Automatic comparison and maintenance method for buoy observation data
By combining convolutional neural networks, Kalman filtering, genetic algorithms, and support vector machines, high-precision position adjustment and energy consumption optimization of buoys under complex sea conditions were achieved, solving the problems of position control and energy management of buoys under complex sea conditions and ensuring the stability and continuity of marine environmental monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- QINGDAO HAIYAN ELECTRONICS CO LTD
- Filing Date
- 2025-05-26
- Publication Date
- 2026-06-16
AI Technical Summary
In complex sea conditions, it is difficult to balance the accuracy of position control and energy management of ocean buoys, which leads to the inability of the system to operate stably for a long time under limited energy conditions and poor data acquisition quality.
Convolutional neural networks and Kalman filtering algorithms are used to predict ocean current trends. Genetic algorithms are combined to optimize the thrust distribution of the propulsion device, dynamically plan energy consumption balance, and adjust the sensor data acquisition frequency through support vector machines to achieve automated comparative testing and maintenance of buoys.
This improved the accuracy of buoy position control and energy management efficiency in the target area, ensuring the accuracy and continuity of marine environmental monitoring.
Smart Images
Figure CN120579047B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of buoy technology, specifically to an automated comparative testing and maintenance method for buoy observation data. Background Technology
[0002] Marine buoy systems face complex technical challenges in practical applications, mainly involving two aspects: position control and energy management. In terms of position control, ensuring accuracy and reliability under complex sea conditions is a major challenge. Regarding energy management, the system needs to balance propulsion efficiency and energy consumption. When battery capacity is insufficient, it must switch to a low-power mode and adjust the data acquisition strategy. Maintaining long-term stable operation under limited energy conditions while ensuring data acquisition quality is another key challenge. These two issues are interconnected; the accuracy of position control directly affects energy consumption, while the energy management strategy restricts position adjustment capabilities, forming a complex technical problem. Summary of the Invention
[0003] In order to solve the problems existing in the prior art, the purpose of this application is to provide an automated comparative testing and maintenance method for buoy observation data.
[0004] The automated comparative testing and maintenance method for buoy observation data described in this application includes:
[0005] Step S101: Obtain real-time buoy position data and surrounding ocean current velocity vector data; extract the current pattern of the current area from the pre-established marine environment database; and use a convolutional neural network algorithm to extract the spatiotemporal features of the ocean current velocity vector to obtain the ocean current change trend.
[0006] Step S102: Based on the trend of ocean current changes, the Kalman filter algorithm is used to fuse the real-time position data of the buoy with the ocean current prediction data, calculate the position deviation vector of the buoy in the future time period, and determine the position deviation amount.
[0007] Step S103: If the position deviation exceeds the preset threshold T1, the sensor calibration status is checked by comparison and inspection to obtain the corrected high-precision real-time position data. Based on the direction and magnitude of the deviation vector, the thrust distribution scheme of the micro-propulsion device is optimized by a genetic algorithm to obtain the thrust control parameters.
[0008] Step S104: Check the integrity of the buoy structure through comparative testing and maintenance, obtain the status data of the outer shell and fixing device, drive the micro propulsion device based on the thrust control parameters, adjust the buoy position to the target monitoring area, obtain the adjusted buoy position data, and determine whether the position deviation is lower than the preset threshold T2.
[0009] Step S105: Obtain real-time energy consumption data of the buoy propulsion device, extract the energy consumption curve of the current thrust distribution scheme from the pre-established energy consumption model, and use dynamic programming algorithm to optimize the balance scheme between thrust distribution and energy consumption to obtain the optimized energy consumption parameters.
[0010] Step S106: Based on the optimized energy consumption parameters, adjust the working mode of the micro propulsion device, obtain the adjusted energy consumption data, and determine whether the remaining battery capacity meets the operating requirements for the future period.
[0011] Step S107: If the remaining battery capacity is lower than the preset threshold T3, the energy management module switches to low power mode, acquires the working status data of the buoy sensor, and uses the support vector machine algorithm to adaptively adjust the sensor data acquisition frequency to obtain the adjusted data acquisition scheme.
