Blade state monitoring device for wind driven generator and wind driven generator

Abstract

The invention discloses a blade state monitoring device for a wind driven generator and the wind driven generator, and relates to the field of data monitoring, and the blade state monitoring device comprises a device integration multi-mode sensor module, a self-cleaning deicing module, a lightning stroke monitoring module, an edge calculation module, an anti-typhoon monitoring module, an energy supply module and a communication gateway module. Strain, vibration and temperature sensors are arranged at the root, the tip and the surface of the blade, and an icing detector is arranged at the front edge. And deicing is realized through hot air injection and the nano coating. And a lightning sensor and ultraviolet imaging are used for positioning lightning stroke. And an AI chip is arranged in the edge computing node to diagnose damage. The fiber grating sensor monitors typhoon strain. The solar film and the miniature wind driven generator supply power. And the communication module supports LoRaWAN and Beidou short messages. And the far-end intelligent diagnosis module predicts the service life through machine learning and provides maintenance suggestions. The system has the advantages that the blade state is accurately monitored in real time through the integrated multi-mode sensor and module technology, operation safety is guaranteed, the service life of equipment is prolonged, and efficient utilization of clean energy is facilitated.

Description

Technical Field

[0001] This invention relates to the field of data monitoring, and in particular to a blade condition monitoring device for wind turbines and a wind turbine. Background Technology

[0002] Onshore wind turbines, due to their clean and efficient characteristics, have become one of the pillars of energy supply. This type of equipment effectively improves the reliability and maintenance efficiency of wind turbines, ensures a stable supply of clean energy, and provides strong technical support for sustainable development.

[0003] Current devices on the market have limited sensor functionality, confined to strain or vibration monitoring, lacking the comprehensive data acquisition capabilities of multi-modal sensor arrays. This inability to fully capture blade conditions results in insufficient diagnostic accuracy. Secondly, most devices lack self-cleaning and de-icing functions, requiring manual intervention to address icing issues, which is inefficient and costly. This device, however, utilizes hot air jets and a hydrophobic coating to automatically address ice accumulation. Furthermore, existing devices have limited capabilities in monitoring lightning damage, struggling to accurately pinpoint strike points and analyze current intensity, increasing maintenance complexity. Regarding typhoon resistance, current devices typically lack dedicated fiber optic grating sensors, making real-time buckling strain monitoring difficult and hindering their ability to cope with extreme wind loads. In terms of power supply, most devices rely on external power sources, lacking independent power supply designs for solar and micro-wind power generation, resulting in insufficient stability. For communication, most devices use a single protocol, making data transmission vulnerable to disruptions during network outages. This device, however, features a multi-protocol gateway supporting LoRaWAN, satellite, and BeiDou communication, ensuring data reliability. Finally, commercially available devices have weak intelligent diagnostic capabilities and lack edge computing and machine learning support, making it difficult to achieve accurate lifespan prediction and optimized maintenance decisions. Their overall technical level is significantly lower than that of this device, making it difficult to meet the high-efficiency operation requirements of wind turbines in complex environments. Summary of the Invention

[0004] To improve existing devices, a blade condition monitoring device and a wind turbine generator are provided. This device integrates technologies such as multimodal sensors, edge computing, self-cleaning de-icing, lightning strike monitoring, typhoon-resistant structure monitoring, and intelligent diagnostics. It can monitor the blade condition in real time and accurately, ensuring operational safety, optimizing maintenance efficiency, extending equipment life, and promoting the efficient use of clean energy.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A blade condition monitoring device for wind turbines, comprising:

[0007] Multimodal sensor array module: The module includes a piezoelectric strain sensor array installed at the root of the blade, a triaxial vibration sensor at the tip of the blade, a distributed fiber optic temperature sensor on the blade surface, and a millimeter-wave radar icing detector at the leading edge of the blade.

[0008] Self-cleaning de-icing module: The module is integrated into the microporous hot air jet unit and hydrophobic nano-coating at the leading edge of the blade;

[0009] Lightning damage monitoring module: The module includes a lightning current sensor embedded in the carbon fiber skin of the blade and an arc locator based on ultraviolet imaging, which can detect the lightning strike location and current intensity in real time.

[0010] Edge computing node module: The module is deployed in a waterproof and corrosion-resistant sealed enclosure inside the nacelle, and has a built-in AI chip for running blade damage diagnosis algorithms;

[0011] Typhoon-resistant structure monitoring module: The module is equipped with an FBG fiber grating sensor array at the main beam position of the blade to monitor the buckling strain threshold under typhoon conditions in real time.

[0012] Power supply module: The module includes a flexible solar film and a vertical axis micro wind turbine installed on the top of the nacelle, providing an independent power supply for the sensor network;

[0013] Multi-protocol communication gateway module: The module supports dual-link redundant transmission of LoRaWAN and satellite communication, and automatically switches to BeiDou short message communication when the ground network is interrupted due to a typhoon;

[0014] Intelligent diagnostic module: The module is a machine learning model deployed on a remote server. The input layer receives data from each module, and the output layer generates blade life prediction and maintenance decision suggestions.

