Intelligent wind power plant fan monitoring system and method based on machine learning and digital twinning

Abstract

The invention discloses an intelligent wind power plant fan monitoring system and method based on machine learning and digital twinning, and relates to the technical field of fan monitoring. Based on the dynamic probability digital twinning technology, the multi-dimensional operation state model of the fan is constructed in real time, and the state distribution and uncertainty quantification result is dynamically updated, so that the real-time operation state of the fan is accurately reflected, and the flexibility and accuracy of state modeling are improved; the machine learning module realizes accurate prediction of key operation parameters through a physical constraint neural network and a deep operation network, and evaluates potential fault risks, so that the equipment fault rate is effectively reduced; the monitoring module can visually display the running state of the fan and calibrate a fault area by using a real-time visualization technology, so that an abnormal problem can be conveniently and quickly positioned; the intelligent optimization control module is based on the deep reinforcement learning technology, dynamic adjustment of fan operation parameters is achieved, the power generation efficiency of the fan is improved, and meanwhile the operation stability and economical efficiency are optimized.

Description

Technical Field

[0001] The present invention relates to the field of wind turbine monitoring technology, and in particular to an intelligent wind farm wind turbine monitoring system and method based on machine learning and digital twins. Background Art

[0002] With the continuous advancement of wind power technology, the scale and complexity of wind farms are also gradually increasing. In modern large-scale wind farms, there are a large number of wind turbines, and the real-time changes in wind turbine operating conditions and environmental conditions have put forward higher requirements for monitoring systems. Traditional wind turbine monitoring systems mainly rely on fixed rules and parameter settings for status monitoring and fault detection. This approach has great limitations and is difficult to cope with the complexity and diversity of wind turbine operating conditions and the diversity of potential faults.

[0003] The development of digital twin technology has provided new possibilities for intelligent monitoring of wind farms. Digital twin technology can reflect the operating status and performance of wind turbines in real time by building a digital model of wind turbines, thereby achieving accurate prediction and diagnosis. In addition, the application of machine learning technology in big data analysis and pattern recognition provides strong support for the intelligent operation and maintenance of wind farms. Through the continuous training and optimization of machine learning algorithms, the monitoring system can realize real-time analysis and status prediction of wind turbine operating data, timely detect potential faults, and avoid equipment damage and downtime.

[0004] However, existing smart wind farm monitoring systems have not yet fully integrated digital twin and machine learning technologies to achieve full life cycle management and fault prediction of wind turbines. Summary of the invention

[0005] In response to the above problems, the present invention proposes an intelligent wind farm wind turbine monitoring system and method based on machine learning and digital twins to improve the accuracy and intelligence level of monitoring, realize accurate monitoring and early warning of wind turbine operating status, and improve the overall operating efficiency and safety of the wind farm.

[0006] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0007] An intelligent wind farm wind turbine monitoring system based on machine learning and digital twins, the system comprising:

[0008] Multi-source data acquisition module, used to collect fan operation data, environmental data and fault-related data, and pre-process and compress the collected data for transmission through distributed edge computing nodes;

[0009] The data processing module is used to build and update the multi-dimensional operating status model of the wind turbine in real time based on the dynamic probabilistic digital twin technology, and generate dynamically updated state distribution data and uncertainty quantification results;

[0010] A machine learning module for predicting key operating parameters of the fan and assessing potential failure risks based on an integrated architecture of a physical constraint neural network and a deep operation network; the key operating parameters include mechanical operating parameters, electrical operating parameters, environmental related parameters, thermodynamic parameters, and failure related parameters;

[0011] The monitoring module is used to display the operating status of the fan through real-time visualization technology, identify potential fault areas, and provide comprehensive monitoring of the fan status;

[0012] Intelligent optimization control module, used to dynamically adjust wind turbine operating parameters through real-time analysis and deep reinforcement learning technology;

[0013] The user interaction module is used to provide users with real-time visualization of the wind turbine operating status, manual control functions and simulation analysis support through a multi-view interface and virtual reality technology, while collecting user feedback to optimize system intelligence.

[0014] As a preferred solution of the present invention, the multi-source data acquisition module includes:

[0015] The fan operation data acquisition unit is used to collect real-time sensor data inside the fan, including speed, vibration, temperature, current, voltage and power generation;

[0016] Environmental data acquisition unit, used to collect external environmental sensor data in real time, including wind speed, wind direction, temperature and humidity, atmospheric pressure and light intensity;

[0017] Fault-related data collection unit, used to collect historical fault logs and marker data of real-time abnormal events;

[0018] The distributed edge computing node unit is used to perform data filtering, data compression, real-time preprocessing and anomaly detection on the wind turbine operation data, environmental data and fault-related data, and to perform timestamp processing on the collected data in combination with time synchronization technology to ensure the consistency of multi-source data.

[0019] As a preferred solution of the present invention, the data processing module includes:

[0020] The high-order outlier removal unit is used for outlier detection and boundary value removal in multi-source data; the outlier detection specifically uses the DBSCAN density clustering algorithm to detect outliers in high-dimensional space, and uses the variational autoencoder to train the normal mode in combination with the time series characteristics to detect data points with high deviation probability; the boundary value removal specifically uses the box plot method to remove extreme values ​​outside the upper and lower quartiles for numerical data, and combines the fan status distribution after real-time update to dynamically remove new outliers;

[0021] The feature extraction unit is used to extract the time-frequency characteristics of the fan vibration signal using short-time Fourier transform, extract the instantaneous time domain characteristics using wavelet transform, and extract the statistics of mean, variance, and skewness to describe the overall change trend of the signal;

[0022] A dimension optimization unit is used to reduce the dimension of the high-dimensional features extracted by the feature extraction unit, reduce redundancy, and retain core information, including kernel PCA dimension reduction and time series embedding; the kernel PCA dimension reduction embeds data into a high-dimensional feature space through nonlinear mapping to extract principal components; the time series embedding uses an autoregressive model or window aggregation technology to convert time series features into embedded representations to capture time series correlations;

[0023] The dynamic probabilistic digital twin unit is used to build a multi-dimensional operating status model of the wind turbine, and dynamically update it using real-time data to generate operating status characteristic data and uncertain quantification results.

