A method and system for safety monitoring of wind turbine generators based on lidar and satellite.

Abstract

This application belongs to the field of wind power generation technology, and particularly relates to a method and system for safety monitoring of wind turbine generators based on lidar and satellite. The method includes: acquiring an external wind condition risk level, which is determined based on advanced sensing data of the wind field ahead of the wind turbine generator collected by the nacelle lidar; synchronously acquiring initial multi-source heterogeneous data of the wind turbine generator using a unified clock reference; preprocessing the initial multi-source heterogeneous data to obtain actual multi-source heterogeneous data; constructing spatiotemporal attitude data based on the actual multi-source heterogeneous data; determining a coupled risk level based on the external wind condition risk level and the spatiotemporal attitude data; and executing a graded safety strategy based on the coupled risk level. This application improves the safety of wind turbine generator operation.

Description

Technical Field

[0001] This application belongs to the field of wind power generation technology, and in particular relates to a method and system for safety monitoring of wind turbine generators based on lidar and satellite. Background Technology

[0002] With the rapid development of wind power generation technology, the single-unit capacity of wind turbines is constantly increasing, and the operating environment of the units is becoming increasingly complex, which places higher demands on safety monitoring and protection technologies. Existing wind turbine safety monitoring solutions typically use lidar for wind condition perception, combined with inertial sensors, tilt sensors, and other equipment to monitor the structural response of the unit.

[0003] However, existing monitoring systems often operate independently, lacking a unified spatiotemporal alignment framework. This makes it difficult to effectively extract and utilize the coupling relationship between external wind impacts and the internal structural response of the wind turbine. Furthermore, existing safety protection strategies are mostly based on single-dimensional threshold judgments, making them susceptible to signal noise interference, leading to misjudgments or missed detections. Moreover, when safety protection is triggered, it typically employs a rigid safety chain instantaneous action, such as emergency stop or brake application. This rigid protection action imposes enormous impact loads on the drivetrain, accelerating fatigue damage to critical components and shortening the turbine's lifespan. Therefore, how to achieve deep fusion of multi-source heterogeneous data, accurately identify precursors of wind-structure coupling instability, and implement flexible graded protection are urgent technical problems to be solved in the field of wind turbine safety monitoring. Summary of the Invention

[0004] To address the aforementioned issues, this application provides a method and system for safety monitoring of wind turbines based on lidar and satellite. It employs a multi-source heterogeneous data synchronous acquisition framework, combining lidar for advanced wind condition perception with satellite positioning-assisted structural attitude monitoring. This approach can accurately identify the coupling risks between wind conditions and structural health status, thereby implementing a graded safety protection strategy. This effectively avoids impact damage caused by hard sudden stops, improving the safety and service life of wind turbine operation.

[0005] Firstly, this application provides a method for safety monitoring of wind turbine generators based on lidar and satellite, the method comprising: The external wind risk level is obtained, which is determined based on the advanced perception data of the wind field in front of the wind turbine collected by the nacelle lidar. Initial multi-source heterogeneous data of the wind turbine are collected synchronously using a unified clock reference. The initial multi-source heterogeneous data is preprocessed to obtain actual multi-source heterogeneous data. Spatiotemporal attitude data is constructed based on the actual multi-source heterogeneous data. The coupling risk level is determined based on the external wind condition risk level and the spatiotemporal attitude data. Based on the aforementioned coupling risk level, a tiered security strategy is implemented.

[0006] Furthermore, Obtain the external wind risk level, specifically including: Acquire advanced sensing data of the wind field in front of the wind turbine; Based on the aforementioned advanced sensing data, wind speed spatiotemporal gradient features are obtained, and extreme wind events are identified based on these features. The risk level of the external wind conditions is determined based on extreme wind events.

[0007] Furthermore, The initial multi-source heterogeneous data is preprocessed to obtain actual multi-source heterogeneous data. Spatiotemporal attitude data is then constructed based on the actual multi-source heterogeneous data, specifically including: The initial multi-source heterogeneous data is processed by timestamp alignment and quality marking to obtain the actual multi-source heterogeneous data; A multidimensional spatiotemporal attitude tensor is constructed based on the actual multi-source heterogeneous data to obtain the spatiotemporal attitude data.

[0008] Furthermore, The initial multi-source heterogeneous data was acquired by an integrated satellite positioning receiver, inertial measurement unit, and tilt sensor; The unified clock reference is the second pulse signal output by the satellite positioning receiver; The initial multi-source heterogeneous data includes at least: the three-dimensional displacement and heading angle output by the satellite positioning receiver, the angular velocity and acceleration output by the inertial measurement unit, and the static tilt angle output by the tilt sensor.