[0015] Preferably, the multimodal sensor array module specifically includes:

[0016] Sensor Unit: The unit includes a piezoelectric strain sensor array installed at the root of the blade, a triaxial vibration sensor at the tip of the blade, a distributed fiber optic temperature sensor on the blade surface, and a millimeter-wave radar icing detector at the leading edge of the blade.

[0017] Unit data integration unit: The unit integrates and preprocesses the data collected by each sensor unit and transmits it to the intelligent diagnostic module.

[0018] Preferably, the self-cleaning de-icing module specifically includes:

[0019] Micro-hole hot air injection unit: The unit is integrated into the leading edge of the blade and connected to the vortex heating device in the nacelle through the air guide pipe arranged in the hollow blade beam. A micro-hole array made of high-strength composite material is pre-embedded in the leading edge of the blade. A high-temperature and high-pressure resistant air guide pipe is arranged in the hollow blade beam, extending from the root of the blade to the leading edge micro-hole array. The vortex heating device located in the nacelle adopts the principle of electromagnetic induction and uses a fan to press hot air into the air guide pipe. The micro-hole array is connected to a micro solenoid valve to control the hot air injection area and flow rate.

[0020] Hydrophobic nanocoating: The coating is a superhydrophobic nanocoating applied by plasma spraying to cover the leading edge of the blade.

[0021] Preferably, the lightning damage monitoring module specifically includes:

[0022] Sensor Unit: The unit monitors the lightning current passing through the carbon fiber skin of the blade in real time through a broadband lightning current sensor, captures the peak value, frequency and duration of the current waveform, and detects the ultraviolet radiation of the electric arc caused by the lightning strike through an arc locator based on ultraviolet imaging technology to locate the three-dimensional spatial coordinates of the lightning strike point.

[0023] Data processing unit: The unit amplifies, filters, and performs analog-to-digital conversion on the signal collected by the broadband lightning current sensor to generate analyzable current waveform data, and simultaneously synchronizes it with the ultraviolet imaging positioning data.

[0024] Data transmission unit: The unit transmits the processed lightning current waveform, spatial coordinates and analysis results to the intelligent diagnostic module through a standard communication interface.

[0025] Preferably, the edge computing node module specifically includes:

[0026] The edge computing node module is deployed in a waterproof and corrosion-resistant sealed enclosure inside the nacelle. It has a built-in AI chip that runs a blade damage diagnosis algorithm. The algorithm includes a strain-vibration time-frequency joint analysis model and an ice thickness inversion model.

[0027] Strain-Vibration Time-Frequency Joint Analysis Model Unit: The unit performs short-time Fourier transform on the data from the piezoelectric strain sensor at the blade root and the triaxial vibration sensor at the blade tip, performs time-frequency decomposition on the strain and vibration signals, extracts the time-frequency features of strain and vibration, and diagnoses blade structural damage and abnormal vibration based on the built-in damage diagnosis model of the convolutional neural network.

[0028] Ice thickness inversion model unit: The unit calculates the ice thickness based on the intensity and phase change of the reflected signal from millimeter-wave radar using an electromagnetic wave propagation model. Combined with temperature data, it constructs an ice accretion physical model. Based on the change of the dielectric constant of ice with temperature, it corrects the inversion results. Through an inversion algorithm, it maps the radar signal and temperature data to the ice thickness distribution, generating a two-dimensional thermal map of the leading edge ice on the blade.

[0029] Preferably, the typhoon-resistant structure monitoring module specifically includes:

[0030] Data acquisition unit: The unit is used to receive and preprocess dynamic buckling strain data acquired by the FBG fiber grating sensor array;

[0031] Dynamic buckling strain analysis unit: The unit uses time-domain analysis and frequency-domain analysis to extract the periodicity, amplitude and frequency characteristics of dynamic buckling strain, generates the time-frequency characteristic map of strain signal through short-time Fourier transform, calculates the buckling critical load based on the pre-constructed mechanical model of blade main beam based on finite element analysis, and trains the mapping relationship between strain characteristics and buckling risk through long short-time memory network to predict the buckling risk level in real time.

[0032] Wind speed classification threshold association unit: The unit sets strain thresholds corresponding to each typhoon level based on historical experimental data, and associates dynamic buckling strain data with typhoon wind speed classification thresholds. Through a neural network model, it predicts the strain trend of the blades under different wind speed levels and generates risk curves.

[0033] Preferably, the power supply module specifically includes:

[0034] Energy Harvesting and Management Unit: This unit is used to monitor and record the energy harvesting process of the flexible CIGS solar thin film and the vertical axis Savonius micro wind turbine in real time, and to optimize the energy output efficiency.

[0035] Intelligent power scheduling unit: The unit monitors the power consumption requirements of the multimodal sensor array module, lightning damage monitoring module, edge computing node module and typhoon-resistant structure monitoring module in real time based on historical data, and establishes a load priority model. Based on the priority, a greedy algorithm is used to divide the energy allocation ratio. The operating point is dynamically adjusted for CIGS solar thin film and Savonius wind turbine through the maximum power point tracking algorithm to obtain the maximum energy harvesting efficiency.