[0012] Preferably, in step S101, a neural network algorithm is used to extract the spatiotemporal features of ocean currents to obtain the trend of ocean current changes, including:
[0013] The original dataset is obtained by acquiring latitude, longitude, and ocean current velocity vectors from the buoy device via a sensor interface.
[0014] A regional ocean current pattern dataset is obtained by matching gridded ocean current records from a marine environment database based on buoy location data.
[0015] A convolutional neural network algorithm was used to perform multi-layer convolution operations on ocean current velocity vectors and regional ocean current pattern datasets to obtain ocean current spatiotemporal feature sets.
[0016] The rate of change in the time dimension is calculated based on the spatiotemporal feature set of ocean currents. If the rate of change exceeds a preset threshold, it is determined to be a significant change in ocean currents.
[0017] The ocean current spatiotemporal feature set is predicted by using a long short-term memory network to output the predicted value of the ocean current velocity vector at the future time step;
[0018] A gridded interpolation method is used to spatially smooth the predicted values, generating dynamic visualization data of ocean currents.
[0019] Preferably, in step S102, the calculation of the buoy position deviation by fusing real-time buoy position data and ocean current prediction data includes:
[0020] The real-time buoy position data and ocean current prediction data are formatted and denoised to obtain a standardized dataset.
[0021] The predicted buoy position is obtained by iteratively updating the state estimate using the Kalman filter algorithm.
[0022] If the deviation between the predicted value and the real-time location data exceeds a preset threshold, adjust the Kalman filter gain and recalculate the predicted location value.
[0023] The position deviation vector is calculated based on the optimized predicted values, resulting in a set of deviation vectors;
[0024] Calculate the magnitude of the deviation vector set to obtain the position deviation value;
[0025] By analyzing the correlation between the deviation values and the ocean current prediction data, the parameters of the ocean current model are updated to obtain optimized ocean current prediction data.
[0026] Preferably, in step S103, high-precision real-time position data is obtained through sensor calibration, and thrust control parameters are generated using an optimization algorithm, including:
[0027] The sensor calibration status is checked through a comparative testing and maintenance process, the calibrated sensor data is obtained, and the sensor calibration parameters are determined.
[0028] Data fusion technology is used to process multi-source location data to obtain high-precision real-time location coordinates;
[0029] Calculate the deviation vector between the real-time position coordinates and the target position, and determine the direction angle and magnitude of the deviation vector;
[0030] A genetic algorithm is used to iteratively calculate the thrust allocation matrix with the deviation vector parameters as input, and an optimized thrust allocation scheme is obtained.
[0031] The control parameters of the micro-propulsion device are generated based on the thrust distribution matrix, and the control parameter set is determined.
[0032] By controlling the set of parameters to drive the micro-propulsion device, the real-time position is adjusted to obtain the new position deviation.
[0033] Preferably, in step S104, obtaining buoy status data through structural integrity detection and adjusting the buoy position to the target area based on thrust control parameters includes:
[0034] The integrity of the buoy shell and fixing device is scanned by the detection equipment to generate a buoy structural status dataset.
[0035] A structural integrity assessment algorithm is used to determine whether the shell and fixing device meet the operating conditions, and the assessment results are obtained.
[0036] If the evaluation results meet the operating conditions, drive commands for the micro-propulsion device are generated based on the thrust control parameters.
[0037] The buoy is moved to the target monitoring area using a miniature propulsion device, and the adjusted position data is collected.
[0038] Extract the current location coordinates from the location data and calculate the deviation value from the target area coordinates;
[0039] If the deviation value is lower than the preset threshold, a position adjustment success flag is generated.
