Wind turbine state real-time monitoring and diagnosis method, system, equipment and medium

Abstract

The invention provides a wind turbine state real-time monitoring and diagnosing method, system and device and a medium. The method comprises the steps that state parameters of key components of a wind turbine generator are collected in real time through a distributed monitoring network; performing cleaning, alignment and feature extraction on the data, and constructing a visual monitoring interface; building an early warning model of a whole machine and subsystem level based on the extracted features, and achieving the automatic recognition and positioning of a fault; and automatically matching a preset fault processing scheme according to the fault type. Through combination of multi-source data fusion, the deep learning model and fault tree reasoning, the accuracy and response speed of fault diagnosis are remarkably improved, real-time monitoring, intelligent diagnosis and operation and maintenance decision support of the state of the wind turbine generator are realized, and the operation and maintenance cost and the non-planed downtime are effectively reduced.

Description

TECHNICAL FIELD

[0001] The present document relates to the technical field of wind turbine fault diagnosis, and in particular to a wind turbine state real-time monitoring and diagnosis method, system, device and medium. BACKGROUND

[0002] As an important part of clean energy, the large-scale development of wind energy puts forward higher requirements for operation and maintenance technology. Traditional wind turbine operation and maintenance mainly relies on the "periodic inspection + after maintenance" mode, which has problems such as fault discovery lag, blind repair, and high operation and maintenance cost. Especially in complex working conditions such as high cold and high wind speed, early faults of key components (such as gearboxes, main shafts, blades, etc.) are difficult to be discovered in time, resulting in a high fault omission rate, which seriously affects the power generation efficiency and operation safety of the unit.

[0003] In the prior art, although some fault diagnosis systems have tried to introduce vibration monitoring, temperature detection and other means, there are still the following shortcomings: Insufficient monitoring: Most systems only focus on a single type of data (such as vibration), and lack of synchronous collection and fusion analysis of multi-dimensional parameters such as oil, electricity, and blade state; Low diagnosis accuracy: general algorithms are difficult to adapt to the fault characteristics of specific models (such as Shanghai Electric EW6.25-202), especially in the case of coupled vibration and small sample early fault; Ineffective response: fault positioning and treatment scheme pushing rely on human experience, and lack intelligent and automated decision support mechanisms; Poor system integration: Most platforms fail to seamlessly integrate with wind turbine original system (such as SCADA) and group-level operation and maintenance platform, resulting in serious data island phenomenon.

[0004] Therefore, there is an urgent need for a wind turbine fault diagnosis and intelligent prediction system that can realize full-state monitoring, high-precision diagnosis, fast response and intelligent decision-making. SUMMARY

[0005] According to the embodiments of the present application, a wind turbine state real-time monitoring and diagnosis method, system, device and medium are provided to solve the above problems.

[0006] According to the embodiments of the present application, a wind turbine state real-time monitoring and diagnosis method is provided, which includes: S1, collecting state parameters of key components of a wind turbine through a distributed monitoring network in real time, the key components including a gearbox, a main shaft, a generator, a blade and a variable pitch system; S2, data cleaning, alignment and feature extraction are performed on the collected state parameters, a monitoring system is built based on a graphical interface framework, and the extracted features, key parameter original values and historical change trends are displayed in a chart form in real time; S3, automatic fault identification and positioning are performed based on the extracted features, including building early warning models at the whole machine level and the subsystem level, and realizing fault positioning through state evaluation indexes and residual error analysis output by the models; S4, according to the fault identification and positioning results, a corresponding processing scheme is automatically matched from a preset fault knowledge base.

[0007] According to the embodiment of the application, a wind turbine state real-time monitoring and diagnosis system is provided, which comprises: A data acquisition module is configured to collect state parameters of key components of a wind turbine in real time through a distributed monitoring network, wherein the key components include a gearbox, a main shaft, a generator, a blade and a variable pitch system. A data processing module is configured to perform data cleaning, alignment and feature extraction on the collected state parameters, and build a monitoring system based on a graphical interface framework, and display the extracted features, key parameter original values and historical change trends in a chart form in real time. A fault positioning module is configured to perform automatic fault identification and positioning based on the extracted features, including building early warning models at the whole machine level and the subsystem level, and realizing fault positioning through state evaluation indexes and residual error analysis output by the models. A processing scheme module is configured to automatically match a corresponding processing scheme from a preset fault knowledge base according to the fault identification and positioning results.