[0024] As a preferred solution of the present invention, the method for constructing the multidimensional operating state model of the wind turbine includes:

[0025] Initial distribution modeling: Use Gaussian mixture model to initialize the distribution of wind turbine operating status:

[0026]

[0027] Where P(x|θ) represents the probability density of data point x under model parameter θ; θ is the parameter set of the Gaussian mixture model, including the mean μ i , covariance matrix∑ i and weight w i ; k is the number of Gaussian components; w i is the weight of the i-th Gaussian component, satisfying Indicates that the i-th Gaussian component has mean μ i and the covariance matrix ∑ i ;

[0028] Dynamic covariance adjustment: Introduce the uncertainty matrix U and dynamically adjust the covariance matrix:

[0029] ∑'=∑+λ·U;

[0030] Where, ∑ is the current covariance matrix, which is used to describe the multidimensional distribution of the wind turbine operating status; ∑' is the updated covariance matrix; λ is the dynamic adjustment factor, which is used to control the weight of the uncertainty matrix U on the covariance matrix adjustment;

[0031] Importance Sampling Optimization: Use importance sampling to optimize the model's weight updates:

[0032]

[0033] In the formula, w (i) is the updated weight of particle i; is the weight of particle i at time t-1; P(D t |θ (i) ) is the likelihood function value of particle i; D t is the real-time observation data at the current time t; θ (i) is the model parameter of particle i.

[0034] As a preferred solution of the present invention, the machine learning module includes:

[0035] A data preprocessing unit is used to receive the state distribution data and uncertainty quantification results of the data processing module, convert them into vector representations acceptable to the neural network, and perform normalization processing;

[0036] The operating parameter prediction unit is used to construct a physical constraint neural network, using the blade motion equation as a constraint to predict the key operating parameters of the wind turbine. The loss function of the physical constraint neural network is expressed as:

[0037] ξ=ξ data +γξ physics ;

[0038] Where ξ is the total loss function of the physical constraint neural network; data is the prediction error loss based on the observed data; ξ physics is the residual loss of the partial differential equation based on physical constraints; γ is the trade-off factor;

[0039] A fault risk assessment unit, which is used to extract fault-related features from the state distribution through a deep operation network to perform fault risk assessment;

[0040] The deep operation network maps the multi-dimensional input of wind turbine status into failure risk probability:

[0041] b=κ(u,a);

[0042] Where b is the predicted output of the fault risk; κ represents the operation network, which is used to extract the nonlinear mapping relationship from the input features; u is the state feature vector; a is the operation variable, which describes the external environment influencing factor;

[0043] Combined with the current state distribution, predict the probability of failure:

[0044] P(FailureD t )=∫P(Failureθ)·P(θD t )dθ;

[0045] Where, P(FailureDt ) indicates that in the current data D t The conditional probability of a fan failure under the given model parameter θ is P(Failureθ), which represents the probability of a fan failure under the given model parameter θ. t ) is based on the current data D t The posterior distribution of , which represents the probability density of the model parameter θ;

[0046] Monte Carlo integration is used for numerical solution to output the prediction results of future operating parameters and fault risk assessment report.

[0047] As a preferred solution of the present invention, the monitoring module includes:

[0048] Multi-mode data visualization unit, including time series charts and thermal imaging, which displays the historical trends and real-time changes of key operating parameters through time series charts, and generates thermal imaging of key wind turbine components based on temperature data to intuitively identify possible abnormal areas;

[0049] 3D modeling unit, used to generate a high-resolution 3D model of the wind turbine in combination with digital twin technology, dynamically reflecting the current status of the wind turbine;

[0050] The alarm and status classification unit is used to classify the wind turbine status into "normal", "abnormal" or "emergency" based on key operating parameters and fault risk assessment results. In the "emergency" state, a fault alarm is triggered to highlight the risk area and provide operating suggestions;

[0051] The status report generation unit is used to generate status reports regularly, providing a report format that combines charts and text, including operating parameter trends, health status of key components and fault warning information.

[0052] As a preferred solution of the present invention, the implementation method of the intelligent optimization control module includes:

[0053] Obtain multi-dimensional operating status models and uncertainty quantification results from the data processing module as basic data for control optimization; obtain key operating parameter prediction and fault risk assessment results from the machine learning module to set optimization goals;

[0054] Build a reinforcement learning model and define the state, action, and reward functions: the state refers to the current operating parameters and environmental data of the wind turbine, the action refers to adjusting the operating parameters of the wind turbine, and the reward function should integrate the wind turbine's power generation efficiency, operating stability, and failure risk;

[0055] Generate optimization strategies through deep Q learning or policy gradient algorithm, adjust wind turbine operating parameters in real time according to the optimization strategies, continuously collect real-time data, update multi-dimensional operating status models, and form a closed-loop optimization mechanism;

[0056] Output dynamically optimized control instructions, which are directly applied to fan operation;

[0057] Monitor changes in key operating parameters after optimization in real time, evaluate optimization effects, and provide optimization suggestion records to support adjustments to long-term operating strategies.

[0058] A wind turbine monitoring method of an intelligent wind farm wind turbine monitoring system based on machine learning and digital twins, the method comprising:

[0059] Collect wind turbine operation data, environmental data, and fault-related data, and perform preprocessing and compressed transmission on the collected data through distributed edge computing nodes. The preprocessing operations include data filtering, data compression, real-time preprocessing, and anomaly detection. In addition, the collected data is timestamped in combination with time synchronization technology to ensure the consistency of multi-source data.

[0060] Based on dynamic probabilistic digital twin technology, a multi-dimensional operating status model of the wind turbine is constructed and updated in real time to generate dynamically updated status distribution data and uncertainty quantification results;

[0061] Based on the integrated architecture of physical constraint neural network and deep operation network, the key operating parameters of wind turbines are predicted and potential failure risks are evaluated;

[0062] Display the fan operating status through real-time visualization technology, identify potential fault areas, and provide comprehensive monitoring of the fan status;

[0063] Dynamically adjust wind turbine operating parameters through real-time analysis and deep reinforcement learning technology;

[0064] Through a multi-view interface and virtual reality technology, it provides users with real-time visualization of the wind turbine's operating status, manual control functions and simulation analysis support, while collecting user feedback to optimize system intelligence.