[0040] Preferably, in step S105, acquiring propulsion device energy consumption data, extracting energy consumption curves from a pre-established energy consumption model, and generating energy consumption parameters using an optimization algorithm includes:
[0041] The real-time energy consumption dataset is obtained by collecting the status of the propulsion device and the data acquisition frequency through the sensor interface;
[0042] Extract the energy consumption curve of the current thrust allocation scheme from the pre-established energy consumption model to determine the feature point set;
[0043] The dynamic programming algorithm is used to analyze the feature point set, and combined with the operating environment and thrust control accuracy, the optimal path of the thrust allocation scheme is calculated.
[0044] If the optimized path meets the preset threshold of the energy balance strategy, output the optimized energy consumption parameters;
[0045] The energy consumption curve of the energy consumption model is updated by updating the energy consumption parameters, and the updated energy consumption model parameters are obtained.
[0046] The data acquisition frequency is adjusted based on the updated energy consumption model parameters to generate the final thrust allocation scheme.
[0047] Preferably, in step S106, adjusting the operating mode of the propulsion device according to energy consumption parameters and determining whether the remaining battery capacity meets the operating requirements includes:
[0048] The optimized energy consumption parameters are obtained from a pre-set energy consumption parameter database to determine the basis for adjustment;
[0049] A mode switching algorithm was used to adjust the working mode of the micro-propulsion device to obtain the adjusted working state.
[0050] Real-time energy consumption data is acquired through the data acquisition module to determine the current energy consumption level;
[0051] If the energy consumption data exceeds the preset threshold, the working mode is adjusted through an iterative optimization algorithm to obtain the optimized energy consumption data;
[0052] The current remaining battery capacity is obtained by using a battery capacity sensor, and a capacity prediction algorithm is used to determine whether the operating requirements are met, thus obtaining the judgment result.
[0053] Adjust the working mode based on the judgment results and determine the final energy consumption data.
[0054] Preferably, in step S107, switching to a low-power mode and using a machine learning algorithm to adjust the sensor data acquisition frequency to obtain a data acquisition scheme includes:
[0055] The battery voltage value and discharge curve parameters are collected by the power monitoring module to determine the remaining battery capacity percentage.
[0056] If the capacity is lower than the preset threshold, the energy management module will shut down non-core function circuits and reduce the operating frequency of the main control chip.
[0057] Acquire environmental and operational parameters such as temperature, humidity, pressure, and acceleration from the sensor;
[0058] The support vector machine algorithm was used to analyze the correlation between historical data collection patterns and environmental changes, and a sampling frequency mapping model was established.
[0059] The acquisition frequency of each sensor is dynamically adjusted based on the mapping relationship model;
[0060] Differential coding and Huffman compression algorithms are applied to sensor data to generate an optimized data acquisition scheme.
[0061] The automated comparative testing and maintenance method for buoy observation data described in this application has the advantage of acquiring real-time buoy position and surrounding ocean current data, extracting the spatiotemporal characteristics of ocean currents using a convolutional neural network, predicting ocean current change trends, and combining position and ocean current data with a Kalman filter algorithm to calculate the future position deviation of the buoy. When the deviation exceeds a threshold, a genetic algorithm is used to optimize the thrust distribution of the micro-propulsion device, driving the buoy back to the target monitoring area. Simultaneously, this invention also employs a dynamic programming algorithm to optimize the thrust-energy balance and adaptively adjusts the sensor operating mode according to battery capacity. This method can effectively maintain the buoy's stable operation in the target area, improve the accuracy and continuity of marine environmental monitoring, and is of great significance to marine scientific research. Attached Figure Description
[0062] Figure 1 This application describes a process for an automated comparative testing and maintenance method for buoy observation data. Figure 1 ;
[0063] Figure 2 This application describes a process for an automated comparative testing and maintenance method for buoy observation data. Figure 2 . Detailed Implementation
[0064] like Figures 1-2 As shown, the automated comparative testing and maintenance method for buoy observation data described in this application includes:
[0065] Step S101: Obtain real-time buoy position data and surrounding ocean current velocity vector data; extract the current pattern of the current area from the pre-established marine environment database; and use a convolutional neural network algorithm to extract the spatiotemporal features of the ocean current velocity vector to obtain the ocean current change trend.