[0036] Power supply stability control unit: The unit is used to ensure that the power supply module provides stable voltage and current to the sensor network, and to prevent power interruption or fluctuation from affecting the operation of the equipment.

[0037] Preferably, the multi-protocol communication gateway module specifically includes:

[0038] Communication Protocol Unit: This unit is used to manage the dual-link protocol of LoRaWAN and satellite communication. Through a dynamic link selection algorithm, it monitors the signal quality of the LoRaWAN link in real time and automatically switches to the satellite communication link when the LoRaWAN link is interrupted.

[0039] Beidou Short Message Communication Unit: When both the LoRaWAN link and the satellite communication link are unavailable, the unit activates the Beidou Short Message Communication Protocol and prioritizes the transmission of high-priority data.

[0040] Data processing and fusion unit: The unit preprocesses, fuses, and prioritizes multi-source sensor data. Through data classification and categorization, the data is divided into real-time alarms, periodic reports, and log data, and integrated to generate a comprehensive status report.

[0041] Preferably, the intelligent diagnostic module specifically includes:

[0042] Data integration and processing unit: This unit performs real-time batch data collection on data from each module, generates time-series and non-time-series data after time synchronization and feature extraction, and constructs a dataset;

[0043] Machine learning model unit: The unit trains the long short-term memory network based on historical data, inputs the time-series and non-time-series data in the dataset into the model, combines the buckling risk data of the typhoon-resistant structure monitoring module, evaluates the impact of extreme wind loads on blade life, and dynamically adjusts the prediction results.

[0044] Decision output unit: The unit is used to output the remaining life prediction and maintenance decision recommendations to the operation and maintenance platform or main control system, providing visualization results and alarm support.

[0045] Compared with the prior art, the advantages of the present invention are:

[0046] The system boasts significant advantages through the integration of advanced technologies such as multimodal sensor arrays, self-cleaning de-icing, lightning damage monitoring, edge computing, typhoon-resistant structural monitoring, independent power supply, multi-protocol communication, and intelligent diagnostics. Its multimodal sensors can monitor blade strain, vibration, temperature, and icing in real time. Combined with AI algorithms from edge computing, it accurately diagnoses blade damage and icing thickness, improving operational safety. The self-cleaning de-icing module effectively prevents ice accumulation through hot air jetting and a hydrophobic coating, ensuring blade performance. The lightning damage monitoring module accurately locates lightning strike points and current intensity, reducing the risk of lightning strikes. The typhoon-resistant module analyzes buckling strain in real time using fiber optic grating sensors, predicting risks under typhoon conditions and ensuring structural stability. The power supply module utilizes solar and micro-wind power to provide a stable power source, enhancing system independence. The multi-protocol communication gateway supports LoRaWAN, satellite, and BeiDou communication, ensuring reliable data transmission in extreme environments. The intelligent diagnostic module uses machine learning to predict blade life and provide maintenance suggestions, optimizing operational efficiency and extending equipment life. This device comprehensively enhances the reliability and economy of wind turbines in complex environments, helps to make efficient use of clean energy, and demonstrates the combination of technological innovation and national scientific and technological strength. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the system proposed in this invention;

[0048] Figure 2 This is a schematic diagram of the multimodal sensor array module proposed in this invention;

[0049] Figure 3 This is a diagram of the self-cleaning de-icing module proposed in this invention;

[0050] Figure 4 This is a diagram of the lightning damage monitoring module proposed in this invention;

[0051] Figure 5 This is a diagram of the edge computing node module proposed in this invention;

[0052] Figure 6 This is a diagram of the typhoon-resistant structural monitoring module proposed in this invention;

[0053] Figure 7 This is a diagram of the power supply module proposed in this invention;

[0054] Figure 8 This is a diagram of the multi-protocol communication gateway module proposed in this invention;

[0055] Figure 9 This is a diagram of the intelligent diagnostic module proposed in this invention. Detailed Implementation

[0056] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0057] See Figure 1 As shown, a blade condition monitoring device for wind turbines includes:

[0058] Multimodal sensor array module: The module includes a piezoelectric strain sensor array installed at the root of the blade, a triaxial vibration sensor at the tip of the blade, a distributed fiber optic temperature sensor on the blade surface, and a millimeter-wave radar icing detector at the leading edge of the blade.

[0059] Self-cleaning de-icing module: The module is integrated into the microporous hot air jet unit and hydrophobic nano-coating at the leading edge of the blade;

[0060] Lightning damage monitoring module: The module includes a lightning current sensor embedded in the carbon fiber skin of the blade and an arc locator based on ultraviolet imaging, which can detect the lightning strike location and current intensity in real time.

[0061] Edge computing node module: The module is deployed in a waterproof and corrosion-resistant sealed enclosure inside the nacelle, and has a built-in AI chip for running blade damage diagnosis algorithms;

[0062] Typhoon-resistant structure monitoring module: The module is equipped with an FBG fiber grating sensor array at the main beam position of the blade to monitor the buckling strain threshold under typhoon conditions in real time.

