. Due to the small number of variables, in order to improve the accuracy of the analysis, redundant variables are not considered here, and the transfer entropy between all variables is directly calculated. There are a total of 16 combinations between the four variables, that is, there may be 16 results of fault transmission. The transfer entropy of the 16 transmissions is calculated, as shown in Table 2. According to the transmission relationship between the variables, the propagation path is drawn and the fault chain is established, as shown in Table 2. Fig. 9 As shown by Fig. 9 It can be seen that variable 2 (bearing outer ring fault) is the key point of multiple propagation paths. Therefore, the fault type of the main bearing of the wind turbine is considered to be a bearing outer ring fault, and the main shaft bearing fault tracing research of this case is realized.

[0059] Table 1: Information entropy table of spindle bearing fault variables

[0060] Table 2: Spindle bearing fault variable transfer entropy table

[0061] Similarly, the above scheme can be used for health status assessment and traceability of large transmission chain components such as wind turbine gearboxes and generators.

[0062] On the third aspect, an embodiment of the present application provides a monitoring system, which includes a memory and a processor, for regularly recording and analyzing the status of major components of a wind turbine, promptly discovering hidden dangers, finding out the causes of the problems, and taking corresponding measures to resolve them.

[0063] It also includes a data acquisition station and a field server. The data acquisition station is an intelligent data acquisition system with a microprocessor as the core. It performs A / D conversion on various sensor signals, and performs time / frequency domain analysis (such as FFT analysis, envelope spectrum analysis, etc.) on the converted vibration signals, and sends the time / frequency domain analysis results to the CMS field server.

[0064] The on-site server is composed of a high-performance professional server and corresponding software. It mainly completes the following functions: long-term storage and management of data; data transmission and display based on B / S or C / S structure; diagnostic analysis and spectrum analysis based on time and frequency domain; system management and settings, including address settings, vibration quantity settings, process quantity settings, and alarm threshold settings for each data acquisition station; sensor and cable fault detection; reporting system faults and restarting the system.

[0065] In the embodiments provided in the present application, it should be understood that the disclosed method and system can also be implemented in other ways. The method and system embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the method and system, method and computer program product according to multiple embodiments 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 a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of boxes in the block diagram and / or the flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0066] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0067] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, a method as in any one of the first aspects described above is implemented. If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or part of the contribution to the prior art or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a disk or an optical disk.

[0068] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0069] It will be apparent to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the present application is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present application. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for monitoring the state of a wind turbine and tracing the source of a fault, characterized in that: The following steps are involved: Collect status sample data of large components of the fan; Extract characteristic indicators from the state sample data of the large components of the wind turbine to obtain corresponding characteristic indicators; Normalize the characteristic index to obtain the corresponding characteristic value; Establish fuzzy fault tree model of large components of wind turbine; Map the fuzzy fault tree model of large components of wind turbines into a Bayesian network model; Optimize the Bayesian network model based on the eigenvalues; The operating state characteristic values ​​of the major components of the wind turbine transmission chain system under various working conditions are input into the optimized Bayesian network model for fault identification, and the operating state evaluation results of the major components of the wind turbine are generated; If the evaluation results of the operating status of major components of a wind turbine show that the major components of the wind turbine are operating abnormally, transfer entropy is introduced to perform fault tracing diagnosis.

2. A method for wind turbine status monitoring and fault tracing according to claim 1, characterized in that: The method for extracting characteristic indicators from the state sample data of the large components of the wind turbine comprises the following steps: Extract and calculate effective value and energy based on the operating data in the status sample data of large components of the wind turbine; According to the status sample data of large components of the wind turbine, the high-frequency acoustic emission power spectrum is obtained through FFT analysis, and the frequency center moment and segment spectrum energy are extracted in real time; Based on the status sample data of large wind turbine components, the gear and bearing defects hidden in the noise are identified through time-frequency analysis; Based on the status sample data of large wind turbine components, envelope detection analysis is used to identify the vibration form and damage characteristics caused by the impact of gears and bearings.

3. A method for wind turbine status monitoring and fault tracing according to claim 2, characterized in that: The formula for calculating the effective value is: , the formula for calculating energy is: , where: N is the number of samples, is the sample data, is the effective value, i is a natural number from 1 to N, T is the sampling period, is the time domain function of the sample data, Is the energy value.