[0066] Further, in step S101, latitude, longitude and ocean current velocity vector data are extracted from the buoy device through the sensor interface to obtain the raw dataset;
[0067] Based on the latitude and longitude information in the original dataset, the current ocean current patterns of the current region are extracted from a pre-established marine environment database to obtain a regional ocean current pattern dataset.
[0068] A convolutional neural network algorithm was used to extract spatiotemporal features from ocean current velocity vector data. The spatial distribution of the vector data was processed through multi-layer convolution operations to obtain a set of ocean current spatial features.
[0069] If the time series of the ocean current spatial feature set is complete, then the convolutional neural network algorithm is used to extract features from the time series of ocean current velocity vector data to obtain the ocean current time feature set;
[0070] By fusing the spatial and temporal feature sets of ocean currents, a spatiotemporal feature set of ocean currents is generated, resulting in a comprehensive feature dataset.
[0071] Based on the comprehensive feature dataset, a convolutional neural network algorithm is used to perform trend analysis on the spatiotemporal feature set of ocean currents to obtain the trend of ocean current changes.
[0072] If the time span of the ocean current change trend data meets the analysis requirements, then the ocean current change trend can be predicted in the short term using a time series prediction model to obtain the predicted trend dataset.
[0073] Based on the predicted trend dataset and the regional ocean current pattern dataset, a data fusion algorithm is used to correct the predicted trend, resulting in a corrected ocean current trend dataset.
[0074] The corrected ocean current trend dataset is spatially mapped using a visualization algorithm to generate an ocean current trend distribution map, thus obtaining the final analysis results.
[0075] Specifically, in step S101, latitude and longitude data (longitude 123.45°E, latitude 25.67°N) and surrounding ocean current velocity vector data (current velocity 0.5m / s, direction 45°) are extracted from the buoy device through the sensor interface to form a raw dataset containing timestamps;
[0076] Based on the buoy's latitude and longitude coordinates, historical ocean current pattern data (current velocity vector field with a grid resolution of 0.1°×0.1°) for the region are matched from the marine environmental database to extract the corresponding ocean current pattern dataset within the spatial range.
[0077] A 3D convolutional neural network (ConvLSTM structure, kernel size 3×3×5) was used to process the spatial distribution of ocean current velocity vectors and extract spatial features (vortex intensity, gradient change);
[0078] If the time series covers more than 24 consecutive hours, the periodicity and trend of flow velocity changes are analyzed through a temporal convolutional layer (TCN network, with a dilation factor of 2), and the time features (tidal components with a period of 12 hours) are output.
[0079] The spatial feature matrix and the temporal feature sequence are concatenated into tensors and input into a fully connected layer to generate a fused spatiotemporal feature set (dimension 256×256×10). Based on the feature set, an attention mechanism (Transformer encoder) is used to calculate the ocean current trend weights and output the change trend for the next 6 hours (current velocity increase of 0.1 m / s / h).
[0080] If the trend data covers more than 72 hours, use the ARIMA model (p=3, d=1, q=2) to predict the flow rate for the next 24 hours and generate the prediction results (mean squared error <0.05).
[0081] The predicted data and regional ocean current models are fused using Kalman filtering (noise covariance 0.01), and the optimized trend (direction correction 3°) is output after bias correction.
[0082] Finally, the vector data is mapped to the electronic nautical chart using an interpolation algorithm (Kriging) to generate a trend distribution map with arrows (color scale range 0-2 m / s).
[0083] Step S102: Based on the trend of ocean current changes, the Kalman filter algorithm is used to fuse the real-time position data of the buoy with the ocean current prediction data, calculate the position deviation vector of the buoy in the future time period, and determine the position deviation amount.
[0084] Further, in step S102, real-time buoy position data and ocean current prediction data are acquired, and the data are formatted and denoised through preprocessing to obtain a standardized dataset;
[0085] The Kalman filter algorithm is used to fuse the standardized dataset and iteratively update the state estimate to obtain the predicted buoy position.