[0063] Power supply module: The module includes a flexible solar film and a vertical axis micro wind turbine installed on the top of the nacelle, providing an independent power supply for the sensor network;

[0064] Multi-protocol communication gateway module: The module supports dual-link redundant transmission of LoRaWAN and satellite communication, and automatically switches to BeiDou short message communication when the ground network is interrupted due to a typhoon;

[0065] Intelligent diagnostic module: The module is a machine learning model deployed on a remote server. The input layer receives data from each module, and the output layer generates blade life prediction and maintenance decision suggestions.

[0066] See Figure 2 As shown, the multimodal sensor array module specifically includes:

[0067] Sensor Unit: The unit includes a piezoelectric strain sensor array installed at the root of the blade, a triaxial vibration sensor at the tip of the blade, a distributed fiber optic temperature sensor on the blade surface, and a millimeter-wave radar icing detector at the leading edge of the blade.

[0068] Unit data integration unit: The unit integrates and preprocesses the data collected by each sensor unit and transmits it to the intelligent diagnostic module.

[0069] Specifically, the piezoelectric strain sensor array is used to monitor the stress at the blade root, capture dynamic strain data, and assess the blade structural load and fatigue state. The triaxial vibration sensor is used to measure the vibration acceleration at the blade tip in three dimensions, analyze the blade's dynamic response and potential flutter phenomena. The distributed fiber optic temperature sensor is used to monitor the blade surface temperature distribution in real time, assess thermal stress, environmental adaptability, and potential overheating risks. The millimeter-wave radar icing detector is used to detect the icing condition at the blade's leading edge, assess the ice thickness and distribution, and ensure the blade's aerodynamic performance and operational safety.

[0070] See Figure 3 As shown, the self-cleaning de-icing module specifically includes:

[0071] Micro-hole hot air injection unit: The unit is integrated into the leading edge of the blade and connected to the vortex heating device in the nacelle through the air guide pipe arranged in the hollow blade beam. A micro-hole array made of high-strength composite material is pre-embedded in the leading edge of the blade. A high-temperature and high-pressure resistant air guide pipe is arranged in the hollow blade beam, extending from the root of the blade to the leading edge micro-hole array. The vortex heating device located in the nacelle adopts the principle of electromagnetic induction and uses a fan to press hot air into the air guide pipe. The micro-hole array is connected to a micro solenoid valve to control the hot air injection area and flow rate.

[0072] Hydrophobic nanocoating: The coating is a superhydrophobic nanocoating applied by plasma spraying to cover the leading edge of the blade.

[0073] Specifically, a micropore array made of high-strength composite material is embedded in the leading edge of the blade. The pore diameter ranges from 0.5 to 1 mm and is evenly distributed on the windward side. The pore spacing is optimized to ensure maximum hot air coverage. The micropores are made of high-temperature resistant and corrosion-resistant materials to prevent wear during long-term use.

[0074] High-temperature and high-pressure resistant air guide tubes are installed inside the hollow blade beam, extending from the blade root to the leading edge micropore array. The inner wall of the air guide tube is smooth to reduce airflow resistance, and it is equipped with a heat insulation layer to reduce heat loss.

[0075] The eddy current heating device located in the cabin uses the principle of electromagnetic induction to quickly heat the air to 50-80°C. The hot air is forced into the air duct by a high-efficiency fan. The heating device is equipped with a temperature control module, which can automatically adjust the heating power according to the data of the icing detector.

[0076] The coating is uniformly adhered to the leading edge surface of the blade by plasma spraying or chemical vapor deposition technology, with a thickness controlled between 10-50μm. This ensures both hydrophobicity and does not affect the aerodynamic shape. The coating has UV resistance, wear resistance, and chemical corrosion resistance, and the surface hardness reaches 6H or higher, making it suitable for long-term exposure to harsh environments such as wind, rain, and ice.

[0077] See Figure 4 As shown, the lightning damage monitoring module specifically includes:

[0078] Sensor Unit: The unit monitors the lightning current passing through the carbon fiber skin of the blade in real time through a broadband lightning current sensor, captures the peak value, frequency and duration of the current waveform, and detects the ultraviolet radiation of the electric arc caused by the lightning strike through an arc locator based on ultraviolet imaging technology to locate the three-dimensional spatial coordinates of the lightning strike point.

[0079] Data processing unit: The unit amplifies, filters, and performs analog-to-digital conversion on the signal collected by the broadband lightning current sensor to generate analyzable current waveform data, and simultaneously synchronizes it with the ultraviolet imaging positioning data.

[0080] Data transmission unit: The unit transmits the processed lightning current waveform, spatial coordinates and analysis results to the intelligent diagnostic module through a standard communication interface.

