A method and system for monitoring sentinel units adapted to unmanned wind farms

Abstract

The application discloses a kind of sentinel machine set monitoring method and system suitable for unmanned wind farm, belong to wind turbine state monitoring technical field.The method includes: obtaining the structural load data collected by the load sensor deployed on at least one wind turbine as sentinel machine set in wind farm, and the basic operation data collected by all wind turbines through standard sensor, wherein the number of sentinel machine set is less than the total number of wind turbines in wind farm;The basic operation data of the wind turbine without deploying load sensor is input into the pre-trained data-driven model, and the model is trained according to the mapping relationship between the structural load data of sentinel machine set and the basic operation data of all wind turbines;Receive the deduction structural load data output by the model.The application realizes the effective coverage of the whole field key structural load by deploying load sensor on a few sentinel machine sets and using data-driven model to extend load sensing capability to all field units at low cost.

Description

Technical Field

[0001] This invention belongs to the field of wind turbine condition monitoring technology, specifically relating to a sentinel turbine monitoring method and system adapted to unmanned wind farms. Background Technology

[0002] Wind turbines operate in complex natural environments such as turbulent winds, gusts, and wind shear for extended periods, making critical components like blades, pitch bearings, and towers susceptible to fatigue damage or sudden failure. Load sensors can directly measure mechanical quantities such as blade root bending moments, pitch bearing loads, and tower bending moments, providing a physical basis for determining whether the structure exceeds limits and assessing the turbine's health status. In contrast, relying solely on conventional operating data such as wind speed, power, and rotational speed cannot directly reveal the structural stress, often leading to monitoring lags and the risk of missed reports.

[0003] Currently, wind farm condition monitoring mainly employs two methods. One method relies on the basic sensors built into the wind turbines at the factory, including speed sensors, voltage sensors, current sensors, and power sensors, to routinely collect the turbine's operating parameters. This method has a limited number of measuring points and a single sensing dimension, making it difficult to comprehensively cover the health status of key components such as blades, main bearings, gearboxes, and towers. It also cannot accurately perceive external wind resource characteristics and environmental coupling effects, primarily used for post-fault tracing, and is insufficient for early warning and in-depth performance optimization. The other method involves installing advanced sensing equipment on all or most of the wind turbines, such as lidar anemometers, structural load sensors, vibration acoustic sensors, and panoramic visual monitoring systems, to obtain richer condition information. While this method improves monitoring capabilities, the high cost of multi-measuring point configuration, complex deployment and maintenance, and difficulty in data integration and processing result in poor overall economic efficiency, hindering large-scale application in the industry.

[0004] It is evident that existing technological solutions present a contradiction between the breadth of monitoring coverage and the controllability of economic costs. How to acquire key structural load data of all wind turbine units across the entire site without excessively increasing investment, and to achieve accurate health status perception and safety early warning, is a pressing issue that needs to be addressed in this field. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a sentinel turbine monitoring method and system adapted to unmanned wind farms. Its purpose is to acquire key structural load data of all wind turbines in the wind farm without excessively increasing the investment in wind farm monitoring hardware, thereby achieving a balance between the breadth of monitoring coverage and the controllability of economic costs, and realizing accurate perception and safety early warning of the health status of unmanned wind farm turbines.

[0006] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: According to a first aspect of the present invention, a sentinel turbine monitoring method adapted to unmanned wind farms is provided, comprising: The system acquires structural load data from load sensors deployed on at least one wind turbine serving as a sentinel unit within the wind farm, and acquires basic operating data from all wind turbines within the wind farm via standard sensors; wherein the number of wind turbines serving as sentinel units is less than the total number of wind turbines within the wind farm. The basic operating data of the wind turbine without the load sensor is input into a pre-trained data-driven model; the data-driven model is trained based on the mapping relationship between the structural load data of the wind turbine acting as a sentinel unit and the basic operating data of all wind turbines. Receive the inferred structural load data corresponding to the wind turbine without the deployed load sensor, output by the data-driven model.

[0007] In one possible implementation of the first aspect, the wind turbine serving as a sentinel unit is further equipped with environmental monitoring sensors; the method also includes: The environmental monitoring sensors collect environmental status data of the wind farm, including at least one of extreme gust data, flood level data, and ice thickness data. When the environmental status data exceeds the corresponding preset environmental threshold, an environmental warning signal is generated.

[0008] In one possible implementation of the first aspect, the wind turbine serving as a sentinel unit is further equipped with safety monitoring sensors; the method further includes: The safety monitoring sensor collects safety status data of the wind turbine unit, which acts as a sentinel unit. The safety status data includes at least one of the following: fire image data, hub speed data, clearance distance between blades and tower, and vibration amplitude data of tower and blades. When the security status data exceeds the corresponding preset security threshold, a security warning signal is generated and a protection action is triggered.