[0086] If the deviation between the predicted value and the real-time location data exceeds a preset threshold, the Kalman filter gain is adjusted, the location prediction value is recalculated, and the optimized prediction value is determined.
[0087] Based on the optimized predicted values, the position deviation vector within the future time period is calculated to obtain the deviation vector set. The vector magnitude is then calculated using the deviation vector set to obtain the position deviation value.
[0088] By using correlation analysis between positional deviation values and ocean current prediction data, the parameters of the ocean current model are updated to obtain optimized ocean current prediction data;
[0089] Based on the optimized ocean current prediction data, the Kalman filter fusion and deviation vector calculation process is repeated to obtain the final buoy position deviation value;
[0090] By using the final buoy position deviation value, deviation correction parameters are generated, and the position correction dataset is determined.
[0091] The buoy position prediction model is updated using the position correction dataset to obtain the optimized position prediction results.
[0092] Specifically, in step S102, real-time buoy position data and ocean current prediction data are acquired, and the data are formatted and denoised through preprocessing to obtain a standardized dataset;
[0093] For example, the real-time position data of the buoy includes latitude and longitude coordinates (120.5°E, 30.2°N), and the ocean current prediction data includes current velocity (0.5m / s) and current direction (45°). The mean filtering algorithm is used to remove noise, the data is unified into a standardized format, the Kalman filtering algorithm is used to fuse the standardized dataset, and the state estimation is iteratively updated to obtain the predicted value of the buoy position.
[0094] For example, the state equation and observation equation of the Kalman filter are based on the buoy motion model and the ocean current dynamics model, respectively. The predicted value (120.6°E, 30.3°N) is obtained through iterative calculation. If the deviation between the predicted value and the real-time position data exceeds a preset threshold (0.1km), the Kalman filter gain is adjusted, the position prediction value is recalculated, and the optimized prediction value (120.55°E, 30.25°N) is determined. Based on the optimized prediction value, the position deviation vector in the future time period is calculated to obtain the deviation vector set.
[0095] For example, if the deviation vector for the next hour is (0.05km, 0.03km), the magnitude of the vector can be calculated using the set of deviation vectors to obtain the position deviation value.
[0096] For example, with a deviation vector magnitude of 0.058 km, the ocean current model parameters are updated by using correlation analysis between the position deviation value and the ocean current prediction data, resulting in optimized ocean current prediction data.
[0097] For example, by fitting the relationship between the deviation and the ocean current data using the least squares method, the current velocity is updated to 0.52 m / s and the current direction is 47°. Based on the optimized ocean current prediction data, the Kalman filter fusion and deviation vector calculation process is repeated to obtain the final buoy position deviation value.
[0098] For example, if the final deviation is 0.06 km, deviation correction parameters are generated based on the final buoy position deviation value, and the position correction dataset is determined.
[0099] For example, the correction parameters include offset angle (5°) and distance (0.06km). Using the position correction dataset, the buoy position prediction model is updated to obtain the optimized position prediction result.
[0100] For example, the optimized prediction value is (120.56°E, 30.26°N).
[0101] In step S103, if the position deviation exceeds the preset threshold T1, the sensor calibration status is checked by comparison and inspection to obtain the corrected high-precision real-time position data. Based on the direction and magnitude of the deviation vector, the thrust distribution scheme of the micro-propulsion device is optimized by a genetic algorithm to obtain the thrust control parameters.
[0102] Furthermore, in step S103, if the position deviation exceeds the preset threshold T1, the sensor calibration status is detected through the comparative testing and maintenance process, the calibrated sensor data is obtained, and the sensor calibration parameters are determined.
[0103] Based on the sensor calibration parameters, data fusion technology is used to process multi-source location data to obtain high-precision real-time location data and obtain real-time location coordinates.
[0104] If there is a deviation between the real-time position coordinates and the target position, the deviation vector V is calculated, where V includes the direction angle θ and the magnitude M, and the deviation vector parameters are determined.