[0081] Specifically, a wideband lightning current sensor is used to monitor the lightning current in real time, obtain the peak value, frequency and duration of the current waveform, and calculate the main frequency through Fourier transform; a Rogowski coil or a high-frequency current transformer is used, with a frequency response range covering 10Hz-10MHz, to capture the wideband characteristics of lightning transient current;

[0082] The arc locator detects the ultraviolet radiation from the arc caused by a lightning strike, outputs the radiation intensity, and calculates the three-dimensional coordinates of the lightning strike point by combining the intensity and shape of the ultraviolet arc spot with a triangulation algorithm through the built-in image processing unit.

[0083] See Figure 5 As shown, the edge computing node module specifically includes:

[0084] The edge computing node module is deployed in a waterproof and corrosion-resistant sealed enclosure inside the nacelle. It has a built-in AI chip that runs a blade damage diagnosis algorithm. The algorithm includes a strain-vibration time-frequency joint analysis model and an ice thickness inversion model.

[0085] Strain-Vibration Time-Frequency Joint Analysis Model Unit: The unit performs short-time Fourier transform on the data from the piezoelectric strain sensor at the blade root and the triaxial vibration sensor at the blade tip, performs time-frequency decomposition on the strain and vibration signals, extracts the time-frequency features of strain and vibration, and diagnoses blade structural damage and abnormal vibration based on the built-in damage diagnosis model of the convolutional neural network.

[0086] Ice thickness inversion model unit: The unit calculates the ice thickness based on the intensity and phase change of the reflected signal from millimeter-wave radar using an electromagnetic wave propagation model. Combined with temperature data, it constructs an ice accretion physical model. Based on the change of the dielectric constant of ice with temperature, it corrects the inversion results. Through an inversion algorithm, it maps the radar signal and temperature data to the ice thickness distribution, generating a two-dimensional thermal map of the leading edge ice on the blade.

[0087] Specifically, the strain signal and vibration signal are decomposed into time-frequency components by short-time Fourier transform to obtain the time-frequency spectrum. The joint analysis algorithm identifies strain-vibration coupling features through feature fusion and detects abnormal patterns, such as the correlation between high-frequency vibration and strain abrupt change. Through the built-in damage diagnosis model, the time-frequency features are compared with the pre-trained damage pattern library based on the convolutional neural network to classify damage types (cracks, fatigue, delamination, etc.) and assess the severity.

[0088] Using an electromagnetic wave propagation model, the ice thickness is calculated based on the intensity and phase changes of the reflected signal from millimeter-wave radar. Combined with temperature data, an icing physical model is constructed, considering the influence of temperature variations on the ice's dielectric constant to correct the inversion results. An inversion algorithm is used to map radar signals and temperature data to the ice thickness distribution, generating a two-dimensional thermal map of the blade's leading edge icing. The model supports zonal analysis, identifying concentrated icing areas, such as the windward side or blade tip, and assessing the potential impact of icing on the blade's aerodynamic performance.

[0089] The formula relating phase difference to ice thickness is:

[0090]

[0091] Where ΔΦ is the phase difference, d(t) is the ice thickness, and ε r λ is the relative permittivity of ice, and λ is the millimeter wave wavelength.

[0092] See Figure 6 As shown, the typhoon-resistant structure monitoring module specifically includes:

[0093] Data acquisition unit: The unit is used to receive and preprocess dynamic buckling strain data acquired by the FBG fiber grating sensor array;

[0094] Dynamic buckling strain analysis unit: The unit uses time-domain analysis and frequency-domain analysis to extract the periodicity, amplitude and frequency characteristics of dynamic buckling strain, generates the time-frequency characteristic map of strain signal through short-time Fourier transform, calculates the buckling critical load based on the pre-constructed mechanical model of blade main beam based on finite element analysis, and trains the mapping relationship between strain characteristics and buckling risk through long short-time memory network to predict the buckling risk level in real time.

[0095] Wind speed classification threshold association unit: The unit sets strain thresholds corresponding to each typhoon level based on historical experimental data, and associates dynamic buckling strain data with typhoon wind speed classification thresholds. Through a neural network model, it predicts the strain trend of the blades under different wind speed levels and generates risk curves.

[0096] Specifically, time-domain analysis is used to calculate the peak, mean, and standard deviation of strain, and frequency-domain analysis (fast Fourier transform) is used to extract the periodicity, amplitude, and frequency characteristics of dynamic buckling strain. Short-time Fourier transform is applied to generate time-frequency characteristic maps of the strain signal to identify abnormal buckling modes under typhoon loads.

[0097] Based on the pre-constructed mechanical model of the blade main beam using finite element analysis, combined with real-time strain data, the buckling critical load and safety margin are calculated. The mapping relationship between strain characteristics and buckling risk is trained using a long short-term memory network, and the buckling risk level is predicted in real time. The model is updated regularly through the cloud to adapt to different blade materials and design parameters.

[0098] The formula for calculating the critical buckling load is:

[0099]

[0100] Among them, P cr Where E is the critical buckling load, I is the elastic modulus of the material, K is the effective length coefficient, and L is the length of the blade main beam.