4. A method for wind turbine status monitoring and fault tracing according to claim 1, characterized in that: The method for mapping the fuzzy fault tree model of a large component of a wind turbine generator set into a Bayesian network model comprises the following steps: Mapping the underlying faults of large components of the fuzzy fault tree model of large components of wind turbines to Bayesian root nodes; The intermediate faults of the fuzzy fault tree model of the large components of the wind turbine are mapped to the intermediate nodes of the Bayesian network; The top-level faults of the fuzzy fault tree model of large components of wind turbines are mapped to leaf nodes of the Bayesian network; The logic gates of the fuzzy fault tree model of large components of wind turbines are mapped to directed edges of the Bayesian network.

5. A method for wind turbine status monitoring and fault tracing according to claim 1, characterized in that: The method for optimizing the Bayesian network model based on eigenvalues ​​comprises the following steps: The characteristic value is used as the input parameter of the Bayesian network model, and the parameter value of each node of the Bayesian network model is updated by parameter learning to obtain the probability distribution table of each node; The Bayesian network model is optimized based on the probability distribution table of each node to obtain the best matching unit.

6. A method for wind turbine status monitoring and fault tracing according to claim 1, characterized in that: The method for introducing transfer entropy and performing fault tracing diagnosis comprises the following steps: According to the evaluation results of the operating status of the major components of the wind turbine, the corresponding fault sample data is obtained, and the corresponding information entropy is calculated; The working condition variable corresponding to the calculated minimum information entropy is used as the benchmark variable, the mutual trust entropy between the remaining variables and the benchmark variable is calculated, and the redundant variables are removed based on the mutual trust entropy; Calculate the transfer entropy between the remaining variables after removing the redundant variables, and determine the transmission relationship of the fault between the variables based on the transfer entropy; Based on the calculation of the corresponding variance and standard deviation of the transfer relationship, the variables are screened to obtain the main variables with obvious transfer relationship in the fault transmission process; A transfer relationship diagram is constructed based on the transfer relationship between the main variables to form a complete fault chain and realize fault tracing.

7. A method for wind turbine status monitoring and fault tracing according to claim 6, characterized in that: The method for obtaining corresponding fault sample data according to the evaluation results of the operating status of the major components of the wind turbine generator set and calculating the corresponding information entropy comprises the following steps: Obtain corresponding fault sample data based on the evaluation results of the operating status of major components of the wind turbine; Normalize the fault sample data to obtain a new sample set , calculate the information entropy of the new sample, the calculation formula of information entropy is as follows: in, For variables The probability of is the information entropy.

8. A method for wind turbine status monitoring and fault tracing according to claim 6, characterized in that: The method for removing redundant variables based on mutual trust entropy comprises the following steps: Calculate the correlation coefficient of each mutual trust entropy, take the variables with correlation coefficient greater than the preset correlation coefficient threshold as redundant variables, and remove the corresponding redundant variables.

9. A method for wind turbine status monitoring and fault tracing according to claim 6, characterized in that: The method for calculating the corresponding variance and standard deviation based on the transfer relationship and screening the variables comprises the following steps: The corresponding variance and standard deviation are calculated based on the transfer relationship, and the variables are screened using Chebyshev's inequality.

10. A wind turbine status monitoring and fault tracing system, characterized in that: It includes sample collection module, feature extraction module, normalization processing module, fault tree model construction module, mapping module, model optimization module, fault identification module and fault tracing module, among which: Sample collection module, used to collect status sample data of large components of the fan; The feature extraction module is used to extract feature indicators from the state sample data of the large components of the wind turbine to obtain corresponding feature indicators; A normalization processing module is used to normalize the characteristic index to obtain the corresponding characteristic value; Fault tree model building module, used to build fuzzy fault tree models of large components of wind turbines; A mapping module, used for mapping the fuzzy fault tree model of the large components of the wind turbine to a Bayesian network model; Model optimization module, used to optimize the Bayesian network model based on eigenvalues; The fault identification module is used to input the operating status characteristic values ​​of the major components of the wind turbine transmission chain system under various working conditions into the optimized Bayesian network model for fault identification and generate the operating status evaluation results of the major components of the wind turbine set; The fault tracing module is used to introduce transfer entropy and perform fault tracing diagnosis when the operating status assessment results of the major components of the wind turbine set show that the major components of the wind turbine set are operating abnormally.

Images