[0105] Based on the deviation vector parameters, a genetic algorithm is used to optimize the thrust distribution scheme of the micro propulsion device. The genetic algorithm takes the direction angle θ and magnitude M as inputs, iteratively calculates the thrust distribution matrix P, and obtains the optimized thrust distribution scheme.
[0106] The control parameters C of the micro-propulsion device are generated by the thrust distribution matrix P, where C includes the thrust intensity and direction of each propulsion unit, and the control parameter set is determined.
[0107] Once the control parameter set C is generated, the micro-propulsion device is driven by the control parameter set C to adjust the real-time position and obtain a new position deviation.
[0108] Based on the new position deviation, repeat the comparative test and maintenance process, update the sensor calibration status, and obtain the latest calibration parameters.
[0109] The buoy structure is scanned for integrity using testing equipment to obtain shell status data and fixing device status data, and a buoy structure status dataset is generated.
[0110] If the buoy structure state dataset meets the operating conditions, then the drive command for the micro propulsion device is generated based on the control parameter set C, and the operating state of the propulsion device is determined.
[0111] Specifically, in step S103, when the position deviation exceeds the preset threshold T1 (5cm), the comparative test and maintenance process is triggered, the sensor calibration status is detected by a laser interferometer, the calibrated sensor data is obtained, the error range is corrected to ±0.1mm, and the calibration parameter matrix K is determined.
[0112] Based on the calibration parameter matrix K, the Kalman filter algorithm is used to fuse GPS and inertial navigation data. The filtering period is set to 10ms, and the high-precision real-time position coordinates (x,y,z) are output with an accuracy of ±2mm.
[0113] If there is a deviation between the real-time coordinates (105.2m, 203.7m, -1.5m) and the target coordinates (100.0m, 200.0m, -1.0m), calculate the deviation vector V (direction angle θ = 32°, magnitude M = 6.3m) and generate the polar coordinate parameter pair (θ, M);
[0114] Based on the deviation vector parameter, the genetic algorithm optimizes the thrust allocation with a population size of 50 and 100 iterations. The fitness function is to minimize energy consumption, and the output thrust allocation matrix P (thruster 1: 3.2N@120°, thruster 2: 1.8N@300°) is used.
[0115] A set of control parameters C is generated using matrix P, which contains the thrust vectors of the four thrusters ([3.2N,120°], [1.8N,300°], [0N,0°], [0N,0°]).
[0116] After the micro-thruster performs position adjustment, the laser rangefinder reports a new deviation of 1.2m, triggering the calibration process again to update parameter K to K';
[0117] Simultaneously, an ultrasonic scanner was used to detect the weld strength of the buoy shell (≥150MPa) and the deformation of the fixing device (≤0.5mm), generating a structural state dataset S;
[0118] If dataset S passes finite element analysis (safety factor ≥ 2.0), then the control parameter set C is called to generate propulsion commands, adjust the buoy to the target area (deviation 0.3m < T2 = 0.5m), and complete the positioning.
[0119] Step S104: Check the integrity of the buoy structure through comparative testing and maintenance, obtain the status data of the outer shell and fixing device, drive the micro propulsion device based on the thrust control parameters, adjust the buoy position to the target monitoring area, obtain the adjusted buoy position data, and determine whether the position deviation is lower than the preset threshold T2.
[0120] Further, in step S104, the buoy structure integrity is scanned through the comparative testing and maintenance process to obtain the shell status data and the fixing device status data, and a buoy structure status dataset is generated.
[0121] Based on the buoy structure status dataset, a preset structural integrity assessment algorithm is used to determine whether the outer shell and fixing device meet the operating conditions, and the structural integrity assessment results are obtained.
[0122] If the structural integrity assessment results meet the operating conditions, then based on the preset thrust control parameters, drive commands for the micro-propulsion device are generated to determine the operating status of the propulsion device.