[0008] According to the embodiment of the application, an electronic device is provided, which comprises: A processor; and A memory arranged to store computer executable instructions, which, when executed, cause the processor to perform the wind turbine state real-time monitoring and diagnosis method.

[0009] According to the embodiment of the application, a storage medium is provided for storing computer executable instructions, which, when executed, implement the wind turbine state real-time monitoring and diagnosis method.

[0010] According to the embodiment of the application, a distributed monitoring network covering key components of a wind turbine is built, realizing all-round and real-time collection and deep fusion of multi-source state parameters, and significantly improving the comprehensiveness and accuracy of device state perception. Through the introduction of feature extraction and fault diagnosis models based on deep learning, the system can effectively capture the weak features of early faults, realize accurate identification and root positioning of faults, and greatly improve the timeliness and reliability of diagnosis. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the one or more embodiments of the present specification or the prior art, the drawings needed to be used in the embodiment or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments described in the present specification, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0012] Figure 1 The flow chart of the wind turbine state real-time monitoring and diagnosis method of the embodiment of the present application; Figure 2 The schematic diagram of the wind turbine state real-time monitoring and diagnosis system of the embodiment of the present application. DETAILED DESCRIPTION

[0013] In order to make the person skilled in the art better understand the technical solutions in the one or more embodiments of the present specification, the technical solutions in the one or more embodiments of the present specification will be described clearly and completely in the following with reference to the drawings in the one or more embodiments of the present specification. Obviously, the described embodiments are only some embodiments of the present specification, not all. Based on the one or more embodiments of the present specification, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the present document.

[0014] Method embodiment According to the embodiment of the present application, a wind turbine state real-time monitoring and diagnosis method is provided, Figure 1 The flow chart of the wind turbine state real-time monitoring and diagnosis method of the embodiment of the present application, according to Figure 1 The wind turbine state real-time monitoring and diagnosis method of the embodiment of the present application specifically includes: S1, collecting the state parameters of the key components of the wind turbine set in real time through a distributed monitoring network, the key components including a gearbox, a main shaft, a generator, a blade and a variable pitch system; The collecting the state parameters of the key components of the wind turbine set in real time through a distributed monitoring network specifically includes: Through a vibration sensor, the vibration amplitude and frequency characteristics of the bearing are monitored to obtain vibration data; Through an oil quality sensor, the viscosity and granularity of the oil are detected online to obtain oil data; Through an electrical parameter sensor adapted to the type of the stator winding of the generator, electrical parameters are collected, the electrical parameters including a stator current, a rotor current and a rotor voltage; Through the monitoring terminal integrated with icing thickness and load sensors fixed on the blade surface, the blade icing and aerodynamic load states are monitored, and blade surface data are obtained. Through the micro temperature sensor array deployed on the gearbox bearing, generator stator and variable pitch motor winding, the temperature changes of the key components are tracked in real time, and temperature data are obtained. Through the voltage and current sensors configured in the frequency converter unit, the three-phase voltage, three-phase current and temperature parameters of the frequency converter are collected, and frequency converter operation data are obtained.

[0015] S2, data cleaning, alignment and feature extraction are performed on the collected state parameters, and a monitoring system is built based on a graphical interface framework, and the extracted features, key parameter original values and historical change trends are displayed in a chart form in real time. The data cleaning, alignment and feature extraction on the collected state parameters specifically include: For vibration data, wavelet denoising processing is adopted, and the root mean square value, peak factor feature in time domain, and 1 / 3 octave spectrum feature in frequency domain are extracted; specifically, Db4 wavelet basis is used for three-layer wavelet packet decomposition and reconstruction to filter out noise, and then the root mean square value and peak factor in time domain of the vibration signal are calculated. Frequency domain analysis is first performed by fast Fourier transform, and then 1 / 3 octave frequency band is divided according to international standards, and the energy proportion of each frequency band is calculated as the spectrum feature.

[0016] For oil data, dynamic thresholds are established based on viscosity and particle size trends to identify abnormal degradation data, and oil degradation rate features are extracted; specifically, a thirty-day sliding window is used to calculate the mean and standard deviation of historical viscosity and particle size data, and data beyond the range of two standard deviations are marked as abnormal. The oil degradation rate is obtained by linear regression calculation of the viscosity change slope and particle size growth rate in the past seven days.