[0009] Furthermore, The coupling risk level is determined based on the external wind risk level and the spatiotemporal attitude data, specifically including: Feature extraction is performed on the spatiotemporal attitude data to obtain spatiotemporal feature vectors; Based on the spatiotemporal feature vector, health status classification is performed to obtain health status classification results; Establish a coupled risk matrix between the external wind condition risk level and the health status classification result; The coupling risk level is determined based on the coupling risk matrix; The health status classification results include five categories: normal, mild deterioration, moderate deterioration, severe deterioration, and dangerous.

[0010] Furthermore, Feature extraction is performed on the spatiotemporal attitude data to obtain a spatiotemporal feature vector, specifically including: The spatiotemporal attitude data is input into a convolutional neural network; By sliding the convolution kernel of a convolutional neural network along the time axis, spatiotemporal feature vectors representing structural health degradation are extracted.

[0011] Furthermore, Based on the spatiotemporal feature vector, health status classification is performed to obtain the health status classification result, which specifically includes: The spatiotemporal feature vector is input into a decision tree classifier to obtain the probability values ​​of the wind turbine under various health states, including normal, mild degradation, moderate degradation, severe degradation, and dangerous conditions. Based on the probability value, the health status classification result is determined.

[0012] Furthermore, The coupling risk level is determined based on the coupling risk matrix, specifically including: Using the external wind risk level and the health status classification result as indexes, the corresponding initial risk level is obtained from the coupled risk matrix; If the external wind risk level reaches the preset high risk level and the health status classification result belongs to at least one of moderate degradation, severe degradation, or danger, the initial risk level is increased to obtain the coupled risk level.

[0013] Furthermore, Based on the aforementioned coupling risk level, a tiered security strategy is implemented, specifically including: In response to the coupling risk level being at the early warning level, the data sampling frequency is increased, an inspection work order is generated, and early warning information is pushed to the central control center; In response to the coupling risk level being protection level, a fault-tolerant command is issued before the hard safety chain is activated to limit the pitch rate limit and lock the yaw system action, thereby achieving fault isolation without triggering a shutdown. In response to the coupling risk level being dangerous, the conventional pitch PID control is bypassed before the hard safety chain is activated, and a flexible pitch retraction action is performed at a preset gentle slope to smoothly transition the speed of the wind turbine to a safe range.

[0014] Secondly, based on the same inventive concept, this application provides a wind turbine safety monitoring system based on lidar and satellite, the system comprising: An external acquisition module is used to acquire the external wind risk level, which is determined based on the advanced perception data of the wind field in front of the wind turbine collected by the nacelle lidar. The data construction module is used to synchronously collect the initial multi-source heterogeneous data of the wind turbine with a unified clock reference, preprocess the initial multi-source heterogeneous data to obtain the actual multi-source heterogeneous data, and construct spatiotemporal attitude data based on the actual multi-source heterogeneous data. The risk level determination module is used to determine the coupling risk level based on the external wind condition risk level and the spatiotemporal attitude data. The policy execution module is used to execute graded security policies based on the coupling risk level.

[0015] Compared with the prior art, this application has the following advantages: This application solves the problem of asynchronous time delays in multi-source heterogeneous data by synchronously collecting multi-source heterogeneous data through a unified clock reference and constructing spatiotemporal attitude data, thus achieving spatiotemporal alignment between external wind conditions and internal responses. By establishing a coupled risk matrix of external wind condition risk level and health status classification results, it achieves accurate capture of wind condition-structure coupling instability precursors, avoiding misjudgments and omissions caused by single-dimensional threshold judgments. Through a graded safety strategy, it implements graded interventions from early warning and attention to fault-tolerant intervention and flexible pitch retrieval based on the coupled risk level. Especially in dangerous situations, flexible pitch retrieval is used instead of hard shutdown, avoiding secondary damage to the transmission chain from instantaneous impact loads and significantly extending the service life of key components of the wind turbine.

[0016] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a wind turbine safety monitoring method based on lidar and satellite according to an embodiment of this application is shown. Figure 2 A structural block diagram of a wind turbine safety monitoring system based on lidar and satellite, according to an embodiment of this application, is shown. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] Figure 1 A flowchart illustrating a wind turbine safety monitoring method based on lidar and satellite according to an embodiment of this application is shown, as follows: Figure 1 As shown in the embodiment of this application, the wind turbine safety monitoring method based on lidar and satellite includes, S1, Obtain the external wind risk level, which is determined based on the advanced perception data of the wind field in front of the wind turbine collected by the nacelle lidar. In this embodiment of the application, step S1 specifically includes: S11, Obtain advanced sensing data of the wind field in front of the wind turbine; S12, obtain the spatiotemporal gradient features of wind speed based on the advanced sensing data, and identify extreme wind events based on the spatiotemporal gradient features of wind speed. S13, Determine the external wind condition risk level based on extreme wind events.

