Ocean buoy wind speed and wind direction monitoring method and system

Abstract

The invention belongs to the technical field of wind speed and wind direction monitoring, and particularly relates to an ocean buoy wind speed and wind direction monitoring method and system, and the method comprises the following steps: S1, setting a sampling period, and obtaining data information through a sensor; s2, preprocessing the data to obtain processed wind speed and wind direction data; s3, calculating the confidence coefficient of each sensor, and troubleshooting fault data; s4, performing weighted fusion on the data, and outputting a final wind direction and a final wind speed; the method has the advantages that in combination with a real-time confidence evaluation mechanism, when any single sensor fails due to salt spray corrosion, mechanical failure or electromagnetic interference, continuous output of wind speed and wind direction data can still be guaranteed, and the availability of the system is improved; motion noise suppression is carried out based on IMU-environment joint compensation; an adaptive decision engine and a self-diagnosis engine are constructed, the weight of a distortion sensor is reduced under extreme working conditions, and the accuracy of measurement is ensured.

Description

Technical Field

[0001] This application belongs to the field of wind speed and direction monitoring technology, specifically relating to a method and system for monitoring wind speed and direction of marine buoys. Background Technology

[0002] Ocean buoys, as unmanned automated observation platforms, play an irreplaceable role in marine meteorological monitoring, disaster early warning, and wind energy development. One of their core functions is the accurate measurement of wind speed and direction, which not only affects the accuracy of weather forecasts but also directly impacts the safety of offshore operations, the effectiveness of typhoon warnings, and the site selection and operational efficiency of offshore wind farms. Buoy wind measurement technology has evolved from mechanical to electronic and then to remote sensing. Early methods mainly relied on mechanical anemometers, converting wind speed into the rotational speed of the cups or propellers, but these methods suffered from mechanical wear and inertial lag. The six degrees of freedom motion of the buoy under wave action (roll, pitch, yaw, sway, and swell) is a core issue affecting wind measurement accuracy. Especially under extreme sea conditions (such as during typhoons), the buoy's violent motion can cause lateral wind speed measurement deviations of up to 30%.

[0003] A single sensor struggles to distinguish between real wind speed fluctuations and noise caused by the buoy's own swaying, especially under extreme conditions such as typhoons and high sea states, where errors can be significant. Existing data fusion methods often employ simple weighted averaging, failing to adequately consider the dynamic correlation between real-time sensor confidence and environmental interference. Summary of the Invention

[0004] This application proposes a marine buoy wind speed and direction monitoring system and method based on multi-sensor data fusion and adaptive decision-making. The system addresses the aforementioned problems through multi-sensor redundant detection, data fusion, and motion compensation. The technical solution is as follows: A method for monitoring wind speed and direction using marine buoys includes the following steps: S1. Set the sampling period and use the sensor to acquire data information; S2. Preprocess the data to obtain the processed wind speed and wind direction data; S3. Calculate the confidence level of each sensor and troubleshoot faulty data; S4. Data weighting and fusion to output the final wind direction and final wind speed.

[0005] Preferably, the data preprocessing steps are as follows: S21. Sensor data input is synchronized with time; S22. An error-state-based Kalman filter is used to fuse IMU and BeiDou satellite data to estimate the buoy's attitude, position, and velocity. At the same time, the sensor error of the IMU is estimated and corrected. BeiDou positioning data is used for error covariance calculation. Combining prediction and observation information, the state update is completed in the Kalman filter update module, and the attitude quaternion of the carrier is output. S23. Using the calculated attitude quaternions, the wind speed data measured by the wind speed and direction sensors in the carrier coordinate system is transformed to the geographic coordinate system to obtain the time-domain wind speed signal after preliminary motion compensation; S24. Frequency Domain Filtering Processing: After windowing, FFT transformation, band-stop filtering based on dynamic setting of stopband based on wave energy, inverse FFT, overlapping and addition, and zero-phase processing, the final output is the motion-compensated wind speed and direction.

[0006] Preferably, the window function in frequency domain filtering is: ; The Fourier transform (FFT) is as follows: ; For the first k The complex result of each frequency component, where n is the index of the time-domain sampling point, ranging from 0 to 1023; Calculate whether the wave energy exceeds a preset threshold: ; ; N is the number of points in the FFT. For frequency resolution, It is the sampling rate. It is the nth time-domain sampled value after windowing.