[0009] In one possible implementation of the first aspect, the wind turbine serving as a sentinel unit is further equipped with an energy efficiency monitoring sensor; the method further includes: The energy efficiency monitoring sensor collects energy efficiency status data of the wind turbine unit, which serves as a sentinel unit. The energy efficiency status data includes at least one of acoustic fingerprint data of the yaw system and image data of the blade cavity. Based on the energy efficiency status data, identify yaw system anomalies or blade stall conditions, and generate energy efficiency optimization prompt signals.

[0010] In one possible implementation of the first aspect, the data-driven model is a neural network model, which is an LSTM network or a CNN network. The basic operating data includes at least one of the following characteristics: wind speed, rotational speed, power, pitch angle, yaw angle, and turbulence intensity.

[0011] According to a second aspect of the present invention, a sentinel turbine monitoring system adapted to unmanned wind farms is provided, comprising: At least one wind turbine serves as a sentinel unit, the number of which is less than the total number of wind turbines in the wind farm; each of the wind turbines serving as a sentinel unit is equipped with a load sensor, which is used to collect structural load data of the wind turbine serving as a sentinel unit. A data acquisition unit is used to acquire basic operating data collected by all wind turbines in the wind farm through standard sensors. The model processing unit stores a pre-trained data-driven model; the data-driven model is trained based on the mapping relationship between the structural load data of the wind turbine unit acting as a sentinel unit and the basic operating data of all wind turbine units. The simulation unit is used to input the basic operating data of the wind turbine without the load sensor into the data-driven model, and to receive the simulation structural load data of the wind turbine without the load sensor output by the data-driven model.

[0012] In one possible implementation of the second aspect, the wind turbine unit serving as a sentinel unit is also equipped with environmental monitoring sensors, which include lidar wind measurement devices and / or environmental video sensors. The environmental monitoring sensor is used to collect environmental status data of the wind farm, and the environmental status data includes at least one of extreme gust data, flood level data, and ice thickness data. The system is also configured to generate an environmental warning signal when the environmental status data exceeds the corresponding preset environmental threshold.

[0013] In one possible implementation of the second aspect, the wind turbine unit serving as a sentinel unit is also equipped with a safety monitoring sensor, which includes at least one of a safety video sensor, a speed sensor, a millimeter-wave radar clearance sensor, and a vibration sensor. The safety monitoring sensor is used to collect safety status data of the wind turbine unit that serves as a sentinel unit. The safety status data includes at least one of the following: fire image data, hub speed data, clearance distance between blades and tower, and vibration amplitude data of tower and blades. The system is also configured to generate a security warning signal and trigger a protection action when the security status data exceeds the corresponding security preset threshold.

[0014] In one possible implementation of the second aspect, the wind turbine serving as a sentinel unit is also equipped with an energy efficiency monitoring sensor, which includes an acoustic fingerprint sensor and / or a blade cavity video sensor. The energy efficiency monitoring sensor is used to collect energy efficiency status data of the wind turbine unit that serves as a sentinel unit. The energy efficiency status data includes at least one of acoustic fingerprint data of the yaw system and image data of the blade cavity. The system is also configured to: identify yaw system anomalies or blade stall states based on the energy efficiency status data, and generate energy efficiency optimization prompt signals.

[0015] In one possible implementation of the second aspect, the data-driven model is a neural network model, which is an LSTM network or a CNN network; the basic operating data includes at least one feature among wind speed, rotational speed, power, pitch angle, yaw angle, and turbulence intensity.

[0016] According to a third aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned sentinel turbine monitoring method adapted to unmanned wind farms.

[0017] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned sentinel turbine monitoring method adapted to unmanned wind farms.

[0018] According to a fifth aspect of the present invention, a computer program product is provided, which, when executed by a processor, implements the aforementioned sentinel turbine monitoring method adapted to unmanned wind farms.

[0019] Compared with the prior art, the present invention has at least the following beneficial effects: This invention selects only a few wind turbines as sentinel units in a wind farm and deploys load sensors on them. By acquiring structural load data from these sentinel units and basic operational data from standard sensors of all units in the farm, a pre-trained data-driven model is used to establish a mapping relationship from basic operational data to structural load data. Furthermore, the basic operational data of wind turbines without deployed load sensors is input into the model to deduce their corresponding structural load data. This point-to-area differentiated monitoring architecture breaks through the dilemma of traditional solutions that either require extensive installation of advanced sensors leading to high costs or rely solely on basic sensors resulting in monitoring lag. It achieves key structural load data for all units in the farm without excessively increasing hardware investment. Effective coverage; because the load sensor directly collects mechanical quantities such as blade root bending moment, pitch bearing load, and tower bending moment, these data are the physical basis for judging whether the structure is within the design limits. Compared with the traditional method of inferring anomalies only through indirect parameters such as wind speed, power, and rotational speed, this invention can more directly and timely reflect the actual stress state of the unit, reducing the risk of monitoring lag and missed reporting. At the same time, the data-driven model enables ordinary units that originally did not have load sensing capabilities to obtain equivalent load monitoring functions without any hardware modification. The derived structural load data can be used for subsequent extreme load judgment and fatigue load spectrum calculation, which solves the contradiction between the breadth of monitoring coverage and the controllability of economic costs in the existing technical solutions. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the specific embodiments of the present invention, the drawings used in the description of the specific embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a flowchart of a sentinel turbine monitoring method adapted to unmanned wind farms according to the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, this invention provides a sentinel turbine monitoring method adapted to unmanned wind farms, specifically including the following steps: Step S1: Obtain structural load data from load sensors deployed on at least one wind turbine serving as a sentinel unit within the wind farm, and obtain basic operating data from all wind turbines within the wind farm collected by standard sensors. The number of wind turbines serving as sentinel units is less than the total number of wind turbines within the wind farm.