[0101] A predefined wind speed-strain mapping table is used. Based on historical experimental data and simulation results, strain thresholds corresponding to each typhoon level are set. A dynamic threshold adjustment algorithm is adopted to correct the thresholds in real time according to the blade running time, material aging state and ambient temperature to ensure the adaptability of the assessment. Through regression analysis or neural network model, the strain trend of the blade under different wind speed levels is predicted to generate risk curves.

[0102] See Figure 7 As shown, the power supply module specifically includes:

[0103] Energy Harvesting and Management Unit: This unit is used to monitor and record the energy harvesting process of the flexible CIGS solar thin film and the vertical axis Savonius micro wind turbine in real time, and to optimize the energy output efficiency.

[0104] Intelligent power scheduling unit: The unit monitors the power consumption requirements of the multimodal sensor array module, lightning damage monitoring module, edge computing node module and typhoon-resistant structure monitoring module in real time based on historical data, and establishes a load priority model. Based on the priority, a greedy algorithm is used to divide the energy allocation ratio. The operating point is dynamically adjusted for CIGS solar thin film and Savonius wind turbine through the maximum power point tracking algorithm to obtain the maximum energy harvesting efficiency.

[0105] Power supply stability control unit: The unit is used to ensure that the power supply module provides stable voltage and current to the sensor network, and to prevent power interruption or fluctuation from affecting the operation of the equipment.

[0106] Specifically, a priority-based greedy algorithm is adopted to optimize the energy allocation ratio based on the real-time output of solar and wind power generation, and to implement a maximum power point tracking algorithm. For CIGS solar thin film and Savonius wind turbine, the operating point is dynamically adjusted to maximize energy harvesting efficiency. Based on the support of multi-energy collaborative scheduling, when the solar output is insufficient, priority is given to switching to wind power generation or backup batteries; when both are insufficient, the battery discharge strategy is optimized.

[0107] Voltage and current regulation is achieved by controlling the DC-DC converter to output a stable voltage. When abnormal current or voltage is detected, the output is automatically cut off and the event is recorded, thus realizing an overvoltage and overcurrent protection algorithm. A fault switching mechanism is designed so that when one energy source fails, it automatically switches to wind power generation or a backup battery. It also supports hot backup function to keep at least one energy source online at all times, ensuring the continuous operation of the sensor network.

[0108] The output power is adjusted in real time according to the dynamic power consumption changes of the sensor network to optimize energy efficiency. Power scheduling parameters are adjusted according to temperature and humidity to prevent extreme environments from affecting power supply stability.

[0109] See Figure 8 As shown, the multi-protocol communication gateway module specifically includes:

[0110] Communication Protocol Unit: This unit is used to manage the dual-link protocol of LoRaWAN and satellite communication. Through a dynamic link selection algorithm, it monitors the signal quality of the LoRaWAN link in real time and automatically switches to the satellite communication link when the LoRaWAN link is interrupted.

[0111] Beidou Short Message Communication Unit: When both the LoRaWAN link and the satellite communication link are unavailable, the unit activates the Beidou Short Message Communication Protocol and prioritizes the transmission of high-priority data.

[0112] Data processing and fusion unit: The unit preprocesses, fuses, and prioritizes multi-source sensor data. Through data classification and categorization, the data is divided into real-time alarms, periodic reports, and log data, and integrated to generate a comprehensive status report.

[0113] Specifically, LoRaWAN protocol management implements the LoRaWAN protocol stack, supports low-power wide-area network communication, configures an adaptive data rate algorithm, dynamically adjusts the data rate according to signal strength and signal-to-noise ratio, optimizes transmission distance and power consumption, supports multi-node access, and manages the data streams of multimodal sensor array modules, lightning damage monitoring modules, edge computing node modules, typhoon-resistant structure monitoring modules, and power supply modules in historical data.

[0114] A dynamic link selection algorithm is adopted to monitor the signal quality of LoRaWAN links in real time and prioritize the use of low-power LoRaWAN links. When the LoRaWAN link is interrupted, such as when a typhoon causes a ground base station failure, it automatically switches to the satellite communication link. Link status caching is supported to record the link performance data of the last 24 hours, which is convenient for fault analysis and optimization.

[0115] The BeiDou protocol integrates the BeiDou-3 short message protocol and compresses key data through data compression algorithms. It is designed with automatic triggering logic to switch to BeiDou short message mode when LoRaWAN and satellite communication links are unavailable, prioritizing the transmission of high-priority data, such as buckling risk alarms from typhoon-resistant structure monitoring modules and damage diagnosis results from edge computing node modules, and standardizing data structures through predefined templates.

[0116] See Figure 9 As shown, the intelligent diagnostic module specifically includes:

[0117] Data integration and processing unit: This unit performs real-time batch data collection on data from each module, generates time-series and non-time-series data after time synchronization and feature extraction, and constructs a dataset;

[0118] Machine learning model unit: The unit trains the long short-term memory network based on historical data, inputs the time-series and non-time-series data in the dataset into the model, combines the buckling risk data of the typhoon-resistant structure monitoring module, evaluates the impact of extreme wind loads on blade life, and dynamically adjusts the prediction results.