[0065] As a preferred solution of the present invention, the method of constructing and updating the multi-dimensional operation status model of the wind turbine in real time based on the dynamic probabilistic digital twin technology, generating dynamically updated state distribution data and uncertainty quantification results, includes:

[0066] The DBSCAN density clustering algorithm is used to detect outliers in high-dimensional space. Combined with the time series characteristics, the variational autoencoder is used to train the normal mode and detect data points with high probability of deviation. At the same time, the box plot method is used to eliminate extreme values ​​outside the upper and lower interquartile ranges.

[0067] Short-time Fourier transform is used to extract the time-frequency characteristics of the fan vibration signal, wavelet transform is used to extract the instantaneous time domain characteristics, and the statistics of mean, variance and skewness are extracted to describe the overall change trend of the signal;

[0068] Perform dimensionality reduction on the extracted high-dimensional features, including kernel PCA dimensionality reduction and time series embedding, to convert time series features into embedded representations to capture time series correlations;

[0069] Construct a multi-dimensional operating status model of the wind turbine, update the model status through real-time data, and generate operating status characteristic data and uncertainty quantification results.

[0070] As a preferred solution of the present invention, the method of predicting key operating parameters of a wind turbine and evaluating potential failure risks based on the integrated architecture of a physical constraint neural network and a deep operation network includes:

[0071] Operation parameter prediction: construct a physical constraint neural network, use the blade motion equation as a constraint, and predict the key operation parameters of the wind turbine. The loss function of the physical constraint neural network includes the prediction error loss and the residual loss of the partial differential equation based on the physical constraint.

[0072] Fault risk assessment: extract fault-related features from state distribution through deep operation network, predict the conditional probability of wind turbine failure, use Monte Carlo integration to numerically solve future operating parameters, and output fault risk assessment report;

[0073] The method of displaying the operation status of the fan by real-time visualization technology, calibrating potential fault areas, and providing comprehensive monitoring of the fan status includes:

[0074] Use time series charts to display historical trends and real-time changes in key operating parameters, and generate thermal images of key components of the fan based on temperature data to visually identify potential abnormal areas;

[0075] Combine digital twin technology to generate a high-resolution 3D model of the wind turbine, dynamically reflecting the current status of the wind turbine;

[0076] Based on key operating parameters and fault risk assessment results, the wind turbine status is classified as "normal", "abnormal" or "emergency". In the "emergency" state, a fault alarm is triggered and the risk area is highlighted, and operation recommendations are generated;

[0077] Generate operating status reports regularly, providing a report format that combines charts and text, including operating parameter trends, health status of key components, and fault warning information.

[0078] The beneficial effects of the present invention are as follows: through the multi-source data acquisition module, the wind turbine operation data, environmental data and fault-related data can be efficiently collected, and data preprocessing and time synchronization are performed through the edge computing node to ensure the consistency and accuracy of multi-source data, and provide reliable data support for subsequent analysis; the data processing module is based on the dynamic probabilistic digital twin technology to build a multi-dimensional operation state model of the wind turbine in real time, and dynamically update the state distribution and uncertainty quantification results, so as to accurately reflect the real-time operation state of the wind turbine and improve the flexibility and accuracy of state modeling; the machine learning module realizes the accurate prediction of key operation parameters through the physical constraint neural network and the deep operation network, and evaluates the It estimates potential failure risks, provides data-driven support for early warning and optimized operation of wind turbines, and effectively reduces equipment failure rates. The monitoring module uses real-time visualization technology to intuitively display the operating status of the wind turbine and calibrate the fault area, which facilitates the rapid location of abnormal problems and improves fault diagnosis and maintenance efficiency. The intelligent optimization control module is based on deep reinforcement learning technology to achieve dynamic adjustment of wind turbine operating parameters, which improves the power generation efficiency of the wind turbine while optimizing the operating stability and economy. The user interaction module provides real-time visualization, multi-perspective interface and simulation analysis functions, providing a convenient interactive platform for users to operate and feedback the optimization system, significantly improving the intelligence level of the system and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0080] in:

[0081] Figure 1 It is a system module diagram of the present invention;

[0082] Figure 2 This is a wind turbine list view in an embodiment of the present invention;

[0083] Figure 3 A columnar view of a fan in an embodiment of the present invention;

[0084] Figure 4 is a cross-sectional view of a single fan in an embodiment of the present invention;

[0085] Figure 5 Schematic diagram of a fan control instruction interface in an embodiment of the present invention;

[0086] Figure 6 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0087] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0088] like Figure 1 As shown in FIG. 1 , an embodiment of the present invention is provided, which provides an intelligent wind farm wind turbine monitoring system based on machine learning and digital twins, including:

[0089] (1) Multi-source data acquisition module

[0090] Used to collect fan operation data, environmental data and fault-related data, and pre-process and compress the collected data for transmission through distributed edge computing nodes;

[0091] In one embodiment, the multi-source data acquisition module includes:

[0092] The wind turbine operation data acquisition unit is used to collect real-time sensor data inside the wind turbine, including speed, vibration, temperature, current, voltage, power generation and other parameters; high-precision sensors are arranged in the core components of the wind turbine (such as gearbox, main shaft, generator) and blades;

[0093] Environmental data acquisition unit, which deploys a weather station or integrated environmental sensor in the wind farm environment to collect external environmental sensor data in real time, including wind speed, wind direction, temperature and humidity, atmospheric pressure, light intensity and other parameters;

[0094] Fault-related data collection unit, used to collect historical fault logs and marker data of real-time abnormal events;

[0095] Adopting industrial IoT protocols (such as MQTT and CoAP) to achieve efficient data transmission between sensors and edge computing nodes;

[0096] Distributed edge computing node units are used to filter, compress, preprocess and detect anomalies of wind turbine operation data, environmental data and fault-related data in real time. They are combined with time synchronization technology (such as the NTP protocol) to timestamp the collected data to ensure the consistency of multi-source data. Edge device caching and batch transmission strategies are used to reduce the delay and network pressure of real-time transmission.