[0021] In this embodiment of the application, the advanced sensing data of the wind field in front of the wind turbine is obtained by installing a lidar (wind measuring radar) on the top of the wind turbine nacelle and emitting a laser beam in front of the turbine (usually tens to hundreds of meters in front of the rotor plane). By measuring the backscattering signal of atmospheric aerosols, the wind field information is inverted. The advanced sensing data includes incoming wind speed, wind direction, turbulence intensity, wind shear distribution, etc.

[0022] In this embodiment, the spatiotemporal gradient feature of wind speed refers to the rate of change of wind speed in spatial location (spatial gradient) and the rate of change of wind speed over time (temporal gradient). By performing spatiotemporal gridding processing on the advanced sensing data, the ratio of the wind speed difference to the distance between adjacent grid points is calculated, thereby extracting the spatiotemporal gradient feature of wind speed.

[0023] In this embodiment, when a sudden change occurs in the spatiotemporal gradient characteristics of wind speed, such as a rapid increase in wind speed over a short distance or a drastic change in wind direction, an extreme wind event is identified. Extreme wind events include, but are not limited to, sudden gusts, extreme turbulence, sudden wind shear changes, and downstream wake interference. For example, if a sudden gust event is detected when the wind speed 100 meters ahead jumps from 8 m / s to 15 m / s within 2 seconds, and the turbulence intensity exceeds a preset threshold, it is identified as such.

[0024] In this embodiment, the identified extreme wind events are quantified into corresponding external wind risk levels based on their type and intensity. For example, external wind risk levels can be divided into three levels: low risk, medium risk, and high risk. If no extreme wind events are identified, or the wind speed gradient changes gently, it is determined to be low risk; if moderate-intensity turbulence or regular gusts are identified, it is determined to be medium risk; if high-risk events such as extreme turbulence or sudden wind shear changes are identified, it is determined to be high risk.

[0025] Specifically, the external wind risk level is not simply a record of the current wind speed, but a prediction of the impending wind impact. Through advanced sensing data, the system can obtain wind field information within a certain distance in front of the generator unit, such as the wind speed vector distribution within a range of 50 to 200 meters ahead. In this embodiment, the external wind risk level can be divided into multiple levels, including low risk, medium risk, and high risk. This division is based on characteristic parameters including, but not limited to, the abrupt change rate of wind speed, turbulence intensity, and wind shear coefficient. These parameters are compared according to set thresholds, and the risk level is determined based on the comparison results. By identifying extreme wind events based on the spatiotemporal gradient characteristics of wind speed, the external wind risk level no longer relies solely on the instantaneous readings of the local anemometer, but is based on a prediction of the evolution trend of the wind field ahead, greatly improving the accuracy and foresight of risk warnings.

[0026] S2, synchronously collect the initial multi-source heterogeneous data of the wind turbine with a unified clock reference, preprocess the initial multi-source heterogeneous data to obtain actual multi-source heterogeneous data, and construct spatiotemporal attitude data based on the actual multi-source heterogeneous data; In this embodiment, wind turbines are typically equipped with various sensors, such as satellite positioning devices for measuring location information, inertial measurement units for measuring motion state, and tilt sensors for measuring tilt angle. These sensors not only have different data formats (heterogeneous), but also, due to differences in transmission paths and processing chips, their output data often exhibits varying degrees of time delay, making them unsuitable for direct high-precision fusion analysis. This step addresses this by unifying the clock reference, for example, using a high-precision second pulse signal output from a satellite positioning system as the network-wide synchronization source, forcing all sensor acquisition channels to trigger sampling at the same time or adding a unified timestamp to the data frame. In this way, the originally discrete and asynchronous initial multi-source heterogeneous data is calibrated onto the same time axis, eliminating the impact of time deviation on fusion accuracy.

[0027] In this embodiment of the application, the preprocessing process also includes the removal of outliers and the imputation of missing values ​​to ensure the reliability of actual multi-source heterogeneous data.

[0028] It should be understood that although this embodiment mentions sensors such as satellite positioning and inertial measurement, in other embodiments, vibration sensors, strain gauges and other data sources may also be introduced according to the unit configuration. As long as they are synchronized through a unified clock reference, they all fall within the protection scope of this application.

[0029] In this embodiment of the application, step S2 specifically includes: S21, perform timestamp alignment and quality marking processing on the initial multi-source heterogeneous data respectively; S22, construct a multidimensional spatiotemporal attitude tensor based on the actual multi-source heterogeneous data to obtain the spatiotemporal attitude data.