[0007] Preferably, the frequency domain wind speed vector obtained by FFT transformation is multiplied by an 8th-order band-stop filter to filter out frequency components within the stopband. Inverse FFT transform and zero-phase processing: The filtered frequency domain signal is converted back to the time domain through inverse FFT transform. After two filtering cycles (forward and reverse), zero-phase processing is achieved. The same 8th-order Butterworth band-stop filter is used. Overlapping and adding reconstruction: ; Zero-phase processing: ; This is the processed signal of the m-th frame. Original input signal, H{} filtering operation, reverse() signal inversion operation, Frame shift time, ( ) No. The position of the segment signal on the time axis; Obtain wind speed and direction data after motion compensation and frequency domain filtering; ; True heading angle.

[0008] Preferred, including: S31. Collect raw data from each sensor and use the time synchronization vector generated in step S2 to generate an anomaly detection matrix; The input data includes: ultrasonic anemometer data, mechanical anemometer data, barometric gradient sensor data, and IMU attitude data. S32. The confidence assessment model is as follows: ; in: The output of the variance detection term is 1 when there is no fault and 0 when there is a fault. : The output of the gradient detection term; defined as , The total number of samples; : Attitude over-limit item; 1 when there is no over-limit, 0 when there is an over-limit; Sensor reliability factor; Let be the weighting coefficient, satisfying ; S33. Fault Isolation: Based on the confidence assessment results, the detected faulty sensor data is removed and not included in the subsequent fusion calculation.

[0009] Preferred, including: S41. Data Preprocessing and Standardization Stage: The motion-compensated data first undergoes data standardization processing; S42. The normalized data is then fed into the weight calculation model; The calculation model is as follows: ; in: Sensor health confidence ; The signal-to-noise ratio factor of sensor r; Environmental adaptability factor: ; Sensor reliability factor; M: The number of effective sensors currently participating in the fusion; S43. Fusion Decision: Based on these weights, data fusion decisions are made to process two different types of measurement data, wind speed and wind direction, respectively. S44. Data Fusion: Wind speed fusion: Wind speed is a scalar quantity and a weighted average algorithm is used; Wind direction fusion: Wind direction is an angle and has periodicity. It is converted into a vector and then synthesized. S45. Confidence Rating and Feedback: The credibility of the fusion results is assessed, and the rating results are fed back to the weight calculation core to form a closed-loop feedback mechanism, which is used to assess the priority of data transmission or trigger system maintenance alarms; S46. After weighted fusion processing, two final results are output: Final wind speed: The wind speed measurement after wind speed fusion. ; Final wind direction: The wind direction measurement after fusion. .

[0010] Preferably, the S44 wind direction fusion steps are as follows: a. Convert wind direction angle to unit vector: For wind direction angle Its unit vector in the northeast celestial coordinate system for: ; X component Represents the east direction; Y component Represents the north direction; b. Perform weighted vector synthesis: ; The sum of the eastward weighted wind direction vectors. This is the weighted northward wind direction vector sum; c. Calculate the final wind direction from the composite vector: ; The atan2 function can correctly process all quadrants, obtaining the final wind direction between 0 and 360 degrees, and the magnitude of the composite vector. It can reflect the consistency of wind direction data; the shorter the modulus, the greater the difference in wind direction between sensors.

[0011] Preferably, the final confidence level of S45 is defined as the weighted sum of the weights of all participating sensors: ; in The closer the result is to 1, the higher the confidence level of the final result, indicating that the final result is determined by a sensor with high confidence.

[0012] A marine buoy wind speed and direction monitoring system includes an information acquisition unit, a motion compensation unit, a self-diagnosis and processing unit, a weighted fusion unit, and an adaptive decision-making unit. Information acquisition unit: Collects raw data from wind speed and direction sensors; Motion compensation unit: Synchronizes multi-source sensor data in time, unifying data from IMU, BeiDou positioning, and wind direction and speed sensors to a 50Hz time series. It fuses IMU and BeiDou positioning data using a Kalman filter, with IMU used for state prediction and BeiDou positioning for state updates, and performs error suppression during the update process. The Kalman filter outputs the buoy's attitude quaternion, used to transform the wind speed vector measured by the wind direction and speed sensors from the buoy coordinate system to the Earth coordinate system. The transformed wind speed signal then undergoes frequency domain filtering, including windowing, FFT transformation, band-stop filtering based on dynamic stopband setting based on wave energy, inverse FFT, overlap addition, and zero-phase processing, finally outputting the motion-compensated wind speed and direction. Self-diagnostic processing unit: Identifies typical fault modes through fault characteristics and quantifies the health status of each sensor, providing a basis for decision-making in subsequent data fusion; Weighted fusion unit: Combines multiple wind speed and direction data streams with the confidence scores of each sensor output by the self-diagnostic processing unit through an adaptive weighting algorithm to output a set of optimal and final wind speed and direction values; Adaptive Decision Unit: Evaluates the credibility of the fusion results and feeds the rating results back to the adaptive weighting algorithm to form a closed-loop feedback mechanism, which is used to evaluate the priority of data transmission or trigger system maintenance alarms.