[0097] Data filtering: remove obviously erroneous data, such as missing values ​​or outliers that exceed the sensor range;

[0098] Data compression: compress the collected data through algorithms (such as LZW and Huffman coding) to reduce transmission bandwidth consumption;

[0099] Real-time preprocessing: preliminary denoising and standardization to reduce the data processing pressure of subsequent modules;

[0100] Anomaly detection: Preliminary identification and labeling of possible abnormal data points through lightweight machine learning models (such as random forests).

[0101] (2) Data processing module

[0102] It is used to build and update the multi-dimensional operating status model of the wind turbine in real time based on dynamic probabilistic digital twin technology, and generate dynamically updated status distribution data and uncertainty quantification results;

[0103] The data processing module includes:

[0104] The high-order outlier removal unit is used for outlier detection and boundary value removal in multi-source data. Outlier detection is to use the DBSCAN density clustering algorithm to detect outliers in high-dimensional space, and to use the variational autoencoder (VAE) to train the normal mode in combination with time series characteristics to detect data points with high probability of deviation. Boundary value removal is to use the box plot method to remove extreme values ​​outside the upper and lower quartiles for numerical data, and to dynamically remove new outliers in combination with the real-time updated fan status distribution, so as to improve the model's adaptability to real-time changes.

[0105] The feature extraction unit is used to extract the time-frequency characteristics of the fan vibration signal using short-time Fourier transform (STFT), and to extract the instantaneous time domain characteristics using wavelet transform to characterize the local change characteristics of the signal, and to extract statistics such as mean, variance, and skewness to describe the overall change trend of the signal;

[0106] The dimension optimization unit is used to reduce the dimension of the high-dimensional features extracted by the feature extraction unit, reduce redundancy, and retain core information, including kernel PCA dimension reduction and time series embedding; kernel PCA dimension reduction embeds data into a high-dimensional feature space through nonlinear mapping and extracts principal components. It is suitable for nonlinear feature extraction of dynamic distribution to enhance the expressiveness of the model; time series embedding uses autoregressive models or window aggregation technology to convert time series features into embedded representations to capture time series correlations;

[0107] The dynamic probabilistic digital twin unit is used to build a multi-dimensional operating status model of the wind turbine, and dynamically update it using real-time data to generate operating status characteristic data and uncertain quantitative results, thereby enhancing the model's ability to describe the operating status.

[0108] In one specific embodiment, a method for constructing a multi-dimensional operating state model of a wind turbine includes:

[0109] Initial distribution modeling: Use Gaussian mixture model (GMM) to initialize the distribution of wind turbine operating status:

[0110]

[0111] Where P(xθ) represents the probability density of data point x under model parameter θ; θ is the parameter set of the Gaussian mixture model, including the mean μ i , covariance matrix∑ i and weight w i ; k is the number of Gaussian components; w i is the weight of the i-th Gaussian component, satisfying Indicates that the i-th Gaussian component has mean μ i and the covariance matrix ∑ i ;

[0112] Dynamic covariance adjustment: Introduce the uncertainty matrix U and dynamically adjust the covariance matrix:

[0113] ∑'=∑+λ·U;

[0114] Where, ∑ is the current covariance matrix, which is used to describe the multidimensional distribution of the wind turbine operating status; ∑' is the updated covariance matrix; λ is the dynamic adjustment factor, which is used to control the weight of the uncertainty matrix U on the covariance matrix adjustment; the uncertainty matrix U reflects the newly added sources of uncertainty in the data distribution (such as real-time data changes);

[0115] Importance Sampling Optimization: Use importance sampling to optimize the model's weight updates:

[0116]

[0117] In the formula, w (i) is the update weight of particle i, reflecting its contribution to the current posterior distribution; is the weight of particle i at time t-1; P(D t |θ (i) ) is the likelihood function value of particle i, indicating the parameter θ (i) In the current data D t The degree of fit under t is the real-time observation data at the current time t; θ (i) is the model parameter of particle i.

[0118] (3) Machine Learning Module

[0119] It is used to predict the key operating parameters of wind turbines and assess potential failure risks based on the integrated architecture of physical constraint neural network (PINN) and deep operation network (DeepONet);

[0120] Key operating parameters include:

[0121] 1) Mechanical operating parameters

[0122] Speed: The rotation speed of the fan blades (unit: rpm), used to reflect the mechanical operating status of the fan;

[0123] Vibration amplitude: The vibration intensity of key components of the fan (such as bearings and gearboxes) is used to detect whether there are mechanical anomalies;

[0124] Yaw angle: The relative angle between the wind turbine and the wind direction, used to evaluate the response of the wind turbine to changes in wind direction;

[0125] 2) Electrical operating parameters

[0126] Power generation: The actual power output of the wind turbine (unit: kW), which directly reflects the power generation efficiency and operating status;

[0127] Current and voltage: The current and voltage of the wind turbine power generation system. Abnormal values ​​may indicate electrical system failure;

[0128] Power factor: The ratio of the effective component to the total component of electrical energy output, used to evaluate the quality of electrical energy;

[0129] 3) Environmental related parameters

[0130] Wind speed and direction: the main environmental factors for wind turbine operation, directly affecting power generation and blade load;

[0131] Ambient temperature: has an important impact on the performance of fan components (such as bearings, lubricating oil, etc.);

[0132] 4) Thermodynamic parameters

[0133] Bearing temperature: The temperature of the bearing of key components, used to monitor whether its operating status is normal;

[0134] Gearbox temperature: The temperature of the fan gearbox, used to determine the lubrication status and wear degree;

[0135] 5) Fault related parameters

[0136] Component stress and fatigue life: stress state and life prediction of key components to assess potential failure risks;

[0137] Transient abnormal parameters: such as load fluctuations, speed jumps, etc., are used to capture abnormal behaviors under non-steady states.