[0030] In this embodiment of the application, the initial multi-source heterogeneous data is acquired by an integrated satellite positioning receiver, inertial measurement unit, and tilt sensor; The unified clock reference is the second pulse signal output by the satellite positioning receiver; The initial multi-source heterogeneous data includes at least: the three-dimensional displacement and heading angle output by the satellite positioning receiver, the angular velocity and acceleration output by the inertial measurement unit, and the static tilt angle output by the tilt sensor.

[0031] In this embodiment, an integrated sensor combination is used to address the problem of insufficient measurement accuracy of a single sensor under complex working conditions, wherein: Satellite positioning receivers (such as BeiDou RTK and GPS RTK) can provide centimeter-level three-dimensional absolute displacement and heading angle information, but their output frequency is usually low (such as 1Hz-10Hz) and they are prone to losing lock in obstructed environments. Inertial measurement units (IMUs) can output angular velocity and acceleration at high frequencies (such as 100Hz-200Hz) and have excellent dynamic response capabilities, but they have drift errors that accumulate over time. Tilt sensors can accurately measure static tilt angles, but they are susceptible to dynamic errors due to acceleration interference when the unit shakes violently. By integrating the above three sensors, the absolute accuracy of satellite positioning is used to correct the drift of inertial measurement, the high-frequency response of inertial measurement is used to compensate for the lag of satellite positioning, and the high accuracy of tilt sensor in static conditions is used to correct attitude calculation. This achieves complementary advantages of multi-source data and significantly improves the accuracy and robustness of attitude monitoring.

[0032] Specifically, while the satellite positioning receiver receives satellite signals and calculates its position information, it outputs a second pulse signal with an extremely steep rising edge. This second pulse signal maintains nanosecond-level synchronization accuracy with the whole second of Coordinated Universal Time (UTC). The system uses this second pulse signal as a network-wide synchronization source, simultaneously connecting it to the trigger ports of the inertial measurement unit and the tilt sensor. When the second pulse signal arrives, all sensors trigger sampling at the same physical moment, or add a unified time stamp to the data frame. This eliminates time discrepancies caused by crystal oscillator frequency deviations or transmission delays between different sensors, ensuring the time consistency of data in subsequent fusion analysis.

[0033] In this embodiment, during data acquisition, the raw data may contain noise or outliers due to electromagnetic interference, occasional sensor malfunctions, or satellite signal obstruction. Therefore, the raw data is quality-labeled. The system performs a quality assessment on each initial piece of multi-source heterogeneous data, for example, detecting whether the ambiguity of satellite positioning is fixed, whether the signal-to-noise ratio is below a threshold, and whether the inertial measurement unit reading exceeds its range. Based on the assessment results, each piece of data is labeled with a quality tag, such as "high-quality," "usable," "suspicious," or "invalid." When constructing spatiotemporal attitude data subsequently, the system can weight or remove data based on the quality tags to prevent outliers from contaminating the feature extraction model, thereby further improving the anti-interference capability of the monitoring system.

[0034] In this embodiment, a multidimensional spatiotemporal attitude tensor is constructed based on actual multi-source heterogeneous data after timestamp alignment and quality labeling. This tensor records the dynamic response history of the unit in the time dimension, records the three-dimensional displacement and attitude angle of the cabin in the spatial dimension, and integrates multiple physical quantities such as displacement, velocity, acceleration, and tilt angle in the feature dimension. This multidimensional spatiotemporal attitude tensor not only contains the state information at a single moment, but also implies the temporal evolution law of the unit's structural response, providing standardized input data for subsequent convolutional neural networks to extract structural health degradation features.

[0035] It should be understood that although this embodiment lists three specific sensor types, in practical applications, vibration sensors, strain gauges, and other data sources can be added as needed. As long as they are synchronized through a unified clock reference and tensors are constructed, they all fall within the protection scope of this application.

[0036] S3, determine the coupling risk level based on the external wind condition risk level and the spatiotemporal attitude data; In the embodiments of this application, traditional monitoring methods often assess external environmental loads and internal structural health separately, such as only triggering an alarm when wind speed exceeds a threshold or only shutting down when vibration amplitude exceeds a standard. However, in actual operation, turbine failure is often the result of the coupling effect of external excitation and internal degradation. For example, in the case of moderate degradation of the turbine structure, even normal wind speeds may induce dangerous vibrations; conversely, in a healthy structural state, the turbine may have the margin to withstand extreme wind conditions. This application establishes a coupling risk level and performs correlation analysis between the external wind risk level and the health state represented by internal spatiotemporal attitude data. This enables the identification of precursory features of wind-structure coupling instability, such as the abnormal amplification of structural response under specific wind excitation, thus comprehensively and accurately identifying the safety status of the wind turbine and effectively reducing the false positive and false negative rates.