[0013] Preferably, the information acquisition unit includes: raw data from ultrasonic anemometers, mechanical anemometers, and barometric pressure gradients; data from IMU 9-axis attitude sensors; temperature and humidity values; and BeiDou satellite positioning data. Self-diagnostic processing unit: Data input phase: Collect raw data from various sensors, use the time synchronization vector generated by the motion compensation unit to generate an anomaly detection matrix; the input data are: ultrasonic anemometer data, mechanical anemometer data, barometric gradient sensor data, and IMU attitude data; Fault type identification: Sensor type, faults are divided into 3 categories: a. The fault type of the mechanical anemometer is mechanical jamming. Calculate the variance of the sequence samples of the mechanical anemometer data within the time window and compare it with the threshold. If the variance of the samples in three consecutive time windows is lower than the threshold, then a mechanical jamming fault can be judged. b. The pitch and roll angles calculated by the IMU sensor exceed the thresholds, which are derived from buoy design data; c. Fault diagnosis of ultrasonic, mechanical anemometers, barometric pressure gradient sensors, and temperature and humidity sensors: Calculate the forward first-order difference of each point in the data sequence and compare it with a threshold. If the threshold is exceeded for p consecutive time windows, a jump fault occurs. The threshold is set by the physical characteristics of the sensor.

[0014] Compared with the prior art, the beneficial effects of this application are as follows: Breaking the reliability limit of a single sensor: By using a redundant architecture of ultrasonic + mechanical + barometric pressure gradient sensors, combined with a real-time confidence assessment mechanism, the system can still ensure continuous output of wind speed and direction data even if any single sensor fails or fails, thus improving the availability of the system.

[0015] Eliminating measurement distortion caused by buoy motion: Based on a 9-axis IMU and North satellite positioning, a motion compensation algorithm was designed, which upgrades the traditional scalar compensation to vector coordinate transformation, effectively suppressing the interference of buoy motion on wind speed and direction measurement and improving the accuracy of wind speed and direction measurement; Achieve adaptive monitoring in extreme environments: Build an adaptive decision model to automatically switch to the optimal data source under extreme conditions (such as shielding mechanical sensors when humidity is high and using air pressure gradient calculation when the tilt angle is large) to improve system availability; System self-diagnostic capability: To achieve real-time identification and dynamic shielding of sensor-level faults, a confidence assessment mechanism is set up, and a data weighted fusion algorithm is designed to improve the measurement accuracy of the system. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the application process. Figure 2 This is a flowchart of the self-diagnosis processing unit; Figure 3 Flowchart of the weighted fusion unit; Figure 4 A comparison chart of the effects of multi-sensor fusion; Figure 5 The result of fault type identification is shown in the image. Detailed Implementation

[0017] The technical solution of this application will be described in detail below with reference to specific embodiments and accompanying drawings. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solution of this application, rather than limitations thereof. Specific technical features can be combined with each other.

[0018] A marine buoy wind speed and direction monitoring system includes an information acquisition unit, a motion compensation unit, a self-diagnosis and processing unit, a weighted fusion unit, and an adaptive decision-making unit. Information acquisition unit: collects raw data from wind speed and direction sensors; sensor installation locations are as follows: ultrasonic sensors and anemometers are installed on the mast; air pressure gradient sensors and temperature and humidity sensors are installed on the mast base; IMU and Beidou module are installed in the buoy cabin.

[0019] Motion compensation unit: Time synchronization of multi-source sensor data, unifying data from IMU, BeiDou positioning, and wind direction and speed sensors to a 50Hz time series, and fusing IMU and BeiDou positioning data through a Kalman filter, where IMU is used for state prediction and BeiDou positioning is used for state update, with error suppression performed during the update process; the Kalman filter outputs the buoy's attitude quaternion, which is used to transform the wind speed vector measured by the wind direction and speed sensors from the buoy coordinate system to the Earth coordinate system; the transformed wind speed signal is then processed by frequency domain filtering, including windowing, FFT transformation, band-stop filtering based on dynamic setting of stopbands based on wave energy, inverse FFT, overlap addition, and zero-phase processing, finally outputting the motion-compensated wind speed and direction.