[0138] In a specific embodiment, the machine learning module includes:

[0139] A data preprocessing unit is used to receive the state distribution data and uncertainty quantification results of the data processing module, convert them into vector representations acceptable to the neural network, and perform normalization processing;

[0140] The operating parameter prediction unit is used to construct a physical constraint neural network, using the blade motion equation as a constraint to predict the key operating parameters of the wind turbine. The loss function of the physical constraint neural network is expressed as:

[0141] ξ=ξ data +γξ physics ;

[0142] Where ξ is the total loss function of the physical constraint neural network; data is the prediction error loss based on observed data (data-driven part); ξ physics is the partial differential equation (PDE) residual loss based on physical constraints (physics-driven part); γ is a trade-off factor used to adjust the relative importance of data-driven and physics-driven losses;

[0143] A fault risk assessment unit, which is used to extract fault-related features from the state distribution through a deep operation network to perform fault risk assessment;

[0144] The deep operation network maps the multi-dimensional input of wind turbine status into failure risk probability:

[0145] b=κ(u,a);

[0146] Where b is the predicted output of the fault risk, such as the fault probability or risk score; κ represents the operation network, which is used to extract the nonlinear mapping relationship from the input features; u is the state feature vector, which describes the key characteristics of the current operating state of the wind turbine; a is the operation variable, which describes the external environmental influencing factors;

[0147] Combined with the current state distribution, predict the probability of failure:

[0148] P(FailureD t )=∫P(Failureθ)·P(θD t )dθ;

[0149] Where, P(Failure|D t ) indicates that in the current data D t The conditional probability of a fan failure under the given model parameter θ is P(Failureθ), which represents the probability of a fan failure under the given model parameter θ. t ) is based on the current data D t The posterior distribution of , which represents the probability density of the model parameter θ;

[0150] Monte Carlo integration is used for numerical solution to output the prediction results of future operating parameters and fault risk assessment report.

[0151] (4) Monitoring module

[0152] Used to display the fan operating status through real-time visualization technology, calibrate potential fault areas, and provide comprehensive monitoring of the fan status to improve the accuracy of fan operating status monitoring;

[0153] Specifically, the monitoring module includes:

[0154] Multi-mode data visualization unit, including time series charts and thermal images, which display the historical trends and real-time changes of key operating parameters (such as speed and power generation) through time series charts, and generate thermal images of key wind turbine components based on temperature data to intuitively calibrate possible abnormal areas;

[0155] 3D modeling unit, used to generate a high-resolution 3D model of the wind turbine in combination with digital twin technology, dynamically reflecting the current status of the wind turbine;

[0156] The alarm and status classification unit is used to classify the status of the wind turbine into "normal", "abnormal" or "emergency" based on key operating parameters and fault risk assessment results. In the "emergency" state, a fault alarm is triggered to highlight the risk area and provide operating suggestions (such as reducing the speed and checking components);

[0157] The status report generation unit is used to regularly generate status reports, including operating parameter trends, key component health status and fault warning information, and provides a report format that combines charts and text to facilitate users to quickly understand.

[0158] (5) Intelligent optimization control module

[0159] Used to dynamically adjust wind turbine operating parameters through real-time analysis and deep reinforcement learning technology to improve power generation efficiency and operating stability;

[0160] Furthermore, the implementation method of the intelligent optimization control module includes:

[0161] Obtain multi-dimensional operating state models and uncertainty quantification results from the data processing module as basic data for control optimization; obtain key operating parameter prediction and fault risk assessment results from the machine learning module to set optimization goals (such as minimizing vibration amplitude and maximizing power generation efficiency);

[0162] Build a reinforcement learning model and define the state, action, and reward functions: the state refers to the current operating parameters and environmental data of the wind turbine, the action refers to adjusting the operating parameters of the wind turbine (such as blade angle and speed), and the reward function should integrate the wind turbine's power generation efficiency, operating stability, and failure risk;

[0163] Generate optimization strategies through deep Q-learning or policy gradient algorithm, adjust wind turbine operating parameters (such as blade angle and power generation) in real time according to the optimization strategies, continuously collect real-time data, update multi-dimensional operating status models, and form a closed-loop optimization mechanism;

[0164] Output dynamically optimized control instructions, which are directly applied to fan operation;

[0165] Monitor changes in key operating parameters after optimization in real time, evaluate optimization effects (such as reduced energy consumption and reduced failure risk), and provide optimization suggestion records to support adjustments to long-term operating strategies.

[0166] (6) User Interaction Module

[0167] It is used to provide users with real-time visualization of wind turbine operating status, manual control functions and simulation analysis support through a multi-perspective interface and virtual reality (VR) technology, while collecting user feedback to optimize system intelligence.

[0168] Figure 2-Figure 5The following are some interface diagrams in the embodiments of the present invention. The fan list view supports the user to customize the measurement points to be monitored, and the display column order of the measurement points can be defined; the measurement points can be sorted by the maximum or minimum value, so that the user can focus on monitoring a certain measurement point. The fan column view can be used to observe which fan power is not reached when the wind speed reaches the full wind speed, and then the customer records it manually. In the single fan interface, the user can view more monitoring information, statistical information, and related charts of the current fan, and use the corresponding shortcut functions to achieve detailed analysis, refined management, etc. Click the search box fan name, double-click the matrix block, and double-click the list to navigate to the single fan view, view more monitoring information, statistical information, and related charts of the current fan, and use the corresponding functions of the current fan. In the single fan interface, the user can select "classic interface", "section view interface" and "benchmark interface". The default view is "classic interface". Click the top icon to switch the corresponding interface. In the classic interface, you can view more monitoring information, statistical information, and related charts of the current fan, use the corresponding functions of the current fan, and play a role in detailed analysis of the current fan. The classic interface is divided into the equipment overview area, real-time trend curve area, real-time data area, and alarm status word area. The cross-sectional view interface is mainly divided into the basic information area, status area, view switching area, component area, cross-sectional view area, measurement point monitoring area, control area, alarm / status area, and basic indicator area. In the single fan interface, you can control the current fan, and the interaction is displayed directly on the interface as much as possible with control instructions, so that users can use it directly. Click the reset, start, stop, lock, and more buttons to pop up the control authority authentication pop-up box first. After passing the authentication, you can perform the corresponding operations. For example, if you turn on "equipment control guardian verification", the "guardian verification" panel will pop up. After the verification is passed, the control instruction can be issued. It supports the fan control log recording function to ensure that each control instruction issued can be traced back to the time of issuance, the person who issued it, the control reason and other information; it supports the instruction execution effect tracking function, and refreshes the key measurement point information of the fan in real time during the execution of the control instruction, such as voltage, current, generator speed, fan status, etc.