[0037] In this embodiment of the application, step S3 specifically includes: S31, Perform feature extraction on the spatiotemporal attitude data to obtain a spatiotemporal feature vector; S32, perform health status classification based on the spatiotemporal feature vector to obtain the health status classification result; S33, Establish a coupled risk matrix between the external wind condition risk level and the health status classification result; S34, Determine the coupling risk level based on the coupling risk matrix; The health status classification results include five categories: normal, mild deterioration, moderate deterioration, severe deterioration, and dangerous.

[0038] In this embodiment of the application, step S31 specifically includes: S311, Input the spatiotemporal attitude data into the convolutional neural network; S312 extracts spatiotemporal feature vectors representing structural health degradation by sliding the convolution kernel of a convolutional neural network along the time axis.

[0039] In the embodiments of this application, the spatiotemporal attitude data includes various physical quantities such as displacement, velocity, acceleration, and tilt angle in the time dimension.

[0040] In this embodiment, the constructed multidimensional spatiotemporal pose tensor is input into the trained convolutional neural network model. The convolutional kernels in the network slide along the time axis, extracting feature maps within a local time window through convolution operations.

[0041] In this embodiment, the time-axis sliding mechanism effectively captures the evolution of the unit's structural response in the time domain, such as minute shifts in the tower's natural frequency, the attenuation trend of the vibration damping ratio, and the asymmetric accumulation of nacelle displacement. These features are often weak precursors to early structural health degradation, easily masked by environmental noise in the raw data. Convolutional neural networks, through multi-layer nonlinear transformations, can automatically extract these deep features characterizing structural health degradation from high-dimensional data, forming a spatiotemporal feature vector.

[0042] It should be understood that although this embodiment uses a convolutional neural network as an example for illustration, in other embodiments, deep learning models such as recurrent neural networks (RNN) or long short-term memory networks (LSTM) can also be used to extract temporal features. As long as the mapping from the original data to the feature vector can be achieved, it falls within the protection scope of this application.

[0043] In this embodiment of the application, step S32 specifically includes: S321, Input the spatiotemporal feature vector into the decision tree classifier to obtain the probability values ​​of the wind turbine under various health states, including normal, mild degradation, moderate degradation, severe degradation, and dangerous conditions; S322, Based on the probability value, determine the health status classification result.

[0044] In this embodiment, the health status classification results are subdivided into five categories to achieve a refined quantitative assessment of the unit's health level. The "normal" state indicates that the operating parameters of all unit components are within the design baseline range; "mild degradation" indicates slight wear or aging, but does not affect normal operation; "moderate degradation" indicates a significant performance decline in key components, requiring inclusion in the watchlist; "severe degradation" indicates a significant structural safety hazard, requiring limiting operating loads; and "dangerous" indicates an impending or already occurring functional failure, requiring immediate protective measures. The decision tree classifier performs a series of logical judgments on the spatiotemporal feature vectors, outputting the probability values ​​of the unit belonging to the above five states. For example, the output results are: P(normal) = 0.1, P(mild degradation) = 0.2, P(moderate degradation) = 0.6, P(severe degradation) = 0.1, P(dangerous) = 0.0. The system selects the state with the highest probability value as the final health status classification result, i.e., determining that the current unit is in the "moderate degradation" state. By using probability values ​​as the output method, not only are classification labels provided, but the credibility of the classification is also given, which provides richer information support for subsequent risk decisions.

[0045] In this embodiment of the application, step S34 specifically includes: S341, using the external wind condition risk level and the health status classification result as indexes, obtain the corresponding initial risk level from the coupled risk matrix; S342, if the external wind risk level reaches the preset high risk level and the health status classification result belongs to at least one of moderate degradation, severe degradation or danger, the initial risk level is increased to obtain the coupled risk level.

[0046] In this embodiment, the coupled risk matrix uses row indices representing external wind condition risk levels (e.g., low, medium, high) and column indices representing health status classification results (e.g., normal, slightly degraded, moderately degraded, severely degraded, dangerous). Each element in the matrix corresponds to an initial risk level. The same external wind condition risk level, when applied to units in different health states, can result in drastically different probabilities of causing an accident. For example, for a "high-risk" external wind condition, if the unit is in a "normal" state, its initial risk level might only be "protective"; however, if the unit is already in a "severely degraded" state, its initial risk level might rise to "dangerous".