[0020] Self-diagnostic processing unit: Identifies typical fault modes through fault characteristics and quantifies the health status of each sensor, providing a basis for decision-making in subsequent data fusion.

[0021] Weighted fusion unit: Combines multiple wind speed and direction data streams with the confidence scores of each sensor output by the self-diagnostic processing unit, and outputs a set of optimal and final wind speed and direction values ​​through an adaptive weighting algorithm.

[0022] Adaptive Decision Unit: Evaluates the credibility of the fusion results and feeds the rating results back to the adaptive weighting algorithm to form a closed-loop feedback mechanism, which is used to evaluate the priority of data transmission or trigger system maintenance alarms.

[0023] The system's final calculation results are sent to a satellite via a satellite module and then forwarded to the customer.

[0024] A method for monitoring wind speed and direction using marine buoys includes the following steps: S1. Set the sampling period and use the sensor to acquire data information; This stage involves collecting raw data from wind speed and direction sensors, including ultrasonic anemometers, mechanical anemometers, and barometric pressure gradients; collecting values ​​from the IMU 9-axis attitude sensor; collecting temperature and humidity values; and collecting BeiDou satellite positioning data. However, due to the inconsistent data transmission cycles of the various sensors, time synchronization is required. In this invention, a 50Hz data sampling cycle is used. BeiDou positioning data is used to calibrate the IMU data to eliminate accumulated errors caused by the IMU.

[0025] S2. Preprocess the data to obtain the processed wind speed and wind direction data; The first step involves time synchronization of data from multiple sensors, unifying data from the IMU, BeiDou positioning, and wind direction and speed sensors to a 50Hz time series. Then, a Kalman filter is used to fuse the IMU and BeiDou positioning data, with the IMU used for state prediction and the BeiDou positioning for state updates. Error suppression is performed during the update process (including HDOP adaptive adjustment of observation noise, zero-bias correction, and zeroing of accumulated errors). The Kalman filter outputs the buoy's attitude quaternion, which is used to transform the wind speed vector measured by the wind direction and speed sensors from the buoy coordinate system to the Earth coordinate system. The transformed wind speed signal then undergoes frequency domain filtering, including windowing, FFT transform, band-stop filtering based on dynamic stopband setting according to wave energy, inverse FFT, overlap addition, and zero-phase processing, finally outputting the motion-compensated wind speed and direction.

[0026] 1) Sensor data input and time synchronization: The algorithm first receives three types of sensor data. All sensor data need to be time synchronized to ensure that the data is aligned on the time axis and to generate a vector with time variables.

[0027] Time synchronization is achieved using linear interpolation.

[0028] The system uses RTC timekeeping internally; after receiving the BeiDou satellite time, it uses the BeiDou satellite time to calibrate the system's RTC time, eliminating system errors caused by the frequency deviation of the RTC crystal oscillator.

[0029] 2) Kalman filtering and motion compensation: a. Kalman filter core processing: The system employs an error-state-based Kalman filter to fuse IMU and BeiDou satellite data to estimate the buoy's attitude, position, and velocity, while simultaneously estimating and correcting IMU sensor errors. IMU data is input to the Kalman prediction module at a frequency of 50Hz. The prediction module model is as follows: ; :exist kThe pose quaternion predicted at time (current time). This is a 4-dimensional vector, typically in the form [q w ,q x ,q y ,q z ] T , where q w It is the real part. It represents the rotational orientation of the object in three-dimensional space.

[0030] The sampling time interval is usually measured in seconds.

[0031] In this application , The sampling frequency.

[0032] ： exist k The raw angular velocity value measured by the gyroscope at any given moment. This is a 3D vector. , , ] T , which represents the angular velocity of an object rotating about the X, Y, and Z axes (the unit is usually rad / s).

[0033] ： Gyroscope bias. This is also a 3D vector. Even when not rotating, a gyroscope has a tiny output; this inherent error is the bias. It's one of the state variables that Kalman filtering needs to estimate. Subtracting the bias from the measurement value yields a more accurate angular velocity.

[0034] A skew-symmetric matrix formed by the corrected angular velocities.

[0035] The 4x4 skew-symmetric matrix is ​​the core of the quaternion derivative equation. Its function is to transform the angular velocity vector into a matrix form that can be multiplied by quaternions, and its physical meaning represents the instantaneous rotational rate.

[0036] BeiDou positioning data is used for error covariance calculation. Combined with prediction and observation information, the state update is completed in the Kalman filter update module, and the attitude quaternion q of the carrier is output.