[0017] For electrical parameters, current harmonic distortion rate and voltage unbalance degree features are extracted through three-phase unbalance degree analysis and harmonic distortion calculation; specifically, the percentage of three-phase voltage negative sequence component and positive sequence component is calculated as the voltage unbalance degree according to the national standard. The current harmonic distortion rate is calculated by fast Fourier transform analysis of the ratio of the square root of the sum of the effective values of the second to fiftieth harmonic currents to the effective value of the fundamental current.

[0018] For blade surface data, combined with environmental temperature and humidity parameters, an icing risk index is established, and blade load distribution unevenness features are extracted; the icing risk index is calculated using an empirical formula, which considers environmental temperature, dew point temperature and relative humidity. When the index is below the set threshold and the blade load increases significantly, it is determined to be high risk. The load distribution unevenness is quantified by calculating the standard deviation between the readings of strain gauges at different positions at the same time.

[0019] For temperature data, a temperature field benchmark model based on physical location is established to extract temperature gradient anomaly and local overheating features; a multivariate linear regression benchmark model is established for each temperature measurement point based on historical normal data and environmental temperature and unit power. Temperature gradient anomaly is identified by calculating the deviation of the temperature difference between adjacent measurement points from the historical normal temperature difference. Local overheating features are defined as a single measurement point temperature that is continuously significantly higher than its benchmark value and lasts for a certain period of time.

[0020] A multi-source data fusion technology is used to combine the above features to construct a comprehensive feature vector of equipment health status.

[0021] The comprehensive feature vector of equipment health status construction step includes: A mechanical vibration intensity feature sub-vector is constructed based on the root mean square value and peak factor of vibration data; A lubrication state feature sub-vector is constructed based on the oil product deterioration rate of oil data; An electrical quality feature sub-vector is constructed based on the harmonic distortion rate and voltage unbalance degree of electrical parameters; A aerodynamic performance feature sub-vector is constructed based on the icing risk index and load distribution unevenness of blade surface data; A thermal state feature sub-vector is constructed based on the temperature gradient anomaly and local overheating features of temperature data; A multi-level feature fusion architecture is used. First, standardization and dimensionality reduction processing is performed within each feature sub-vector, and then a primary fusion feature is constructed through weighted splicing. Within each sub-vector, numerical features are normalized by minimum and maximum, and spectral features are reduced to three to five dimensions by principal component analysis. The primary fusion uses weighted splicing, and the weights are determined based on the information gain of the features in the historical failure data. Features with high importance are given higher weights.

[0022] A feature interaction model based on deep neural network is established to mine the coupling relationship between different feature sub-vectors and generate high-order cross features; a multi-layer perceptron model is constructed with the primary fusion features as input, which automatically learns the complex correlation between vibration and temperature and other subsystem features through hidden layer nonlinear transformation, and the output layer is the mined high-order cross features.

[0023] An operating condition adaptive mechanism is introduced to dynamically adjust the fusion weights of each feature sub-vector according to the current wind speed, power output and environmental conditions; Finally, a comprehensive feature vector of equipment health status is outputted, which includes original features, high-order cross features and operating condition weighted features.

[0024] S3, based on the extracted features, automatic fault identification and positioning is performed, including constructing early warning models at the whole machine level and subsystem level, and realizing fault positioning through the state evaluation index and residual analysis of the model output; The S3 includes a whole machine level and a subsystem level early warning model, and the early warning model specifically includes: Based on the historical normal operation data of the equipment, a whole machine operation benchmark model and a subsystem operation benchmark model are respectively established; in the embodiment of the present application, the historical data of at least six months of failure-free records are collected. The whole machine benchmark model uses Mahalanobis distance to describe the overall correlation between multiple variables. The subsystem benchmark model establishes a univariate or multivariate Gaussian distribution model for each subsystem key parameter such as transmission chain vibration energy and generator current harmonic.

[0025] For the whole machine level, a multi-variable state estimation technology is used to construct a whole machine health state evaluation model, a memory matrix is used to store the multi-variable data relationship under the normal operation state, and a nonlinear state estimation algorithm is used to calculate the residual error between the real-time state and the benchmark state; the memory matrix is composed of thousands of historical normal state vectors. For a new observation vector, the optimal similarity selection method in the multi-variable state estimation technology is used to find the k nearest neighbors from the memory matrix, and an estimated vector is generated by weighted average. The residual error is the difference between the observation vector and the estimated vector in each dimension.