[0119] Decision output unit: The unit is used to output the remaining life prediction and maintenance decision recommendations to the operation and maintenance platform or main control system, providing visualization results and alarm support.

[0120] Specifically, deep learning models are used to process the fusion of time-series and non-time-series data;

[0121] The input layer receives multimodal feature vectors, including strain-vibration time-frequency features, icing thickness distribution, lightning strike parameters, wind speed classification data, and environmental parameters; the hidden layer contains multiple fully connected or convolutional layers to fuse multimodal features and extract deep representations of damage patterns and environmental impacts; the output layer generates two types of outputs:

[0122] Blade remaining life prediction: in hours; maintenance decision recommendations: categorized results, such as "continue operation", "operate with reduced load", "immediate maintenance";

[0123] The model has a built-in damage mode library, including cracks, fatigue, delamination, lightning damage, and icing effects. Through feature matching and probabilistic reasoning, it identifies the damage type and severity corresponding to the current blade state. Combined with buckling risk data from the typhoon-resistant structure monitoring module, it assesses the impact of extreme wind loads on lifespan and dynamically adjusts the prediction results.

[0124] Outputs include remaining life prediction and maintenance decision recommendations, including remaining life prediction, maintenance decision recommendations, damage analysis reports, and visualization results.

[0125] Furthermore, this solution also proposes a wind turbine generator equipped with the aforementioned blade condition monitoring device for wind turbine generators.

[0126] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0127] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0128] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A blade condition monitoring device for wind turbine generators, characterized in that, include: Multimodal sensor array module: The module includes a piezoelectric strain sensor array installed at the root of the blade, a triaxial vibration sensor at the tip of the blade, a distributed fiber optic temperature sensor on the blade surface, and a millimeter-wave radar icing detector at the leading edge of the blade. Self-cleaning de-icing module: The module is integrated into the microporous hot air jet unit and hydrophobic nano-coating at the leading edge of the blade; Lightning damage monitoring module: The module includes a lightning current sensor embedded in the carbon fiber skin of the blade and an arc locator based on ultraviolet imaging, which can detect the lightning strike location and current intensity in real time. Edge computing node module: The module is deployed in a waterproof and corrosion-resistant sealed enclosure inside the nacelle, and has a built-in AI chip for running blade damage diagnosis algorithms; Typhoon-resistant structure monitoring module: The module is equipped with an FBG fiber grating sensor array at the main beam position of the blade to monitor the buckling strain threshold under typhoon conditions in real time. Power supply module: The module includes a flexible solar film and a vertical axis micro wind turbine installed on the top of the nacelle, providing an independent power supply for the sensor network; Multi-protocol communication gateway module: The module supports dual-link redundant transmission of LoRaWAN and satellite communication, and automatically switches to BeiDou short message communication when the ground network is interrupted due to a typhoon; Intelligent diagnostic module: The module is a machine learning model deployed on a remote server. The input layer receives data from each module, and the output layer generates blade life prediction and maintenance decision suggestions.

2. The blade condition monitoring device for wind turbines according to claim 1, characterized in that, The multimodal sensor array module specifically includes: Sensor Unit: The unit includes a piezoelectric strain sensor array installed at the root of the blade, a triaxial vibration sensor at the tip of the blade, a distributed fiber optic temperature sensor on the blade surface, and a millimeter-wave radar icing detector at the leading edge of the blade. Unit data integration unit: The unit integrates and preprocesses the data collected by each sensor unit and transmits it to the intelligent diagnostic module.

3. The blade condition monitoring device for wind turbines according to claim 1, characterized in that, The self-cleaning de-icing module specifically includes: Micro-hole hot air injection unit: The unit is integrated into the leading edge of the blade and connected to the vortex heating device in the nacelle through the air guide pipe arranged in the hollow blade beam. A micro-hole array made of high-strength composite material is pre-embedded in the leading edge of the blade. A high-temperature and high-pressure resistant air guide pipe is arranged in the hollow blade beam, extending from the root of the blade to the leading edge micro-hole array. The vortex heating device located in the nacelle adopts the principle of electromagnetic induction and uses a fan to press hot air into the air guide pipe. The micro-hole array is connected to a micro solenoid valve to control the hot air injection area and flow rate. Hydrophobic nanocoating: The coating is a superhydrophobic nanocoating applied by plasma spraying to cover the leading edge of the blade.

4. The blade condition monitoring device for wind turbines according to claim 1, characterized in that, The lightning damage monitoring module specifically includes: Sensor Unit: The unit monitors the lightning current passing through the carbon fiber skin of the blade in real time through a broadband lightning current sensor, captures the peak value, frequency and duration of the current waveform, and detects the ultraviolet radiation of the electric arc caused by the lightning strike through an arc locator based on ultraviolet imaging technology to locate the three-dimensional spatial coordinates of the lightning strike point. Data processing unit: The unit amplifies, filters, and performs analog-to-digital conversion on the signal collected by the broadband lightning current sensor to generate analyzable current waveform data, and simultaneously synchronizes it with the ultraviolet imaging positioning data. Data transmission unit: The unit transmits the processed lightning current waveform, spatial coordinates and analysis results to the intelligent diagnostic module through a standard communication interface.