[0123] The buoy is moved to the target monitoring area by performing a position adjustment operation through a micro propulsion device, and the adjusted position data is collected to generate a buoy position dataset;
[0124] A position data processing algorithm is used to extract the current position coordinates from the buoy position dataset, calculate the deviation value from the target area coordinates, and obtain the position deviation result.
[0125] If the position deviation result is lower than the preset threshold T2, a position adjustment success flag is generated to confirm that the buoy position adjustment is complete.
[0126] If the deviation value is higher than or equal to the threshold T2, return to the drive command generation step and regenerate the drive command for the micro propulsion device.
[0127] Based on the successful position adjustment indicator, collect real-time monitoring data of the buoy in the target area to generate a monitoring dataset;
[0128] By using data fusion technology to process multi-source data in the monitoring dataset, high-precision real-time monitoring data is obtained, and real-time monitoring parameters are determined.
[0129] Based on real-time monitoring parameters, a preset data analysis algorithm is used to determine whether the monitoring data meets the monitoring requirements of the target area, and the monitoring results are obtained.
[0130] Specifically, in step S104, the buoy shell and fixing device are inspected by ultrasonic scanner and stress sensor in the comparative inspection process, and the shell thickness data (5.2mm±0.1mm) and fixing device bolt torque value (45N·m±5N·m) are collected to generate a dataset containing structural parameters.
[0131] A structural integrity assessment algorithm based on finite element analysis is adopted, and a yield strength threshold (235MPa) is set. If the test data is lower than the threshold, the structure is deemed qualified.
[0132] If the structural assessment is successful, propulsion commands are generated through a PID control algorithm based on the preset thrust control parameters (maximum thrust of 2.5N per thruster). The micro-thruster adjusts the thrust direction (deflection angle of 15°) and intensity (1.8N) with a period of 0.1 seconds according to the commands, driving the buoy to move.
[0133] The real-time coordinates of the buoy (longitude 118.75°E, latitude 24.45°N) were collected using GPS and an inertial navigation system, and the Euclidean distance deviation (52m) between the buoy and the target point (118.76°E, 24.44°N) was calculated.
[0134] If the deviation is less than the threshold T2 (50m), the marker position adjustment is successful;
[0135] Otherwise, recalculate the thrust parameters. After successful adjustment, start the multispectral sensor and pH sensor to collect water quality data at a frequency of 10Hz (chlorophyll concentration 2.3μg / L, pH value 8.1).
[0136] The sensor data was fused using a Kalman filter algorithm, and after noise removal, the dissolved oxygen (6.5 mg / L) and turbidity (12 NTU) parameters were output.
[0137] The validity of the monitoring data is determined by analyzing whether the parameters are within the preset range (dissolved oxygen threshold > 5 mg / L) using a fuzzy logic algorithm.
[0138] Step S105: Obtain real-time energy consumption data of the buoy propulsion device, extract the energy consumption curve of the current thrust distribution scheme from the pre-established energy consumption model, and use dynamic programming algorithm to optimize the balance scheme between thrust distribution and energy consumption to obtain the optimized energy consumption parameters.
[0139] Further, in step S105, real-time energy consumption data and operating environment parameters of the buoy propulsion device are collected through the sensor interface to determine the real-time energy consumption dataset;
[0140] Based on the real-time energy consumption dataset, the energy consumption curve corresponding to the current thrust allocation scheme is extracted from the preset energy consumption model to obtain the feature point set of the energy consumption curve.
[0141] The dynamic programming algorithm is used to analyze the feature point set. Combined with the buoy's operating environment and thrust control accuracy, the optimal path of the thrust allocation scheme is calculated, and preliminary optimization parameters are obtained.
[0142] If the initial optimized parameters meet the preset threshold of the energy balance strategy, then the optimized energy consumption parameters are output.
[0143] If the conditions are not met, the thrust distribution scheme is adjusted, the optimized path is recalculated, and the optimized energy consumption parameters are obtained.
[0144] By updating the energy consumption curve of the preset energy consumption model using the optimized energy consumption parameters, the updated energy consumption model parameters are obtained.