[0169] like Figure 6 As shown, another embodiment of the present invention provides a method for monitoring wind turbines in an intelligent wind farm based on machine learning and digital twins. Based on the intelligent wind farm wind turbine monitoring system based on machine learning and digital twins as described above, the method includes the following steps:

[0170] S1: Collect wind turbine operation data, environmental data and fault-related data, and perform preprocessing and compressed transmission on the collected data through distributed edge computing nodes. The preprocessing operations include data filtering, data compression, real-time preprocessing and anomaly detection. In addition, the collected data is timestamped in combination with time synchronization technology to ensure the consistency of multi-source data.

[0171] S2: Based on the dynamic probabilistic digital twin technology, a multi-dimensional operating status model of the wind turbine is constructed and updated in real time to generate dynamically updated state distribution data and uncertainty quantification results;

[0172] S2.1: Use the DBSCAN density clustering algorithm to detect outliers in high-dimensional space. Combined with the time series characteristics, use the variational autoencoder to train the normal mode and detect data points with high probability of deviation. At the same time, use the box plot method to eliminate extreme values ​​outside the upper and lower interquartile ranges.

[0173] S2.2: Use short-time Fourier transform to extract the time-frequency characteristics of the fan vibration signal, use wavelet transform to extract the instantaneous time domain characteristics, and extract statistics (mean, variance, skewness) to describe the overall change trend of the signal;

[0174] S2.3: Perform dimensionality reduction processing on the extracted high-dimensional features, including kernel PCA dimensionality reduction and time series embedding, converting the time series features into embedded representations to capture the temporal correlation;

[0175] S2.4: Construct a multi-dimensional operating status model of the wind turbine, and update the model status through real-time data to generate operating status characteristic data and uncertainty quantification results.

[0176] S3: Based on the integrated architecture of physical constraint neural network and deep operation network, the key operating parameters of the wind turbine are predicted and the potential failure risk is evaluated;

[0177] S3.1: Operation parameter prediction, construct a physical constraint neural network, use the blade motion equation as a constraint, and predict the key operating parameters of the wind turbine. The loss function of the physical constraint neural network includes the prediction error loss and the residual loss of the partial differential equation based on the physical constraint;

[0178] S3.2: Fault risk assessment, extract fault-related features from the state distribution through a deep operation network, and predict the conditional probability of wind turbine failure. Use Monte Carlo integration to numerically solve future operating parameters and output a fault risk assessment report.

[0179] S4: Displays the fan operating status through real-time visualization technology, identifies potential fault areas, and provides comprehensive monitoring of the fan status;

[0180] S4.1: Display the historical trends and real-time changes of key operating parameters in time series charts, and generate thermal images of key components of the wind turbine based on temperature data to visually identify potential abnormal areas;

[0181] S4.2: Combine digital twin technology to generate a high-resolution 3D model of the wind turbine to dynamically reflect the current status of the wind turbine;

[0182] S4.3: Based on the key operating parameters and fault risk assessment results, the wind turbine status is classified as "normal", "abnormal" or "emergency" status. In the "emergency" status, a fault alarm is triggered and the risk area is highlighted, and operation recommendations are generated;

[0183] S4.4: Generate operating status reports regularly, providing a report format that combines charts and text, including operating parameter trends, health status of key components, and fault warning information.

[0184] S5: Dynamically adjust wind turbine operating parameters through real-time analysis and deep reinforcement learning technology;

[0185] S6: Through a multi-view interface and virtual reality technology, it provides users with real-time visualization of the wind turbine operating status, manual control functions and simulation analysis support, while collecting user feedback to optimize system intelligence.

[0186] In summary, the present invention collects fan operation data, environmental data and fault-related data through a multi-source data acquisition module, and uses distributed edge computing nodes to pre-process the data in real time to ensure the consistency and integrity of the data. This module realizes the efficient collection and processing of multi-type and multi-dimensional data, laying the foundation for the subsequent fan operation status modeling and analysis.

[0187] The data processing module uses dynamic probabilistic digital twin technology to build and update the multi-dimensional operating status model of the wind turbine in real time, generating dynamically updated state distribution data and uncertainty quantification results. This function can accurately reflect the operating status of the wind turbine, timely capture the dynamic changes of operating parameters, and provide a reliable basis for subsequent state prediction and fault assessment.

[0188] The machine learning module uses a physical constraint neural network and a deep operation network integrated architecture to accurately predict the key operating parameters of the wind turbine (such as mechanical operating parameters, electrical operating parameters, environmental related parameters, thermodynamic parameters and fault related parameters), and can also assess potential failure risks. This module can promptly identify abnormal conditions in wind turbine operation and discover potential failure risks in advance, thereby effectively reducing equipment failure rates and downtime, and improving the overall operating efficiency of wind farms.

[0189] The monitoring module uses real-time visualization technology to intuitively display the operating status of the fan, identify potential fault areas, and provide comprehensive monitoring of the fan operating status. By combining multi-mode visualization (such as thermal imaging, real-time trend charts, etc.), users can quickly locate the fault area and take corresponding maintenance measures, improving the efficiency of operation and maintenance.

[0190] The intelligent optimization control module combines deep reinforcement learning technology to dynamically adjust the operating parameters of the wind turbine based on real-time analysis and multi-dimensional operating status data. By optimizing the wind turbine operation strategy, the module effectively improves the power generation efficiency of the wind turbine, reduces operating costs, and enhances the operating stability of the wind farm.

[0191] The user interaction module provides users with a real-time visual display of the wind turbine operating status through a multi-view interface and virtual reality technology, and supports manual control functions and simulation analysis. Users can intuitively understand the operating status of the wind turbine through the interactive interface and quickly execute control commands. At the same time, the user data fed back can further optimize the system's intelligence and user experience.

[0192] This system fully combines digital twin technology and machine learning technology, breaks through the limitations of traditional monitoring systems, and realizes complete closed-loop management from wind turbine data collection, state modeling, fault prediction to dynamic optimization control. Through precise monitoring and prediction functions, the present invention not only improves the reliability and safety of wind turbines, but also reduces the cost of manual operation and maintenance, providing strong support for the intelligent operation and maintenance of wind farms.