[0026] For the wind wheel subsystem, a performance degradation evaluation model based on aerodynamic characteristics is established, the dynamic relationship between the blade pitch angle, wind speed and power output is analyzed, the aerodynamic efficiency benchmark curve is established, and the aerodynamic efficiency deviation is monitored in real time; below the rated wind speed, the wind speed-power benchmark curve is fitted based on the historical data according to the Betz theory. The ratio of the actual power at the current wind speed to the benchmark power is defined as the aerodynamic efficiency. If the efficiency value continuously falls below the benchmark curve by a certain percentage for a period of time, a performance degradation early warning is triggered.

[0027] For the transmission subsystem, a mechanical fault early warning model based on vibration propagation characteristics is established, vibration sensors are arranged on the gearbox and main bearing, time domain and frequency domain features of the vibration signal are extracted, and a vibration energy distribution benchmark model is established; the benchmark model establishes statistical control limits for the amplitude of characteristic frequencies such as gear box shaft rotation frequency and its harmonics and meshing frequency sidebands. Real-time monitoring of whether the amplitudes of these characteristic frequencies are out of limits, and combining envelope spectrum analysis to determine the bearing fault type.

[0028] For the generator subsystem, an electrical fault early warning model based on electromagnetic characteristics is established, the harmonic components of the stator current and rotor current are analyzed, an electromagnetic characteristic benchmark model is established, and abnormal changes in electrical parameters are monitored.

[0029] Fault positioning is achieved through state evaluation indicators and residual error analysis of the model output, and specifically includes: A dynamic threshold evaluation mechanism based on a sliding window is established, a kernel density estimation algorithm is used to estimate the probability density of the residual sequence under normal state, and a dynamic early warning threshold is determined according to the preset confidence; At the system level, multivariate state estimation technology is used to calculate the residual norm of the real-time observation vector and the memory matrix estimation vector. When the residual norm continuously exceeds the dynamic warning threshold for a preset duration, a system-level warning is triggered. For the subsystem level, a residual contribution analysis model is established. By calculating the contribution of each subsystem's characteristic parameters to the overall residual, the main sources of anomalies are identified. By employing fault propagation path analysis technology and combining the energy transfer relationships between subsystems, primary faults and secondary faults are distinguished, and the root cause of the fault is determined.

[0030] S4. Based on the fault identification and location results, automatically match the corresponding handling solution from the preset fault knowledge base.

[0031] Furthermore, automatically matching corresponding handling solutions from a pre-set fault knowledge base specifically includes: Construct a fault knowledge base based on a graph structure to store the multidimensional relationships between fault modes, fault causes, handling solutions, and historical cases; A graph neural network algorithm is used to calculate the semantic similarity between the current fault features and historical cases in the knowledge base, and a list of processing solutions ranked by matching degree is generated. A case confidence assessment mechanism is introduced, which combines the frequency of failure, the success rate of solution execution, and the equipment operating environment to assign a confidence score to each matched solution; Output the top-K recommended solutions and display the maintenance records and effect evaluations of similar historical cases to support operation and maintenance decisions.

[0032] This invention, through the construction of a distributed monitoring network covering key components of wind turbine generators, achieves comprehensive, real-time acquisition and deep fusion of multi-source status parameters, significantly improving the comprehensiveness and accuracy of equipment status perception. By introducing deep learning-based feature extraction and fault diagnosis models, the system can effectively capture the subtle features of early faults, achieving accurate fault identification and root cause localization, greatly improving the timeliness and reliability of diagnosis.

[0033] System Implementation Examples According to embodiments of the present invention, a real-time monitoring and diagnostic system for wind turbine status is provided. Figure 2 This is a schematic diagram of a real-time monitoring and diagnostic system for wind turbine status according to an embodiment of the present invention. Figure 2 As shown, the wind turbine status real-time monitoring and diagnosis system of this invention specifically includes: Data acquisition module 20 is used to collect status parameters of key components of wind turbine in real time through a distributed monitoring network. The key components include gearbox, main shaft, generator, blades and pitch system. The data processing module 22 is used for data cleaning, alignment and feature extraction of the collected state parameters, and constructs a monitoring system based on a graphical interface framework, and displays the extracted features, key parameter original values and historical change trends in a chart form in real time. The fault positioning module 24 is used for automatic fault identification and positioning based on the extracted features, including construction of a pre-warning model at a whole machine level and a subsystem level, and implementation of fault positioning through state evaluation indexes and residual error analysis of the model output. The processing scheme module 26 is used for automatic matching of a corresponding processing scheme from a pre-set fault knowledge base according to the fault identification and positioning results.