[0037] The system adaptively adjusts the filter parameters based on the satellite geometric distribution quality factor (HDOP), corrects the zero-point drift of the sensor in real time, and periodically clears the accumulated errors in the filter based on BeiDou satellite positioning data, thereby improving the system's stability and accuracy.

[0038] c. Coordinate system transformation: Using the calculated attitude quaternions, the wind speed data measured by the wind speed and direction sensor in the carrier coordinate system is transformed to the geographic coordinate system to obtain the time-domain wind speed signal after preliminary motion compensation.

[0039] The formula for calculating quaternions is: ; The motion-compensated time-domain wind speed signal is: ; The wind speed and direction values ​​are located at the northeast celestial coordinates. The values ​​represent wind speed and direction in the buoy coordinate system.

[0040] 3) Frequency domain filtering: a. Preprocessing stage: The time-domain wind speed signal is windowed using a Hamming window with N=1024 points. The time-domain signal is converted to the frequency domain using FFT transformation, which facilitates feature extraction and analysis in the frequency domain.

[0041] The window function is: ; The FFT transformation is as follows: ; Calculate whether the wave energy exceeds a preset threshold: ; ; For the first k The complex result of each frequency component is the complex spectrum calculated by FFT; n Here, N represents the index of the time-domain sampling points, ranging from 0 to 1023, where N is the number of points in the FFT. For frequency resolution, It is the sampling rate (50Hz in this application). It is the nth time-domain sampled value after windowing.

[0042] If the wave energy exceeds the threshold, set the stopband upper limit to 0.4Hz; if the wave energy does not exceed the threshold, set the stopband upper limit to 0.3Hz.

[0043] c. An 8th-order Butterworth band-stop filter is used for filtering, with a stopband range from 0.05Hz to a dynamically set upper limit of 0.4Hz / 0.3Hz, to filter out wind speed noise caused by waves. The 8th-order Butterworth band-stop filter is: ; The filter frequency represents the frequency of the signal after passing through the filter. The gain (attenuation or amplification factor) of the component.

[0044] The frequency-domain wind speed vector obtained by FFT transformation is multiplied by an 8th-order Butterworth band-stop filter to filter out frequency components within the stopband.

[0045] d. Inverse FFT Transform and Zero-Phase Processing: The filtered frequency domain signal is converted back to the time domain through inverse FFT transformation. Since 50% overlap is used in the windowed frame processing, the overlapping parts of adjacent frames need to be added in the time domain to reconstruct the continuous signal. Signals that undergo continuous time-domain-frequency-time-domain conversion will have phase distortion problems. Therefore, the design is to achieve zero-phase processing by filtering twice in both forward and reverse directions. The same 8th-order Butterworth band-stop filter is used. Overlapping and adding reconstruction: ; Zero-phase processing: ; This is the processed signal of the m-th frame. Original input signal, H{} filtering operation, reverse() signal inversion operation, Frame shift time, ( ) No. The position of the segment signal on the time axis.

[0046] Obtain wind speed and direction data after motion compensation and frequency domain filtering; ; True heading angle; The eastward velocity component is the velocity of the carrier in the eastward direction, usually measured in m / s. The northward velocity component represents the velocity of the carrier in the northward direction.

[0047] S3. Calculate the confidence level of each sensor and troubleshoot faulty data; This stage uses sensor-collected data to identify typical fault modes through fault characteristics and quantifies the health status (confidence level) of each sensor, providing a basis for decision-making in subsequent data fusion.

[0048] 1) Data Input Phase: The system collects raw data from various sensors and uses the time synchronization vector generated in Phase 2 to generate an anomaly detection matrix. Input data includes: ultrasonic anemometer data, mechanical anemometer data, barometric gradient sensor data, and IMU attitude data.

[0049] 2) Fault type determination: Sensor type, faults are divided into 3 categories: a. The fault type of the mechanical anemometer is mechanical jamming, which manifests as no change or minimal change in the data collected by the anemometer over time, resulting in a significant reduction in the variance of the data sequence.

[0050] Calculate the sequence sample variance of the mechanical anemometer data within the time window (sampling period value) and compare it with a threshold. If the sample variance value of three consecutive time windows is lower than the threshold, a mechanical jamming fault can be judged. The threshold can be determined experimentally.

[0051] b. The pitch and roll angles calculated by the IMU sensor exceed the thresholds, which are derived from buoy design data.