5. The blade condition monitoring device for wind turbines according to claim 1, characterized in that, The edge computing node module specifically includes: The edge computing node module is deployed in a waterproof and corrosion-resistant sealed enclosure inside the nacelle. It has a built-in AI chip that runs a blade damage diagnosis algorithm. The algorithm includes a strain-vibration time-frequency joint analysis model and an ice thickness inversion model. Strain-Vibration Time-Frequency Joint Analysis Model Unit: The unit performs short-time Fourier transform on the data from the piezoelectric strain sensor at the blade root and the triaxial vibration sensor at the blade tip, performs time-frequency decomposition on the strain and vibration signals, extracts the time-frequency features of strain and vibration, and diagnoses blade structural damage and abnormal vibration based on the built-in damage diagnosis model of the convolutional neural network. Ice thickness inversion model unit: The unit calculates the ice thickness based on the intensity and phase change of the reflected signal from millimeter-wave radar using an electromagnetic wave propagation model. Combined with temperature data, it constructs an ice accretion physical model. Based on the change of the dielectric constant of ice with temperature, it corrects the inversion results. Through an inversion algorithm, it maps the radar signal and temperature data to the ice thickness distribution, generating a two-dimensional thermal map of the leading edge ice on the blade.

6. The blade condition monitoring device for a wind turbine according to claim 1, characterized in that, The typhoon-resistant structural monitoring module specifically includes: Data acquisition unit: The unit is used to receive and preprocess dynamic buckling strain data acquired by the FBG fiber grating sensor array; Dynamic buckling strain analysis unit: The unit uses time-domain analysis and frequency-domain analysis to extract the periodicity, amplitude and frequency characteristics of dynamic buckling strain, generates the time-frequency characteristic map of strain signal through short-time Fourier transform, calculates the buckling critical load based on the pre-constructed mechanical model of blade main beam based on finite element analysis, and trains the mapping relationship between strain characteristics and buckling risk through long short-time memory network to predict the buckling risk level in real time. Wind speed classification threshold association unit: The unit sets strain thresholds corresponding to each typhoon level based on historical experimental data, and associates dynamic buckling strain data with typhoon wind speed classification thresholds. Through a neural network model, it predicts the strain trend of the blades under different wind speed levels and generates risk curves.

7. The blade condition monitoring device for wind turbines according to claim 1, characterized in that, The power supply module specifically includes: Energy Harvesting and Management Unit: This unit is used to monitor and record the energy harvesting process of the flexible CIGS solar thin film and the vertical axis Savonius micro wind turbine in real time, and to optimize the energy output efficiency. Intelligent power scheduling unit: The unit monitors the power consumption requirements of the multimodal sensor array module, lightning damage monitoring module, edge computing node module and typhoon-resistant structure monitoring module in real time based on historical data, and establishes a load priority model. Based on the priority, a greedy algorithm is used to divide the energy allocation ratio. The operating point is dynamically adjusted for CIGS solar thin film and Savonius wind turbine through the maximum power point tracking algorithm to obtain the maximum energy harvesting efficiency. Power supply stability control unit: The unit is used to ensure that the power supply module provides stable voltage and current to the sensor network, and to prevent power interruption or fluctuation from affecting the operation of the equipment.

8. The blade condition monitoring device for a wind turbine according to claim 1, characterized in that, The multi-protocol communication gateway module specifically includes: Communication Protocol Unit: This unit is used to manage the dual-link protocol of LoRaWAN and satellite communication. Through a dynamic link selection algorithm, it monitors the signal quality of the LoRaWAN link in real time and automatically switches to the satellite communication link when the LoRaWAN link is interrupted. Beidou Short Message Communication Unit: When both the LoRaWAN link and the satellite communication link are unavailable, the unit activates the Beidou Short Message Communication Protocol and prioritizes the transmission of high-priority data. Data processing and fusion unit: The unit preprocesses, fuses, and prioritizes multi-source sensor data. Through data classification and categorization, the data is divided into real-time alarms, periodic reports, and log data, and integrated to generate a comprehensive status report.

9. A blade condition monitoring device for a wind turbine according to claim 1, characterized in that, The intelligent diagnostic module specifically includes: Data integration and processing unit: This unit performs real-time batch data collection on data from each module, generates time-series and non-time-series data after time synchronization and feature extraction, and constructs a dataset; Machine learning model unit: The unit trains the long short-term memory network based on historical data, inputs the time-series and non-time-series data in the dataset into the model, combines the buckling risk data of the typhoon-resistant structure monitoring module, evaluates the impact of extreme wind loads on blade life, and dynamically adjusts the prediction results. Decision output unit: The unit is used to output the remaining life prediction and maintenance decision recommendations to the operation and maintenance platform or main control system, providing visualization results and alarm support.

10. A wind turbine generator, characterized in that, It is equipped with a blade condition monitoring device for wind turbines as described in claims 1-9.

Images