[0047] In this embodiment, when the external wind risk level reaches a preset high-risk level (such as extreme turbulence or gusts), and the unit health status classification result is moderately degraded, severely degraded, or dangerous, it indicates that the unit structure can no longer withstand the expected extreme loads. At this time, the system will automatically upgrade the initial risk level found in the matrix, for example, upgrading "protection level" to "danger level".

[0048] S4. Based on the aforementioned coupling risk level, implement a graded security policy.

[0049] In this embodiment of the application, step S4 specifically includes: S41, in response to the coupling risk level being a warning level, increase the data sampling frequency, generate an inspection work order, and push the warning information to the central control center; S42, in response to the coupling risk level being protection level, a fault-tolerant command is issued before the hard safety chain is activated to limit the pitch rate limit and lock the yaw system action, thereby achieving fault isolation without triggering a shutdown; S43, in response to the coupling risk level being dangerous, bypass the conventional pitch PID control before the hard safety chain is activated, and perform a flexible pitch retraction action at a preset gentle slope to smoothly transition the speed of the wind turbine to a safe range.

[0050] In this embodiment, for the early warning level, the application automatically increases the sampling frequency of key sensors (such as displacement sensors and vibration sensors), for example, increasing the sampling frequency from the conventional 1Hz to 10Hz, to capture transient changes in the unit's status. Simultaneously, it automatically generates inspection work orders and pushes them to the mobile terminals of maintenance personnel, indicating the locations of components requiring close monitoring. The advantage of this strategy is that it allows for early intervention in monitoring without affecting the unit's power generation efficiency, accumulating data for subsequent decision-making.

[0051] In this application embodiment, for the protection level, the application proactively issues fault-tolerant instructions before the hard safety chain is activated. These "fault-tolerant instructions" refer to parameter adjustment instructions preset for specific fault modes. For example, limiting the upper limit of the pitch rate to prevent the pitch system from overloading due to drastic movements; locking the yaw system's operation to avoid a surge in additional loads during yaw against wind. This strategy reduces the unit's operating boundaries by limiting some functions or performance, proactively avoiding risky loads without triggering shutdowns or losing power generation revenue, thus achieving refined management of operation even with defects.

[0052] In this embodiment, for hazardous levels, before the hard safety chain activates, the application bypasses conventional pitch PID control and takes over the pitch system. Specifically, the system disables the conventional speed regulation PID output and directly executes a gentle pitch retraction action according to a preset, gradual slope (e.g., a pitch rate changing by 2 to 5 degrees per second). The pitch angle increases smoothly over time, causing the aerodynamic energy absorbed by the impeller to gradually decrease, resulting in a smooth downward curve in the unit speed, eventually transitioning to a safe speed range. This eliminates the impact peak caused by instantaneous seizure, significantly reduces the ultimate load during shutdown, and effectively extends the service life of critical unit components. It should be understood that the specific value of the preset, gradual slope can be adaptively adjusted according to the unit model, capacity, and current operating conditions. As long as a smooth speed transition can be achieved, it falls within the protection scope of this application.

[0053] To more intuitively demonstrate the practical application effect of the technical solution of this application, this embodiment takes the monitoring and protection process of a certain onshore 2.5MW wind turbine as an example for detailed explanation.

[0054] This wind turbine is located in a complex terrain wind field, making it susceptible to sudden gusts and turbulence. A lidar is installed on the top of the nacelle for feedforward wind measurement, and satellite positioning receivers, inertial measurement units, and tilt sensors are integrated inside the nacelle and on the top of the tower. The system sampling frequency is 10Hz.

[0055] During one operation, the unit encountered an unforeseen sudden gust of wind. The specific handling process is as follows: During the sensing phase, the lidar scans the wind field within a range of 100 to 200 meters in front of the unit, acquiring advanced sensing data. Based on this data, the system calculates the spatiotemporal gradient characteristics of wind speed, discovering that at a distance of 150 meters in front, the wind speed rapidly increases from 8 m / s to 16 m / s within 3 seconds, with a wind speed gradient value reaching 2.67 m / s. 2 The wind speed exceeded the preset gust threshold. Based on this, the system identified the extreme wind event and determined the external wind risk level to be high. This proactive perception process provided the unit's control system with a response time window of approximately 5 to 8 seconds, enabling the unit to shift from passively enduring the situation to actively responding.