[0052] c. Data collected by ultrasonic sensors, mechanical anemometers, barometric pressure gradient sensors, and temperature and humidity sensors exhibit large-scale jumps over time. The forward first-order difference of each point in the data sequence is calculated and compared to a threshold. If the threshold is exceeded for 10 consecutive time windows, a jump fault occurs. The threshold is set based on the physical characteristics of the sensors and can be determined from the manual. The first-order difference model is as follows: ; 3) Confidence Assessment: This module outputs a confidence score between 0 and 1 for each sensor to quantify its health status. A confidence score of 1 indicates complete reliability, while a confidence score of 0 indicates complete failure, rendering the data infeasible and requiring isolation. The confidence assessment model is as follows: ; in: The output of the variance detection term is 1 when there is no fault and 0 when there is a fault; it can also be designed as a continuous function, such as... ; : The output of the gradient detection term; defined as ; The total number of samples; : Attitude over-limit item; 1 when there is no over-limit, 0 when there is an over-limit; Sensor reliability factor (e.g., the proportion of time without failure in the past 24 hours); For the weighting coefficients, satisfying .

[0053] Fault isolation: Based on the confidence assessment results, the system performs fault isolation operations, separating the detected faulty sensors or functional modules from the normal system and preventing them from participating in subsequent fusion calculations, thus preventing the fault from spreading and affecting the overall system performance.

[0054] Figure 5 As shown, by employing an adaptive weighted dynamic strategy, accuracy is improved by 36% under normal operating conditions, 78% when mechanical sensor malfunctions, 79% when ultrasonic data jumps, and 57% under severe sea conditions.

[0055] Figure 4 As shown, the fault detection accuracy improved by 98.2%, the system availability improved by 99.8%, and the wind speed and direction measurement accuracy improved by 42.1%.

[0056] S4. Data weighting and fusion to output the final wind direction and final wind speed.

[0057] The core task of data weighted fusion is to receive multiple wind speed and direction data streams from the motion compensation unit and the confidence scores of each sensor output by the self-diagnostic processing unit, and then output a set of optimal final wind speed and direction values ​​through an adaptive weighting algorithm. Its core idea is to give greater weight to more reliable sensor data in the final result.

[0058] Data preprocessing and standardization stage: The motion-compensated data is first standardized by means of mean-variance normalization to eliminate the dimensional differences between data from different sensors and by using the Z-score standardization method.

[0059] 2) Weight Calculation Model: The normalized data enters the core weight calculation module, which comprehensively considers the following factors to determine the weight of each sensor data point: a. Sensor health status: Assess the current operating status and performance of each sensor; b. Environmental parameters: Adjust the suitability weights of each sensor according to the current environmental conditions; c. Weight Arbitrator: Arbitrates and optimizes the calculation results of different weights; d. Historical reliability; The calculation model is as follows: ; in: Sensor health confidence ; : The signal-to-noise ratio (SNR) factor of sensor r; the higher the SNR, the greater the weight. It can be approximated as... (Ratio of mean to standard deviation), and normalize the results; Environmental adaptability factor: ; Sensor reliability factor; M: The number of effective sensors currently participating in the fusion.

[0060] 3) Fusion Decision: The fusion decision unit receives the calculated weight information and makes data fusion decisions based on these weights, processing two different types of measurement data, wind speed and wind direction, respectively.

[0061] 4) Data Fusion: The fusion decision unit divides the processing into two parallel paths: Wind speed fusion: Wind speed is a scalar quantity and a weighted average algorithm is used; Wind direction fusion: Wind direction is an angle and has periodicity. It is converted into a vector and then synthesized.

[0062] a. Convert wind direction angle to unit vector: For wind direction angle Its unit vector in the northeast celestial coordinate system for: ; X component Represents the east direction; Y component Represents the north direction; b. Perform weighted vector synthesis: ; The sum of the eastward weighted wind direction vectors. This is the north-weighted wind direction vector sum.

[0063] c. Calculate the final wind direction from the composite vector: ; The atan2 function can correctly process all quadrants, obtaining the final wind direction between 0 and 360 degrees, and the magnitude of the composite vector. It can reflect the consistency of wind direction data; the shorter the modulus, the greater the difference in wind direction between sensors.

[0064] 5) Confidence rating and feedback: The credibility of the fusion results is evaluated and the rating results are fed back to the weight calculation core to form a closed-loop feedback mechanism, which is used to evaluate the priority of data transmission or trigger system maintenance alarms.

[0065] Final confidence level: Defined as the weighted sum of the weights of all sensors involved in the fusion.

[0066] ; in The closer the result is to 1, the higher the confidence level of the final result, indicating that the final result is determined by a sensor with high confidence.

[0067] Consistency index: Vector magnitude Consistency index for wind direction; weighted standard deviation of wind speed is a consistency index for wind speed.