[0193] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An intelligent wind farm wind turbine monitoring system based on machine learning and digital twins, characterized in that: The system includes: a multi-source data acquisition module for collecting fan operation data, environmental data and fault-related data, and performing pre-processing operations and compression transmission on the collected data through distributed edge computing nodes; The data processing module is used to build and update the multi-dimensional operating status model of the wind turbine in real time based on the dynamic probabilistic digital twin technology, and generate dynamically updated state distribution data and uncertainty quantification results; A machine learning module for predicting key operating parameters of the fan and assessing potential failure risks based on an integrated architecture of a physical constraint neural network and a deep operation network; the key operating parameters include mechanical operating parameters, electrical operating parameters, environmental related parameters, thermodynamic parameters, and failure related parameters; The monitoring module is used to display the operating status of the fan through real-time visualization technology, identify potential fault areas, and provide comprehensive monitoring of the fan status; Intelligent optimization control module, used to dynamically adjust wind turbine operating parameters through real-time analysis and deep reinforcement learning technology; The user interaction module is used to provide users with real-time visualization of the wind turbine operating status, manual control functions and simulation analysis support through a multi-view interface and virtual reality technology, while collecting user feedback to optimize system intelligence.

2. The intelligent wind farm wind turbine monitoring system based on machine learning and digital twins according to claim 1, characterized in that: The multi-source data acquisition module comprises: The fan operation data acquisition unit is used to collect real-time sensor data inside the fan, including speed, vibration, temperature, current, voltage and power generation; Environmental data acquisition unit, used to collect external environmental sensor data in real time, including wind speed, wind direction, temperature and humidity, atmospheric pressure and light intensity; Fault-related data collection unit, used to collect historical fault logs and marker data of real-time abnormal events; The distributed edge computing node unit is used to perform data filtering, data compression, real-time preprocessing and anomaly detection on the wind turbine operation data, environmental data and fault-related data, and to perform timestamp processing on the collected data in combination with time synchronization technology to ensure the consistency of multi-source data.

3. The intelligent wind farm wind turbine monitoring system based on machine learning and digital twins according to claim 1, characterized in that: The data processing module comprises: The high-order outlier removal unit is used for outlier detection and boundary value removal in multi-source data; the outlier detection specifically uses the DBSCAN density clustering algorithm to detect outliers in high-dimensional space, and uses the variational autoencoder to train the normal mode in combination with the time series characteristics to detect data points with high deviation probability; the boundary value removal specifically uses the box plot method to remove extreme values ​​outside the upper and lower quartiles for numerical data, and combines the fan status distribution after real-time update to dynamically remove new outliers; The feature extraction unit is used to extract the time-frequency characteristics of the fan vibration signal using short-time Fourier transform, extract the instantaneous time domain characteristics using wavelet transform, and extract the statistics of mean, variance, and skewness to describe the overall change trend of the signal; A dimension optimization unit is used to reduce the dimension of the high-dimensional features extracted by the feature extraction unit, reduce redundancy, and retain core information, including kernel PCA dimension reduction and time series embedding; the kernel PCA dimension reduction embeds data into a high-dimensional feature space through nonlinear mapping to extract principal components; the time series embedding uses an autoregressive model or window aggregation technology to convert time series features into embedded representations to capture time series correlations; The dynamic probabilistic digital twin unit is used to build a multi-dimensional operating status model of the wind turbine, and dynamically update it using real-time data to generate operating status characteristic data and uncertain quantification results.

4. The intelligent wind farm wind turbine monitoring system based on machine learning and digital twins according to claim 3, characterized in that: The method for constructing the multidimensional operating state model of the wind turbine includes: Initial distribution modeling: Use Gaussian mixture model to initialize the distribution of wind turbine operating status: In the formula, P(x | θ) represents the probability density of data point x under model parameters θ; θ is the parameter set of the Gaussian mixture model, including the mean μ i , covariance matrix∑ i and weight w i ; k is the number of Gaussian components; w i is the weight of the i-th Gaussian component, satisfying Indicates that the i-th Gaussian component has mean μ i and the covariance matrix ∑ i ; Dynamic covariance adjustment: Introduce the uncertainty matrix U and dynamically adjust the covariance matrix: ∑'=∑+λ·U; Where, ∑ is the current covariance matrix, which is used to describe the multidimensional distribution of the wind turbine operating status; ∑' is the updated covariance matrix; λ is the dynamic adjustment factor, which is used to control the weight of the uncertainty matrix U on the covariance matrix adjustment; Importance Sampling Optimization: Use importance sampling to optimize the model's weight updates: In the formula, w (i) is the updated weight of particle i; is the weight of particle i at time t-1; P(D t| θ (i) ) is the likelihood function value of particle i; D t is the real-time observation data at the current time t; θ (i) is the model parameter of particle i.

5. The intelligent wind farm wind turbine monitoring system based on machine learning and digital twins according to claim 4, characterized in that: The machine learning module includes: A data preprocessing unit is used to receive the state distribution data and uncertainty quantification results of the data processing module, convert them into vector representations acceptable to the neural network, and perform normalization processing; The operating parameter prediction unit is used to construct a physical constraint neural network, using the blade motion equation as a constraint to predict the key operating parameters of the wind turbine. The loss function of the physical constraint neural network is expressed as: ξ=ξ data +gx physics ; Where ξ is the total loss function of the physical constraint neural network; data is the prediction error loss based on the observed data; ξ physics is the residual loss of the partial differential equation based on physical constraints; γ is the trade-off factor; A fault risk assessment unit, which is used to extract fault-related features from the state distribution through a deep operation network to perform fault risk assessment; The deep operation network maps the multi-dimensional input of wind turbine status into failure risk probability: b=κ(u,a); Where b is the predicted output of the fault risk; κ represents the operation network, which is used to extract the nonlinear mapping relationship from the input features; u is the state feature vector; a is the operation variable, which describes the external environment influencing factor; Combined with the current state distribution, predict the probability of failure: P(FailureD t )=∫P(Failureθ)·P(θD t )dθ; In the formula, P(Failure | D t ) indicates that in the current data D t The conditional probability of a fan failure under the given model parameter θ is P(Failureθ), which represents the probability of a fan failure under the given model parameter θ. t ) is based on the current data D t The posterior distribution of , which represents the probability density of the model parameter θ; Monte Carlo integration is used for numerical solution to output the prediction results of future operating parameters and fault risk assessment report.