[0056] During the monitoring phase, as the gust front approached, the unit structure began to exhibit a weak response. The second pulse signal output from the satellite positioning receiver served as a unified clock reference, synchronously triggering the inertial measurement unit (IMU) and tilt sensor to acquire data. The monitoring module detected an abnormal oscillation in the three-dimensional displacement of the tower top, with an amplitude reaching ±15 mm. Simultaneously, angular velocity spectrum analysis output from the IMU showed a slight shift of approximately 0.02 Hz in the tower's first natural frequency. The system performed timestamp alignment and quality marking processing on the initial multi-source heterogeneous data, constructing a multi-dimensional spatiotemporal attitude tensor, and fully recording the dynamic response history of the unit under extreme wind conditions.

[0057] During the fusion and decision-making phase, the system inputs the constructed spatiotemporal attitude tensor into a pre-trained convolutional neural network. The convolutional kernel slides along the time axis, automatically extracting deep feature vectors characterizing structural health degradation, including features of tower damping ratio decay and nacelle asymmetric displacement accumulation. The decision tree classifier outputs a health status classification result based on this feature vector, showing a probability of 0.75 for the unit currently in a "moderate degradation" state and 0.15 for a "severe degradation" state. The system establishes a coupled risk matrix between the external wind risk level and the health status classification result, using "high risk" and "moderate degradation" as indices for querying. Although a conventional matrix query might point to a lower initial risk level, if the system detects that the external wind risk level has reached the preset high risk level and the health status classification result is moderate degradation, it triggers a risk level escalation logic, ultimately raising the coupled risk level to "protection level."

[0058] During the execution phase, the control module responds to the "protection level" of coupling risk and proactively intervenes before the hard safety chain activates. The system issues a fault-tolerant command, limiting the pitch rate limit from the usual 8 degrees / second to 3 degrees / second and locking the yaw system to prevent a surge in additional loads during yaw. This strategy effectively suppresses drastic fluctuations in aerodynamic loads and achieves fault isolation without triggering unit shutdown and ensuring power generation revenue.

[0059] Subsequently, the gust of wind acted fully on the unit. Because the pitch rate was limited, the aerodynamic energy absorbed by the impeller was controlled, and although the unit speed increased, it did not exceed the danger threshold. If a traditional hard safety chain protection strategy were used, the brakes would lock immediately upon detecting abnormal speed, causing an impact torque to the transmission chain exceeding three times the rated load. However, this embodiment, through the early intervention of a tiered safety strategy, reduced the impact load by approximately 40%, significantly extending the fatigue life of the gearbox and main shaft bearings. After the gust of wind passed, the system automatically released the restrictions, and the unit resumed normal operation.

[0060] As can be seen from this embodiment, the method provided in this application can accurately identify wind condition-structure coupling risks and effectively avoid secondary damage caused by hard shutdown through graded flexible protection.

[0061] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application, such as using other types of deep learning models to replace convolutional neural networks for feature extraction, or adjusting the specific triggering thresholds and execution parameters of the hierarchical security strategy, should be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0062] Based on the above method, this application also provides a wind turbine safety monitoring system based on lidar and satellite, corresponding to the above method. Figure 2 A structural block diagram of a wind turbine safety monitoring system based on lidar and satellite according to an embodiment of this application is shown. See also... Figure 2 As shown, the system includes: External acquisition module 10 is used to acquire the external wind risk level, which is determined based on the advanced perception data of the wind field in front of the wind turbine collected by the nacelle lidar. Data construction module 20 is used to synchronously collect initial multi-source heterogeneous data of the wind turbine with a unified clock reference, preprocess the initial multi-source heterogeneous data to obtain actual multi-source heterogeneous data, and construct spatiotemporal attitude data based on the actual multi-source heterogeneous data. The risk level determination module 30 is used to determine the coupling risk level based on the external wind condition risk level and the spatiotemporal attitude data. The policy execution module 40 is used to execute a graded security policy based on the coupling risk level.

[0063] Based on the same inventive concept disclosed above, this application also provides an electronic device. The electronic device of this application includes at least one processor and at least one memory electrically connected to the processor. The memory is electrically connected to the processor, wherein the memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method described above.

[0064] It should be noted that the electrical connections between the above-mentioned units do not necessarily represent the connections between lines. Indirect connections are applicable to the embodiments of this application as long as they achieve the purpose of this application.

[0065] Based on the same inventive concept, this application also provides a computer storage medium storing a computer program, which, when executed by a processor, implements the steps of the above method.

[0066] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications 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 this application.

Claims

1. A method for safety monitoring of wind turbine generators based on lidar and satellite, characterized in that, The method includes, The external wind risk level is obtained, which is determined based on the advanced perception data of the wind field in front of the wind turbine collected by the nacelle lidar. Initial multi-source heterogeneous data of the wind turbine are collected synchronously using a unified clock reference. The initial multi-source heterogeneous data is preprocessed to obtain actual multi-source heterogeneous data. Spatiotemporal attitude data is constructed based on the actual multi-source heterogeneous data. The coupling risk level is determined based on the external wind condition risk level and the spatiotemporal attitude data. Based on the aforementioned coupling risk level, a tiered security strategy is implemented.