[0068] 6) After weighted fusion processing, the system outputs two final results: Final wind speed: The wind speed measurement after wind speed fusion. .

[0069] Final wind direction: The wind direction measurement after fusion. .

[0070] The entire data weighted fusion engine achieves intelligent fusion of multi-source sensor data through multi-dimensional weight calculation and closed-loop feedback mechanism, thereby improving the accuracy and reliability of wind speed and direction measurement.

[0071] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for monitoring wind speed and direction using marine buoys, characterized in that, Includes the following steps: S1. Set the sampling period and use the sensor to acquire data information; S2. Preprocess the data to obtain the processed wind speed and wind direction data; S3. Calculate the confidence level of each sensor and troubleshoot faulty data; S4. Data weighting and fusion to output the final wind direction and final wind speed.

2. The method for monitoring wind speed and direction using a marine buoy according to claim 1, characterized in that, The data preprocessing steps are as follows: S21. Sensor data input is synchronized with time; S22. An error-state-based Kalman filter is used to fuse IMU and BeiDou satellite data to estimate the buoy's attitude, position, and velocity. At the same time, the sensor error of the IMU is estimated and corrected. BeiDou positioning data is used for error covariance calculation. Combining prediction and observation information, the state update is completed in the Kalman filter update module, and the attitude quaternion of the carrier is output. S23. Using the calculated attitude quaternions, the wind speed data measured by the wind speed and direction sensors in the carrier coordinate system is transformed to the geographic coordinate system to obtain the time-domain wind speed signal after preliminary motion compensation; S24. Frequency Domain Filtering Processing: After windowing, FFT transformation, band-stop filtering based on dynamic setting of stopband based on wave energy, inverse FFT, overlapping and addition, and zero-phase processing, the final output is the motion-compensated wind speed and direction.

3. The method for monitoring wind speed and direction using a marine buoy according to claim 2, characterized in that, The window function in frequency domain filtering is: ； The Fourier transform (FFT) is as follows: ； For the first The complex result of each frequency component, where n is the index of the time-domain sampling point, ranging from 0 to 1023; Calculate whether the wave energy exceeds a preset threshold: ； ; N is the number of points in the FFT. For frequency resolution, It is the sampling rate. It is the nth time-domain sampled value after windowing.

4. The method for monitoring wind speed and direction of a marine buoy according to claim 3, characterized in that, The frequency domain wind speed vector obtained by FFT transformation is multiplied by an 8th-order band-stop filter to filter out frequency components within the stopband. Inverse FFT transform and zero-phase processing: The filtered frequency domain signal is converted back to the time domain through inverse FFT transform. After two filtering cycles (forward and reverse), zero-phase processing is achieved. The same 8th-order Butterworth band-stop filter is used. Overlapping and adding reconstruction: ； Zero-phase processing: ； This is the processed signal of the m-th frame. Original input signal, H{} filtering operation, reverse() signal inversion operation, Frame shift time, ( ) No. The position of the segment signal on the time axis; Obtain wind speed and direction data after motion compensation and frequency domain filtering; ； True heading angle.

5. The method for monitoring wind speed and direction using a marine buoy according to claim 3, characterized in that, include: S31. Collect raw data from each sensor and use the time synchronization vector generated in step S2 to generate an anomaly detection matrix; The input data includes: ultrasonic anemometer data, mechanical anemometer data, barometric gradient sensor data, and IMU attitude data. S32. The confidence assessment model is as follows: ； in: The output of the variance detection term is 1 when there is no fault and 0 when there is a fault. : The output of the gradient detection term; defined as , The total number of samples; : Attitude over-limit item; 1 when there is no over-limit, 0 when there is an over-limit; Sensor reliability factor; Let be the weighting coefficient, satisfying ; S33. Fault Isolation: Based on the confidence assessment results, the detected faulty sensor data is removed and not included in the subsequent fusion calculation.

6. The method for monitoring wind speed and direction of a marine buoy according to claim 3, characterized in that, include: S41. Data Preprocessing and Standardization Stage: The motion-compensated data first undergoes data standardization processing; S42. The normalized data is then fed into the weight calculation model; The calculation model is as follows: ； in: Sensor health confidence ; The signal-to-noise ratio factor of sensor r; Environmental adaptability factor: ; Sensor reliability factor; M: The number of effective sensors currently participating in the fusion; S43. Fusion Decision: Based on these weights, data fusion decisions are made to process two different types of measurement data, wind speed and wind direction, respectively. S44. Data Fusion: Wind speed fusion: Wind speed is a scalar quantity and a weighted average algorithm is used; Wind direction fusion: Wind direction is an angle and has periodicity. It is converted into a vector and then synthesized. S45. Confidence Rating and Feedback: The credibility of the fusion results is assessed, and the rating results are fed back to the weight calculation core to form a closed-loop feedback mechanism, which is used to assess the priority of data transmission or trigger system maintenance alarms; S46. After weighted fusion processing, two final results are output: Final wind speed: The wind speed measurement after wind speed fusion. ; Final wind direction: The wind direction measurement after fusion. .