[0024] It should be noted that the standard sensors refer to the basic sensors built into the wind turbine at the factory, including speed sensors, voltage sensors, current sensors, and power sensors, which are used to collect routine operating data such as wind speed, power, and speed. Load sensors are used to collect mechanical quantities such as blade root bending moment, pitch bearing load, and tower bending moment.

[0025] Step S2: Input the basic operating data of the wind turbines without the deployed load sensors into the pre-trained data-driven model. The data-driven model is trained based on the mapping relationship between the structural load data of the wind turbines acting as sentinel units and the basic operating data of all wind turbines.

[0026] This mapping relationship is established through machine learning methods, enabling the model to output corresponding structural load data after inputting basic operational data.

[0027] Step S3: Receive the simulated structural load data corresponding to the wind turbine without the deployed load sensor, output by the data-driven model.

[0028] It should be understood that the structural load data obtained from the simulation can be used for subsequent load assessment, life prediction, or safety warning.

[0029] Taking an unmanned wind farm as an example, this wind farm has 50 wind turbines. Three of these turbines are selected as sentinel turbines, and fiber optic strain gauge load sensors are installed on each sentinel turbine to monitor the bending moment at the blade root, the pitch bearing load, and the bending moment at the bottom and middle of the tower. The remaining 47 wind turbines only collect basic operating data such as wind speed, rotational speed, and power through standard sensors. The data-driven model uses a Long Short-Term Memory (LSTM) network. The input layer receives the basic operating data features of the sentinel turbines, including mean wind speed, turbulence intensity, and pitch angle change rate, while the output layer corresponds to each load channel. After training the model with data from the sentinel turbines, the basic operating data of the 47 ordinary turbines are input into the model, allowing the derivation of the structural load time series data for each ordinary turbine.

[0030] Specifically, the full-field load simulation process includes data alignment and preprocessing, feature engineering construction, neural network model design, model training, model validation, simulation to units without load sensors, load result post-processing, and model update closed-loop steps. Data alignment and preprocessing requires time alignment of SCADA data and CMS load data from the sentinel unit, removal of outliers and downtime data, and a unified sampling frequency of 1Hz or 10Hz. Feature engineering extracts load-related features from SCADA data, such as mean wind speed, standard deviation of wind speed, turbulence intensity, pitch angle change rate, fluctuation characteristics of speed and power, yaw error, and nacelle vibration amplitude; and filters and extracts statistical features from CMS load data, such as maximum values ​​and fatigue equivalent loads. The neural network model adopts a feedforward neural network structure, with the loss function being the mean squared error (MSE) or the load spectrum distribution difference loss. After model training, a portion of the data from the sentinel unit is reserved for validation, comparing the errors between predicted and measured loads, such as mean absolute error (MAE) and coefficient of determination (R²). 2 Load spectrum overlap. The simulation results are post-processed to calculate extreme loads and fatigue load spectra, and combined with wind condition markers for abnormal events, finally outputting a load simulation report for each unit without load sensors. The model is periodically retrained or fine-tuned using new data from sentinel units, for example, updated quarterly.

[0031] This implementation method deploys load sensors on a small number of sentinel turbine units and extends load sensing capabilities to all wind turbine units in the field that do not have load sensors deployed by using a data-driven model. This avoids the huge investment required to install expensive load sensors on all units, while obtaining key structural load data for all units in the field.

[0032] In one implementation, the wind turbine, acting as a sentinel unit, is also equipped with environmental monitoring sensors. The method further includes: collecting environmental status data of the wind farm through the environmental monitoring sensors, the environmental status data including at least one of extreme gust data, flood level data, and ice thickness data. When the environmental status data exceeds a corresponding preset environmental threshold, an environmental early warning signal is generated.