6. The intelligent wind farm wind turbine monitoring system based on machine learning and digital twins according to claim 1, characterized in that: The monitoring module comprises: Multi-mode data visualization unit, including time series charts and thermal imaging, which displays the historical trends and real-time changes of key operating parameters through time series charts, and generates thermal imaging of key wind turbine components based on temperature data to intuitively identify possible abnormal areas; The 3D modeling unit is used to generate a high-resolution 3D model of the wind turbine in combination with the digital twin technology to dynamically reflect the current status of the wind turbine. The alarm and status classification unit is used to classify the status of the wind turbine into "normal", "abnormal" or "emergency" based on key operating parameters and fault risk assessment results. In the "emergency" state, a fault alarm is triggered to highlight the risk area and provide operating suggestions. The status report generation unit is used to generate status reports regularly, providing a report format that combines charts and text, including operating parameter trends, health status of key components and fault warning information.

7. The intelligent wind farm wind turbine monitoring system based on machine learning and digital twins according to claim 1, characterized in that: The implementation method of the intelligent optimization control module includes: Obtain multi-dimensional operating status models and uncertainty quantification results from the data processing module as basic data for control optimization; obtain key operating parameter prediction and fault risk assessment results from the machine learning module to set optimization goals; Build a reinforcement learning model and define the state, action, and reward functions: the state refers to the current operating parameters and environmental data of the wind turbine, the action refers to adjusting the operating parameters of the wind turbine, and the reward function should integrate the wind turbine's power generation efficiency, operating stability, and failure risk; Generate optimization strategies through deep Q learning or policy gradient algorithm, adjust wind turbine operating parameters in real time according to the optimization strategies, continuously collect real-time data, update multi-dimensional operating status models, and form a closed-loop optimization mechanism; Output dynamically optimized control instructions, which are directly applied to fan operation; Monitor changes in key operating parameters after optimization in real time, evaluate optimization effects, and provide optimization suggestion records to support adjustments to long-term operating strategies.

8. A wind turbine monitoring method of an intelligent wind farm wind turbine monitoring system based on machine learning and digital twins according to any one of claims 1 to 7, characterized in that: The method comprises: Collect wind turbine operation data, environmental data, and fault-related data, and perform preprocessing and compressed transmission on the collected data through distributed edge computing nodes. The preprocessing operations include data filtering, data compression, real-time preprocessing, and anomaly detection. In addition, the collected data is timestamped in combination with time synchronization technology to ensure the consistency of multi-source data. Based on dynamic probabilistic digital twin technology, a multi-dimensional operating status model of the wind turbine is constructed and updated in real time to generate dynamically updated status distribution data and uncertainty quantification results; Based on the integrated architecture of physical constraint neural network and deep operation network, the key operating parameters of wind turbines are predicted and potential failure risks are evaluated; Display the fan operating status through real-time visualization technology, identify potential fault areas, and provide comprehensive monitoring of the fan status; Dynamically adjust wind turbine operating parameters through real-time analysis and deep reinforcement learning technology; Through a multi-view interface and virtual reality technology, it provides users with real-time visualization of the wind turbine's operating status, manual control functions and simulation analysis support, while collecting user feedback to optimize system intelligence.

9. The wind turbine monitoring method of the intelligent wind farm wind turbine monitoring system based on machine learning and digital twins according to claim 8, characterized in that: The method of constructing and updating a multi-dimensional operating state model of a wind turbine based on dynamic probabilistic digital twin technology in real time, generating dynamically updated state distribution data and uncertainty quantification results, includes: The DBSCAN density clustering algorithm is used to detect outliers in high-dimensional space. Combined with the time series characteristics, the variational autoencoder is used to train the normal mode and detect data points with high probability of deviation. At the same time, the box plot method is used to eliminate extreme values ​​outside the upper and lower interquartile ranges. Short-time Fourier transform is used to extract the time-frequency characteristics of the fan vibration signal, wavelet transform is used to extract the instantaneous time domain characteristics, and the statistics of mean, variance and skewness are extracted to describe the overall change trend of the signal; Perform dimensionality reduction on the extracted high-dimensional features, including kernel PCA dimensionality reduction and time series embedding, to convert time series features into embedded representations to capture time series correlations; Construct a multi-dimensional operating status model of the wind turbine, update the model status through real-time data, and generate operating status characteristic data and uncertainty quantification results.

10. The wind turbine monitoring method of the intelligent wind farm wind turbine monitoring system based on machine learning and digital twins according to claim 9, characterized in that: The physical constraint neural network and deep operation network integrated architecture are used to predict key operating parameters of the wind turbine and evaluate potential failure risks. The method includes: Operation parameter prediction: construct a physical constraint neural network, use the blade motion equation as a constraint, and predict the key operation parameters of the wind turbine. The loss function of the physical constraint neural network includes the prediction error loss and the residual loss of the partial differential equation based on the physical constraint. Fault risk assessment: extract fault-related features from state distribution through deep operation network, predict the conditional probability of wind turbine failure, use Monte Carlo integration to numerically solve future operating parameters, and output fault risk assessment report; The method of displaying the operation status of the fan by real-time visualization technology, calibrating potential fault areas, and providing comprehensive monitoring of the fan status includes: Use time series charts to display historical trends and real-time changes in key operating parameters, and generate thermal images of key components of the fan based on temperature data to visually identify potential abnormal areas; Combine digital twin technology to generate a high-resolution 3D model of the wind turbine, dynamically reflecting the current status of the wind turbine; Based on key operating parameters and fault risk assessment results, the wind turbine status is classified as "normal", "abnormal" or "emergency". In the "emergency" state, a fault alarm is triggered and the risk area is highlighted, and operation recommendations are generated; Generate operating status reports regularly, providing a report format that combines charts and text, including operating parameter trends, health status of key components, and fault warning information.

Images