7. The method for monitoring wind speed and direction of a marine buoy according to claim 6, characterized in that, The steps for S44 wind direction blending are as follows: a. Convert wind direction angle to unit vector: For wind direction angle Its unit vector in the northeast celestial coordinate system for: ； X component Represents the east direction; Y component Represents the north direction; b. Perform weighted vector synthesis: ; The sum of the eastward weighted wind direction vectors. This is the weighted northward wind direction vector sum; c. Calculate the final wind direction from the composite vector: ; The atan2 function can correctly process all quadrants, obtaining the final wind direction between 0 and 360 degrees, and the magnitude of the composite vector. It can reflect the consistency of wind direction data; the shorter the modulus, the greater the difference in wind direction between sensors.

8. A method for monitoring wind speed and direction using a marine buoy according to claim 6, characterized in that, S45 Final Confidence: Defined as the weighted sum of the weights of all participating sensors: ； in The closer the result is to 1, the higher the confidence level of the final result, indicating that the final result is determined by a sensor with high confidence.

9. A marine buoy wind speed and direction monitoring system, employing the marine buoy wind speed and direction monitoring method as described in any one of claims 1-8, characterized in that, It includes an information acquisition unit, a motion compensation unit, a self-diagnosis processing unit, a weighted fusion unit, and an adaptive decision-making unit; Information acquisition unit: Collects raw data from wind speed and direction sensors; Motion compensation unit: Synchronizes multi-source sensor data in time, unifying data from IMU, BeiDou positioning, and wind direction and speed sensors to a 50Hz time series. It fuses IMU and BeiDou positioning data using a Kalman filter, with IMU used for state prediction and BeiDou positioning for state updates, and performs error suppression during the update process. The Kalman filter outputs the buoy's attitude quaternion, used to transform the wind speed vector measured by the wind direction and speed sensors from the buoy coordinate system to the Earth coordinate system. The transformed wind speed signal then undergoes frequency domain filtering, including windowing, FFT transformation, band-stop filtering based on dynamic stopband setting based on wave energy, inverse FFT, overlap addition, and zero-phase processing, finally outputting the motion-compensated wind speed and direction. Self-diagnostic processing unit: Identifies typical fault modes through fault characteristics and quantifies the health status of each sensor, providing a basis for decision-making in subsequent data fusion; Weighted fusion unit: Combines multiple wind speed and direction data streams with the confidence scores of each sensor output by the self-diagnostic processing unit through an adaptive weighting algorithm to output a set of optimal and final wind speed and direction values; Adaptive Decision Unit: Evaluates the credibility of the fusion results and feeds the rating results back to the adaptive weighting algorithm to form a closed-loop feedback mechanism, which is used to evaluate the priority of data transmission or trigger system maintenance alarms.

10. A marine buoy wind speed and direction monitoring system according to claim 9, characterized in that, Information acquisition unit: Raw data includes ultrasonic anemometer, mechanical anemometer, and barometric pressure gradient values; acquired IMU 9-axis attitude sensor values, temperature and humidity values, and BeiDou satellite positioning data; Self-diagnostic processing unit: Data input phase: Collect raw data from various sensors, use the time synchronization vector generated by the motion compensation unit to generate an anomaly detection matrix; the input data are: ultrasonic anemometer data, mechanical anemometer data, barometric gradient sensor data, and IMU attitude data; Fault type identification: Sensor type, faults are divided into 3 categories: a. The fault type of the mechanical anemometer is mechanical jamming. Calculate the variance of the sequence samples of the mechanical anemometer data within the time window and compare it with the threshold. If the variance of the samples in three consecutive time windows is lower than the threshold, then a mechanical jamming fault can be judged. b. The pitch and roll angles calculated by the IMU sensor exceed the thresholds, which are derived from buoy design data; c. Fault diagnosis of ultrasonic, mechanical anemometers, barometric pressure gradient sensors, and temperature and humidity sensors: Calculate the forward first-order difference of each point in the data sequence and compare it with a threshold. If the threshold is exceeded for p consecutive time windows, a jump fault occurs. The threshold is set by the physical characteristics of the sensor.

Images