[0033] Specifically, the environmental monitoring sensors include a lidar wind measurement device, an environmental video sensor, and a blade video sensor. The lidar wind measurement device is fixedly installed on the top of the nacelle of the sentinel unit in an area unobstructed by blades. This device is small in size and lightweight, making it easy to integrate and install. Its basic principle is based on pulse lidar technology and the Doppler frequency shift coherent detection principle. It emits a laser beam into the atmosphere through a laser head, receives the backscattered signals of atmospheric particles, and combines it with four-beam scanning vector synthesis technology to achieve forward monitoring of the wind field within a range of 10m to 400m directly in front of the wind turbine. It can simultaneously measure wind field data at multiple user-defined locations and calculate extreme gusts and sudden wind changes in real time. When the instantaneous wind speed exceeds a preset threshold and the rate of change of wind speed exceeds a set value, an extreme gust warning signal is generated, triggering the wind turbine load protection and control strategy adjustment.

[0034] An environmental video sensor is mounted on the top of the nacelle to capture images of the environment surrounding the aircraft. This sensor meets IP66 protection standards, operates within a temperature range of -40℃ to 70℃, and uses a wide-angle telephoto lens, enabling it to identify targets as small as 2m x 2m at a maximum distance of 3000m. It automatically analyzes the images using AI image recognition algorithms to identify flood levels. When the water level reaches a warning threshold of 0.5m from the foundation platform, a flood warning signal is generated, triggering equipment protection and personnel evacuation commands.

[0035] The blade video sensor is installed below the wind turbine tower to capture images of the surface condition of the turbine blades and tower. AI image recognition algorithms automatically analyze the images to identify the icing condition of the turbine blades and tower surface. When the icing thickness reaches 5mm or more, or the icing rate exceeds a preset value, an icing warning signal is generated, providing a basis for blade de-icing strategies and wind turbine start-up and shutdown decisions.

[0036] This implementation method achieves early detection of macro-environmental risks at wind farms by deploying environmental monitoring sensors on sentinel turbine units. LiDAR can detect extreme gusts hundreds of meters in advance, buying time for feedforward control of the units; video sensors combined with AI algorithms can automatically identify floods and freezing disasters, overcoming the lag of manual inspections. These environmental early warning signals enable wind farms to autonomously cope with severe weather without human intervention, reducing the risk of equipment damage and improving the safe operation level of unmanned wind farms.

[0037] In one implementation, the wind turbine serving as a sentinel unit is further equipped with safety monitoring sensors. The method also includes: collecting safety status data of the wind turbine serving as a sentinel unit through the safety monitoring sensors. The safety status data includes at least one of the following: fire image data, hub rotation speed data, clearance distance data between the blades and the tower, and vibration amplitude data of the tower and blades. When the safety status data exceeds a corresponding preset safety threshold, a safety warning signal is generated and a protective action is triggered.

[0038] Specifically, safety monitoring sensors include safety video sensors, speed sensors, millimeter-wave radar clearance sensors, and vibration sensors.

[0039] Safety video sensors are installed on the top of the cabin and automatically analyze images using AI image recognition algorithms to identify open flames and smoke inside the cabin and wheel hubs. When the image detects open flame features or the smoke concentration exceeds a preset threshold, a fire warning signal is generated, which in turn triggers the fire extinguishing device and a power-off command to the system.

[0040] The speed sensor employs a Hall effect type, fixedly installed in a non-rotational interference area near the main shaft inside the hub. The probe is aligned with the metal gear ring or fixed magnet on the main shaft, maintaining a slight gap with the gear tip or magnet to avoid contact wear. The Hall element senses changes in the magnetic field of the rotating magnet, converting the displacement signal into a voltage signal. After conditioning, it outputs a pulse signal, thereby calculating the hub speed in real time. When the speed exceeds 120% of the rated speed and the speed acceleration is positive, an overspeed warning signal is generated, triggering the propeller retraction action and a protective shutdown. For example, when the rated speed is 10 rpm, the overspeed threshold is 12 rpm.

[0041] The millimeter-wave radar clearance sensor, based on 24GHz or 77GHz millimeter-wave reflection detection technology, is installed beside the lidar on the top of the nacelle and at the rear of the nacelle, avoiding obstructions from blade rotation trajectories and equipment. This ensures complete coverage of the entire blade rotation radius and the critical clearance areas between the blade tips and the tower and nacelle. The sensor continuously transmits millimeter-wave signals into the blade rotation area and calculates the clearance distance between the blade and the tower in real time after receiving the reflected signals. When the clearance distance falls below a preset safety threshold of 1.5m, an alarm signal is immediately output, triggering the unit control system to perform protective actions such as speed reduction and shutdown, effectively preventing blade-tower-sweeping accidents.

[0042] Vibration sensors employ piezoelectric accelerometers, deployed at both the tower and blades. The sensors acquire vibration signals from the tower and blades, analyzing the vibration frequency and amplitude. When an abnormal shift in vibration frequency or an amplitude exceeding a safe threshold is detected, a vibration warning signal is generated, indicating structural fatigue damage or resonance risk.