2. The method according to claim 1, characterized in that, Obtain the external wind risk level, specifically including: Acquire advanced sensing data of the wind field in front of the wind turbine; Based on the aforementioned advanced sensing data, wind speed spatiotemporal gradient features are obtained, and extreme wind events are identified based on these features. The risk level of the external wind conditions is determined based on extreme wind events.

3. The method according to claim 1, characterized in that, The initial multi-source heterogeneous data is preprocessed to obtain actual multi-source heterogeneous data. Spatiotemporal attitude data is then constructed based on the actual multi-source heterogeneous data, specifically including: The initial multi-source heterogeneous data is processed by timestamp alignment and quality marking to obtain the actual multi-source heterogeneous data; A multidimensional spatiotemporal attitude tensor is constructed based on the actual multi-source heterogeneous data to obtain the spatiotemporal attitude data.

4. The method according to claim 3, characterized in that, The initial multi-source heterogeneous data was acquired by an integrated satellite positioning receiver, inertial measurement unit, and tilt sensor; The unified clock reference is the second pulse signal output by the satellite positioning receiver; The initial multi-source heterogeneous data includes at least: the three-dimensional displacement and heading angle output by the satellite positioning receiver, the angular velocity and acceleration output by the inertial measurement unit, and the static tilt angle output by the tilt sensor.

5. The method according to claim 1, characterized in that, The coupling risk level is determined based on the external wind risk level and the spatiotemporal attitude data, specifically including: Feature extraction is performed on the spatiotemporal attitude data to obtain spatiotemporal feature vectors; Based on the spatiotemporal feature vector, health status classification is performed to obtain health status classification results; Establish a coupled risk matrix between the external wind condition risk level and the health status classification result; The coupling risk level is determined based on the coupling risk matrix; The health status classification results include five categories: normal, mild deterioration, moderate deterioration, severe deterioration, and dangerous.

6. The method according to claim 5, characterized in that, Feature extraction is performed on the spatiotemporal attitude data to obtain a spatiotemporal feature vector, specifically including: The spatiotemporal attitude data is input into a convolutional neural network; By sliding the convolution kernel of a convolutional neural network along the time axis, spatiotemporal feature vectors representing structural health degradation are extracted.

7. The method according to claim 5, characterized in that, Based on the spatiotemporal feature vector, health status classification is performed to obtain the health status classification result, which specifically includes: The spatiotemporal feature vector is input into a decision tree classifier to obtain the probability values ​​of the wind turbine under various health states, including normal, mild degradation, moderate degradation, severe degradation, and dangerous conditions. Based on the probability value, the health status classification result is determined.

8. The method according to claim 5, characterized in that, The coupling risk level is determined based on the coupling risk matrix, specifically including: Using the external wind risk level and the health status classification result as indexes, the corresponding initial risk level is obtained from the coupled risk matrix; If the external wind risk level reaches the preset high risk level and the health status classification result belongs to at least one of moderate degradation, severe degradation, or danger, the initial risk level is increased to obtain the coupled risk level.

9. The method according to claim 8, characterized in that, Based on the aforementioned coupling risk level, a tiered security strategy is implemented, specifically including: In response to the coupling risk level being at the early warning level, the data sampling frequency is increased, an inspection work order is generated, and early warning information is pushed to the central control center; In response to the coupling risk level being protection level, a fault-tolerant command is issued before the hard safety chain is activated to limit the upper limit of the pitch rate and lock the yaw system action, thereby achieving fault isolation without triggering a shutdown. In response to the coupling risk level being dangerous, the conventional pitch PID control is bypassed before the hard safety chain is activated, and a flexible pitch retraction action is performed at a preset gentle slope to smoothly transition the speed of the wind turbine to a safe range.

10. A wind turbine safety monitoring system based on lidar and satellite, characterized in that, The system includes: An external acquisition module is used to acquire the external wind risk level, which is determined based on the advanced perception data of the wind field in front of the wind turbine collected by the nacelle lidar. The data construction module is used to synchronously collect the initial multi-source heterogeneous data of the wind turbine with a unified clock reference, preprocess the initial multi-source heterogeneous data to obtain the actual multi-source heterogeneous data, and construct spatiotemporal attitude data based on the actual multi-source heterogeneous data. The risk level determination module is used to determine the coupling risk level based on the external wind condition risk level and the spatiotemporal attitude data. The policy execution module is used to execute graded security policies based on the coupling risk level.

Images