[0034] Embodiment one of the device According to the embodiment of the present application, an electronic device is provided, comprising: a processor; and a memory arranged to store computer executable instructions that, when executed, cause the processor to perform the wind turbine state real-time monitoring and diagnosis method as described above.

[0035] Embodiment two of the device According to the embodiment of the present application, a storage medium is provided for storing computer executable instructions that, when executed, implement the wind turbine state real-time monitoring and diagnosis method as described above.

[0036] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for real-time monitoring and diagnosing the state of a wind turbine, characterized in that The method comprises the following steps: S1, collecting the state parameters of key components of a wind turbine in real time through a distributed monitoring network, wherein the key components include a gearbox, a main shaft, a generator, a blade, and a variable pitch system; S2, performing data cleaning, alignment, and feature extraction on the collected state parameters, and constructing a monitoring system based on a graphical interface framework, and visualizing the extracted features, key parameter original values, and historical change trends in a chart form in real time; S3, performing automatic fault identification and positioning based on the extracted features, including constructing early warning models at the whole machine level and the subsystem level, and realizing fault positioning through state evaluation indexes and residual error analysis output by the models; S4, automatically matching corresponding processing schemes from a pre-set fault knowledge base according to the fault identification and positioning results.

2. The method of claim 1, wherein, The step of collecting the state parameters of key components of a wind turbine in real time through a distributed monitoring network comprises the following steps: monitoring the vibration amplitude and frequency characteristics of bearings through vibration sensors to obtain vibration data; detecting the viscosity and granularity of oil in real time through oil quality sensors to obtain oil data; collecting electrical parameters through electrical parameter sensors adapted to the type of the stator winding of the generator, wherein the electrical parameters include stator current, rotor current, and rotor voltage; monitoring the icing and aerodynamic load state of blades through a monitoring terminal fixed to the curved surface of the blade and integrated with icing thickness and load sensors to obtain blade surface data; tracking the temperature changes of key components in real time through a micro temperature sensor array deployed on the bearings of the gearbox, the stator of the generator, and the winding of the variable pitch motor to obtain temperature data; collecting the three-phase voltage, three-phase current, and temperature parameters of the frequency converter through voltage and current sensors configured in the frequency converter unit to obtain frequency converter operation data.

3. The method of claim 2, wherein, The step of performing data cleaning, alignment, and feature extraction on the collected state parameters comprises the following steps: for vibration data, performing wavelet denoising processing, and extracting the root mean square value, peak factor characteristics in the time domain, and 1 / 3 octave spectrum characteristics in the frequency domain; for oil data, establishing dynamic thresholds based on the viscosity and granularity change trends, identifying abnormal degradation data, and extracting oil degradation rate characteristics; for electrical parameters, extracting current harmonic distortion rate and voltage imbalance degree characteristics through three-phase imbalance degree analysis and harmonic distortion calculation; for blade surface data, establishing an icing risk index in combination with environmental temperature and humidity parameters, and extracting blade load distribution unevenness characteristics; for temperature data, establishing a temperature field reference model based on physical location, and extracting temperature gradient anomalies and local overheating characteristics; using multi-source data fusion technology, combining the above characteristics to construct a comprehensive feature vector of equipment health status.

4. The method of claim 3, wherein, The step of constructing the comprehensive feature vector of equipment health status comprises the following steps: constructing a mechanical vibration intensity feature sub-vector based on the root mean square value and peak factor of vibration data; constructing a lubrication state feature sub-vector based on the oil degradation rate of oil data; constructing an electrical quality feature sub-vector based on the harmonic distortion rate and voltage imbalance degree of electrical parameters; constructing an aerodynamic performance feature sub-vector based on the icing risk index and load distribution unevenness of blade surface data; constructing a thermal state feature sub-vector based on the temperature gradient anomalies and local overheating characteristics of temperature data; The multi-level feature fusion architecture is adopted, first, the standardization and dimension reduction processing are performed in each feature sub-vector, and then the primary fusion features are constructed by weighted splicing; A feature interaction model based on deep neural network is established to mine the coupling relationship between different feature sub-vectors and generate high-order cross features; An operating condition adaptive mechanism is introduced to dynamically adjust the fusion weights of each feature sub-vector according to the current wind speed, power output and environmental conditions; Finally, the device health status comprehensive feature vector is output, which includes original features, high-order cross features and operating condition weighted features.