[0043] This implementation method deploys multi-dimensional safety monitoring sensors on the sentinel turbine units, enabling real-time monitoring and rapid response to sudden safety incidents such as fires, overspeeding, tower sweeping, and resonance. Safety video sensors can issue alarms in the early stages of a fire, much earlier than traditional temperature sensors; speed sensors combined with acceleration analysis can avoid false alarms; millimeter-wave radar clearance sensors directly measure the distance between the blades and the tower, making it the most effective means of preventing tower sweeping accidents; vibration sensors can identify early fatigue damage. These safety warnings and protective actions together constitute the active safety protection system of unmanned wind farms, greatly reducing the probability of catastrophic accidents.

[0044] In one implementation, the wind turbine serving as a sentinel unit is further equipped with an energy efficiency monitoring sensor. The method further includes: collecting energy efficiency status data of the wind turbine serving as a sentinel unit through the energy efficiency monitoring sensor, the energy efficiency status data including at least one of acoustic signature data of the yaw system and image data of the blade interior. Based on the energy efficiency status data, yaw system anomalies or blade stall states are identified, and an energy efficiency optimization prompt signal is generated.

[0045] Specifically, energy efficiency monitoring sensors include acoustic fingerprint sensors and blade cavity video sensors.

[0046] Acoustic sensors are installed on the outside of the yaw gearbox at the bottom of the nacelle, next to the yaw motor, and near the yaw slewing support connecting the nacelle and the tower. Typically, two to three sensors are installed to form multi-point monitoring, avoiding areas with strong vibration and noise interference. The sensors have built-in high-sensitivity microphones and signal conditioning modules to collect acoustic signals generated by gear meshing, motor operation, and braking during yaw system operation in real time. After filtering and noise reduction, these signals are compared and analyzed with a preset normal acoustic template. When abnormal deviations in acoustic frequency or amplitude are detected and exceed safety thresholds, a yaw system fault warning signal is output. This signal indicates faults such as yaw gear wear, bearing wear, abnormal meshing, or brake loosening. The system then alerts maintenance personnel to investigate, thereby optimizing yaw control strategies, reducing wind error, and improving power generation efficiency.

[0047] Internal video sensors are installed inside the cavities of each blade in the wind turbine generator set to collect internal images of the blade during operation. Intelligent acquisition and analysis are performed by an edge-based intelligent control and analysis unit located on the turbine side, automatically identifying phenomena such as blade penetration cracks, obvious bulges, large-area whitening, complex deformation, and blade stall. Blade stall refers to a state where airflow separates from the blade surface, leading to a sharp drop in aerodynamic efficiency. By analyzing the blade deformation characteristics in the internal images or combining them with aerodynamic parameters, it can be determined whether blade stall has occurred. When blade stall is detected, an energy efficiency optimization alert signal is generated. Maintenance personnel can remotely confirm the blade's health status without shutting down the turbine or climbing the tower, and formulate targeted pitch control strategies or blade aerodynamic optimization schemes to avoid energy loss and major blade failure.

[0048] This implementation method achieves refined monitoring of wind turbine operating efficiency by deploying energy efficiency monitoring sensors on the sentinel turbine units. Acoustic sensors can detect mechanical faults in the yaw system in advance, avoiding power generation losses caused by inaccurate wind alignment; blade internal cavity video sensors can directly observe the internal structural state of the blades and stall phenomena, providing intuitive data for aerodynamic optimization and operation and maintenance decisions. The energy efficiency optimization alerts provided by this implementation method help improve the overall power generation of the wind farm and extend the service life of components.

[0049] In one possible implementation, the data-driven model is a neural network model, specifically an LSTM network or a CNN network. The basic operational data includes at least one of the following features: wind speed, engine speed, power, pitch angle, yaw angle, and turbulence intensity.

[0050] LSTM (Long Short-Term Memory) networks are suitable for processing time-series data and can capture the dynamic characteristics of parameters such as wind speed and power changing over time. CNN (Convolutional Neural Network) networks are suitable for extracting local features and can be used to analyze data fluctuation patterns within short time windows. Both networks can establish a nonlinear mapping from basic operational data to structural load data through supervised learning.

[0051] Basic operating data can be obtained directly through standard sensors or through simple calculations, without the need for additional load sensors.

[0052] This invention also provides a sentinel turbine monitoring system adapted for unmanned wind farms. The system includes: at least one wind turbine serving as a sentinel turbine, a data acquisition unit, a model processing unit, and a simulation unit.

[0053] The number of wind turbines serving as sentinel units is less than the total number of wind turbines in the wind farm. Each of these sentinel wind turbines is equipped with a load sensor for collecting structural load data. The load sensors include fiber optic strain sensors, installed at key locations on the pitch bearing, blade root, and tower. For example, four strain sensors are installed on the pitch bearing, evenly spaced at 90° intervals around the bearing circumference. Blade load sensors are installed 1.5m from the blade root, with sensor 1 located at the center of the leeward side, and sensors 2 and 3 installed clockwise at 120° intervals. Tower load sensors are installed at the top, middle, and bottom of the tower. The top sensor is installed 2m above the lower flange of the last tower section, and the bottom sensor is installed 2m below the upper flange of the bottom tower section.

[0054] It should also be noted that the fiber optic demodulator and edge acquisition unit are installed in the cabinet at the bottom of the tower. The edge acquisition unit is connected to the wind turbine ring network and uploads the data and processed characteristic values ​​to the upper computer intelligent early warning system.

[0055] The data acquisition unit is used to acquire basic operating data from all wind turbines within the wind farm through standard sensors. These standard sensors include speed sensors, voltage sensors, current sensors, and power sensors that are pre-installed on each wind turbine at the factory.

[0056] The model processing unit stores a pre-trained data-driven model. This data-driven model is trained based on the mapping relationship between the structural load data of the wind turbine unit acting as a sentinel unit and the basic operating data of all wind turbine units.

[0057] The simulation unit is used to input the basic operating data of the wind turbine without the load sensor into the data-driven model, and to receive the simulation structural load data of the wind turbine without the load sensor output by the data-driven model.

[0058] In practical deployment, the data acquisition unit, model processing unit, and simulation unit can be integrated into the host computer system of the wind farm's booster station or deployed on a cloud server. Sentinel turbines upload structural load data and basic operation data to the model processing unit via a fiber optic ring network; all ordinary turbines upload basic operation data to the data acquisition unit via a fiber optic ring network. The model processing unit uses the data from the sentinel turbines for model training or updates; the simulation unit applies the trained model to the basic operation data of the ordinary turbines, outputs simulated structural load data in real time, and displays it in the load simulation report on the host computer software platform.

[0059] In one implementation, the wind turbine serving as a sentinel unit is further equipped with environmental monitoring sensors, including lidar wind measurement devices and / or environmental video sensors. These sensors are used to collect environmental status data of the wind farm, including at least one of extreme gust data, flood level data, and frost thickness data. The system is further configured to generate an environmental warning signal when the environmental status data exceeds a corresponding preset environmental threshold.

[0060] The lidar wind measurement device is fixedly installed on the top of the sentry wind turbine's nacelle in an area free of blade obstruction. Based on pulse lidar technology and the Doppler frequency shift coherent detection principle, it employs 4-beam scanning vector synthesis technology to achieve forward-looking monitoring of the wind field within a range of 10m to 400m directly in front of the wind turbine. This device can output parameters such as incoming wind speed, wind direction, and turbulence intensity in real time, providing the system with a basis for judging extreme gusts and sudden wind changes.

[0061] The environmental video sensor is installed on the top of the Sentinel aircraft's nacelle, possessing IP66 protection and the ability to withstand operating temperatures from -40℃ to 70℃. It employs a wide-angle, long-focal-length lens, with a maximum monitoring distance of 3000m based on a 2m*2m target. This sensor automatically identifies flood level changes using AI image recognition algorithms. The blade video sensor is positioned below the tower to identify icing conditions on the blades and tower surface.

[0062] The system is internally configured with a threshold comparison module, where preset environmental thresholds include flood level warning thresholds (e.g., 0.5m from the base platform), ice thickness warning thresholds (e.g., ice thickness reaching 5mm), and extreme gust wind speed thresholds. When the collected environmental status data exceeds the corresponding preset environmental threshold, the system generates an environmental warning signal and issues an alarm through the site monitoring platform.

[0063] This embodiment integrates environmental monitoring sensors and threshold judgment and signal generation functions into the system, enabling unmanned wind farms to autonomously perceive external environmental threats and provide early warnings. The forward-looking wind measurement capability of lidar provides time margin for the unit's feedforward control, and video AI recognition technology replaces manual inspections, improving the level of unmanned operation of the wind farm under severe weather conditions.

[0064] In one implementation, the wind turbine serving as a sentinel unit is further equipped with safety monitoring sensors, including at least one of a safety video sensor, a speed sensor, a millimeter-wave radar clearance sensor, and a vibration sensor. The safety monitoring sensors are used to collect safety status data of the wind turbine serving as a sentinel unit, including at least one of fire image data, hub speed data, clearance distance data between the blades and the tower, and vibration amplitude data of the tower and blades. The system is further configured to generate a safety warning signal and trigger protective actions when the safety status data exceeds a corresponding preset safety threshold.

[0065] Safety video sensors are installed on the top of the Sentinel aircraft's cabin, using AI image recognition algorithms to automatically identify open flames and smoke inside the cabin and wheel hubs. When the concentration of open flames or smoke exceeds a preset threshold, the system generates a fire warning signal and triggers a fire extinguishing device and a power-off command.

[0066] The speed sensor adopts the Hall effect type and is fixedly installed inside the hub near the main shaft. The sensor collects the hub speed in real time. When the speed exceeds 120% of the rated speed and the speed acceleration is positive, the system generates an overspeed warning signal, triggers the propeller retraction action, and performs a protective shutdown.