5. The method of claim 1, wherein, The S3 includes: Based on the historical normal operation data of the device, the whole machine operation reference model and the subsystem operation reference model are established respectively; For the whole machine level, a multi-variable state estimation technology is used to construct a whole machine health status evaluation model, a memory matrix is used to store the multi-variable data relationship in the normal operation state, and a nonlinear state estimation algorithm is used to calculate the residual error between the real-time state and the reference state; For the wind wheel subsystem, a performance degradation evaluation model based on aerodynamic characteristics is established, the dynamic relationship between the blade pitch angle, wind speed and power output is analyzed, the aerodynamic efficiency reference curve is established, and the aerodynamic efficiency deviation is monitored in real time; For the transmission subsystem, a mechanical fault early warning model based on vibration propagation characteristics is established, time domain and frequency domain features of vibration signals are extracted through vibration sensors arranged in the gearbox and main bearing, and a vibration energy distribution reference model is established; For the generator subsystem, an electrical fault early warning model based on electromagnetic characteristics is established, the harmonic components of the stator current and the rotor current are analyzed, the electromagnetic characteristic reference model is established, and the abnormal change of electrical parameters is monitored.

6. The method of claim 1, wherein, The S3 includes: A dynamic threshold evaluation mechanism based on a sliding window is established, a kernel density estimation algorithm is used to estimate the probability density of the residual sequence in the normal state, and a dynamic early warning threshold is determined according to the preset confidence; For the whole machine level, a multi-variable state estimation technology is used to calculate the residual norm of the real-time observation vector and the memory matrix estimation vector, and when the residual norm continuously exceeds the dynamic early warning threshold for a preset time, the whole machine level early warning is triggered; For the subsystem level, a residual contribution degree analysis model is established, the contribution degree of each subsystem feature parameter to the total residual error is calculated to identify the main abnormal source; A fault propagation path analysis technology is used to distinguish between primary faults and secondary faults based on the energy transmission relationship between subsystems, and the root cause of the fault is determined.

7. The method of claim 1, wherein, The S4 includes: A fault knowledge base based on graph structure is constructed to store the multi-dimensional association relationship among fault modes, fault causes, treatment schemes and historical cases; A graph neural network algorithm is used to calculate the semantic similarity between the current fault features and the historical cases in the knowledge base, and a treatment scheme list sorted by matching degree is generated; A case confidence evaluation mechanism is introduced to assign a confidence score to each matching scheme based on the fault occurrence frequency, the success rate of scheme execution and the device operating environment. Output top-K recommendation processing scheme, and associate display similar historical case maintenance record and effect evaluation, provide support for operation and maintenance decision.

8. A wind turbine condition real-time monitoring and diagnosis method, characterized in that, Comprise: Data acquisition module, used for collecting state parameters of key components of wind turbine in real time through distributed monitoring network, the key components include gearbox, main shaft, generator, blade and variable pitch system; Data processing module, used for data cleaning, alignment and feature extraction of collected state parameters, and based on graphical interface framework, build monitoring system, real-time visual display of extracted features, key parameters original value and historical change trend in chart form; Fault positioning module, used for automatic fault identification and positioning based on extracted features, including building early warning model of whole machine level and subsystem level, realizing fault positioning through state evaluation index and residual error analysis output by model; Processing scheme module, used for automatically matching corresponding processing scheme from pre-set fault knowledge base according to fault identification and positioning result. 9.An electronic device comprising: a processor; and a memory arranged to store computer-executable instructions that, when executed, cause the processor to perform the wind turbine state real-time monitoring and diagnosis method of any one of claims 1-8. 10.A storage medium for storing computer-executable instructions that, when executed, implement the wind turbine state real-time monitoring and diagnosis method of any one of claims 1-8.

Images