[0067] The millimeter-wave radar clearance sensor, based on 24GHz or 77GHz millimeter-wave reflection detection technology, is installed next to the lidar on the top of the nacelle and at the rear of the nacelle. The sensor outputs the clearance distance between the blades and the tower in real time. When the clearance distance is lower than the preset safety threshold of 1.5m, the system immediately outputs an alarm signal, triggering the unit control system to perform speed reduction or shutdown protection actions.

[0068] Vibration sensors employ piezoelectric accelerometers, deployed at the tower and blades. The sensors acquire time-domain vibration signals, and the system analyzes the vibration frequency and amplitude. When an abnormal shift in vibration frequency or an amplitude exceeding a safe threshold is detected, a vibration warning signal is generated, indicating structural fatigue damage or resonance risk.

[0069] Safety preset thresholds include fire smoke concentration threshold, overspeed threshold (120% of rated speed), clearance distance threshold (1.5m), vibration amplitude threshold, and frequency deviation threshold. Protective actions include propeller retraction, speed reduction, shutdown, power cut-off, and activation of fire extinguishing devices.

[0070] By integrating safety monitoring sensors and corresponding threshold judgment and protection trigger functions into the system, an active accident prevention system for unmanned wind farms has been constructed. Safety early warning and protection actions can automatically intervene in the early stages of an accident or even before the accident occurs, avoiding accidents such as blade sweeping, turbine overrun, and nacelle fire.

[0071] In one implementation, the wind turbine serving as a sentinel unit is further equipped with energy efficiency monitoring sensors, including acoustic fingerprint sensors and / or blade cavity video sensors. The energy efficiency monitoring sensors are used to collect energy efficiency status data of the wind turbine serving as a sentinel unit, including at least one of acoustic fingerprint data of the yaw system and image data of the blade cavity. The system is further configured to: identify yaw system anomalies or blade stall states based on the energy efficiency status data, and generate an energy efficiency optimization prompt signal.

[0072] Acoustic sensors are installed on the outside of the yaw gearbox at the bottom of the nacelle, next to the yaw motor, and near the yaw swing support, typically deployed in groups of 2 to 3 to form multi-point monitoring. The sensors have built-in high-sensitivity microphones and signal conditioning modules to collect acoustic signals from the yaw system in real time. The system compares the collected acoustic signals with preset normal acoustic templates to analyze abnormal deviations in frequency and amplitude. When fault characteristics such as yaw gear wear, bearing wear, abnormal meshing, or loose brakes are detected, a yaw system malfunction warning signal is generated, suggesting adjustments to the yaw control strategy or scheduling maintenance.

[0073] Internal video sensors are installed in the cavity of each blade to capture internal images during blade operation. The system analyzes the images in real time through an edge intelligent control and analysis unit, automatically identifying phenomena such as blade penetration cracks, obvious bulges, large-area whitening, complex deformation, and blade stall. When blade stall is detected, an energy efficiency optimization prompt signal is generated, suggesting adjustments to the pitch angle or blade aerodynamic optimization.

[0074] Energy efficiency optimization prompts can further trigger automatic control logic, such as adjusting yaw wind accuracy, optimizing pitch rate curves, and recommending the installation of blade aerodynamic accessories, in order to improve the unit's power generation efficiency.

[0075] In one possible implementation, the data-driven model is a neural network model, specifically an LSTM network or a CNN network. The basic operational data includes at least one of the following features: wind speed, engine speed, power, pitch angle, yaw angle, and turbulence intensity.

[0076] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used for the operation of a sentinel turbine monitoring method adapted to unmanned wind farms.

[0077] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be Random Access Memory (RAM) or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the sentinel turbine monitoring method adapted to unmanned wind farms in the above embodiments.

[0078] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.

[0079] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0080] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0081] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0082] This invention also provides a computer program product for executing any of the above-described sentinel turbine monitoring methods adapted to unmanned wind farms. Since the computer program product provided by this invention belongs to the same inventive concept as the sentinel turbine monitoring method adapted to unmanned wind farms described above, it possesses all the advantages of the sentinel turbine monitoring method adapted to unmanned wind farms described above. Therefore, the beneficial effects of the computer program product provided by this invention will not be elaborated upon here.

[0083] In this invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0084] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the scope of the technology disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention.

Claims

1. A method for monitoring sentinel turbine units adapted to unmanned wind farms, characterized in that, include: The system acquires structural load data from load sensors deployed on at least one wind turbine serving as a sentinel unit within the wind farm, and acquires basic operating data from all wind turbines within the wind farm via standard sensors; wherein the number of wind turbines serving as sentinel units is less than the total number of wind turbines within the wind farm. The basic operating data of the wind turbine without the load sensor is input into a pre-trained data-driven model; the data-driven model is trained based on the mapping relationship between the structural load data of the wind turbine acting as a sentinel unit and the basic operating data of all wind turbines. Receive the inferred structural load data corresponding to the wind turbine without the deployed load sensor, output by the data-driven model.

2. The sentinel turbine monitoring method adapted to unmanned wind farms according to claim 1, characterized in that, The wind turbine unit, which serves as a sentinel unit, is also equipped with environmental monitoring sensors; the method further includes: The environmental monitoring sensors collect environmental status data of the wind farm, including at least one of extreme gust data, flood level data, and ice thickness data. When the environmental status data exceeds the corresponding preset environmental threshold, an environmental warning signal is generated.

3. The sentinel turbine monitoring method adapted to unmanned wind farms according to claim 1, characterized in that, The wind turbine unit, which serves as a sentinel unit, is also equipped with safety monitoring sensors; the method further includes: The safety monitoring sensor collects safety status data of the wind turbine unit, which acts as a sentinel unit. The safety status data includes at least one of the following: fire image data, hub speed data, clearance distance between blades and tower, and vibration amplitude data of tower and blades. When the security status data exceeds the corresponding preset security threshold, a security warning signal is generated and a protection action is triggered.

4. The sentinel turbine monitoring method adapted to unmanned wind farms according to claim 1, characterized in that, The wind turbine unit, which serves as a sentinel unit, is also equipped with energy efficiency monitoring sensors; the method further includes: The energy efficiency monitoring sensor collects energy efficiency status data of the wind turbine unit, which serves as a sentinel unit. The energy efficiency status data includes at least one of acoustic fingerprint data of the yaw system and image data of the blade cavity. Based on the energy efficiency status data, identify yaw system anomalies or blade stall conditions, and generate energy efficiency optimization prompt signals.

5. The sentinel turbine monitoring method adapted to unmanned wind farms according to claim 1, characterized in that, The data-driven model is a neural network model, which is either an LSTM network or a CNN network. The basic operating data includes at least one of the following characteristics: wind speed, rotational speed, power, pitch angle, yaw angle, and turbulence intensity.

6. A sentinel turbine monitoring system adapted for unmanned wind farms, characterized in that, include: At least one wind turbine serves as a sentinel unit, the number of which is less than the total number of wind turbines in the wind farm; each of the wind turbines serving as a sentinel unit is equipped with a load sensor, which is used to collect structural load data of the wind turbine serving as a sentinel unit. A data acquisition unit is used to acquire basic operating data collected by all wind turbines in the wind farm through standard sensors. The model processing unit stores a pre-trained data-driven model; the data-driven model is trained based on the mapping relationship between the structural load data of the wind turbine unit acting as a sentinel unit and the basic operating data of all wind turbine units. The simulation unit is used to input the basic operating data of the wind turbine without the load sensor into the data-driven model, and to receive the simulation structural load data of the wind turbine without the load sensor output by the data-driven model.

7. A sentinel turbine monitoring system adapted to unmanned wind farms according to claim 6, characterized in that, The wind turbine unit, which serves as a sentinel unit, is also equipped with environmental monitoring sensors, including lidar wind measurement devices and / or environmental video sensors. The environmental monitoring sensor is used to collect environmental status data of the wind farm, and the environmental status data includes at least one of extreme gust data, flood level data, and ice thickness data. The system is also configured to generate an environmental warning signal when the environmental status data exceeds the corresponding preset environmental threshold.

8. A sentinel turbine monitoring system adapted to unmanned wind farms according to claim 6, characterized in that, The wind turbine unit, which serves as a sentinel unit, is also equipped with safety monitoring sensors, including at least one of a safety video sensor, a speed sensor, a millimeter-wave radar clearance sensor, and a vibration sensor. The safety monitoring sensor is used to collect safety status data of the wind turbine unit that serves as a sentinel unit. The safety status data includes at least one of the following: fire image data, hub speed data, clearance distance between blades and tower, and vibration amplitude data of tower and blades. The system is also configured to generate a security warning signal and trigger a protection action when the security status data exceeds the corresponding security preset threshold.

9. A sentinel turbine monitoring system adapted to unmanned wind farms according to claim 6, characterized in that, The wind turbine unit, which serves as a sentinel unit, is also equipped with energy efficiency monitoring sensors, including acoustic fingerprint sensors and / or blade cavity video sensors. The energy efficiency monitoring sensor is used to collect energy efficiency status data of the wind turbine unit that serves as a sentinel unit. The energy efficiency status data includes at least one of acoustic fingerprint data of the yaw system and image data of the blade cavity. The system is also configured to: identify yaw system anomalies or blade stall states based on the energy efficiency status data, and generate energy efficiency optimization prompt signals.

10. A sentinel turbine monitoring system adapted to unmanned wind farms according to claim 6, characterized in that, The data-driven model is a neural network model, which is an LSTM network or a CNN network; the basic operating data includes at least one feature among wind speed, rotational speed, power, pitch angle, yaw angle, and turbulence intensity.

Images