[0145] Based on the updated energy consumption model parameters, the data acquisition frequency is adjusted and the calculation efficiency is optimized to generate the final thrust allocation scheme.
[0146] By adopting the final thrust distribution scheme and controlling the state of the propulsion device, the optimized energy consumption operating parameters of the buoy propulsion device are obtained;
[0147] By using a pre-set energy consumption parameter database, the optimized energy consumption parameters are obtained, and the adjustment basis for the micro propulsion device is determined.
[0148] Based on the adjustment criteria of the micro-propulsion device, a mode switching algorithm is used to adjust the working mode of the micro-propulsion device and obtain the adjusted working state.
[0149] Specifically, in step S105, the current and voltage data of the buoy propulsion device, as well as the ambient water temperature and flow velocity parameters, are collected through the sensor interface at a sampling frequency of 10Hz. After noise reduction by Kalman filtering, a real-time energy consumption dataset containing timestamps is generated.
[0150] Based on this dataset, the energy consumption curve corresponding to the current thrust allocation coefficient of 0.85 is called from the preset LSTM energy consumption model, and 12 feature points on the curve with thrust step change exceeding 5% are extracted to form a feature point set.
[0151] The Bellman-Ford dynamic programming algorithm is used with a thrust control accuracy of 0.1N as a constraint. The Pareto front of the energy consumption increment and thrust gain is calculated in the feature point set, and the preliminary optimization parameters with an 8% reduction in energy consumption are output.
[0152] If the parameter meets the condition of energy consumption fluctuation threshold ±2W, then the optimized parameter is stored.
[0153] Otherwise, adjust the thrust distribution coefficient in increments of 0.05 and recalculate iteratively until the target is met;
[0154] The final optimized parameters are input into the energy consumption model training module, and the LSTM network weights are updated using the gradient descent method to generate model parameters v2.1 containing the new energy consumption curve. Based on the confidence interval of model parameters v2.1, the data acquisition frequency is dynamically adjusted to 5-15Hz. At the same time, CUDA parallel computing is enabled to accelerate the optimization process and generate the final allocation scheme containing 3 sets of thrust vectors.
[0155] The allocation scheme is written to the thruster controller via the CAN bus, and the PWM duty cycle is adjusted to 62%, so that the system enters the optimized energy consumption operation state.
[0156] Read the 10 most recent optimization parameters from the Redis cache database and calculate the moving average as the threshold for switching the micro-booster mode;
[0157] When the roll rate exceeds 0.5 rad / s, switch to high-efficiency mode and increase the propeller speed to 1200 rpm to complete the adjustment of the working status.
[0158] Step S106: Based on the optimized energy consumption parameters, adjust the working mode of the micro propulsion device, obtain the adjusted energy consumption data, and determine whether the remaining battery capacity meets the operating requirements for the future period.
[0159] Furthermore, in step S106, optimized energy consumption parameters are obtained through a preset energy consumption parameter database to determine the adjustment basis for the micro propulsion device;
[0160] Based on the obtained optimized energy consumption parameters, the working mode of the micro propulsion device is adjusted using a mode switching algorithm to obtain the adjusted working state;
[0161] The data acquisition module obtains real-time energy consumption data from the adjusted working status to determine the current energy consumption level.
[0162] If the energy consumption data exceeds the preset energy consumption threshold, the working mode is adjusted through an iterative optimization algorithm to obtain optimized energy consumption data.
[0163] The battery capacity sensor is used to obtain the current remaining battery capacity and determine the battery's available energy.
[0164] Based on the optimized energy consumption data and operational requirements, a capacity prediction algorithm is used to determine whether the remaining battery capacity meets the subsequent operational requirements, and the judgment result is obtained.
[0165] If the judgment result indicates that the battery capacity is insufficient, the working mode of the micro-propulsion device is adjusted through dynamic programming algorithm to obtain the final energy consumption data.
[0166] Based on the final energy consumption data, update the energy consumption curve of the preset energy consumption model to obtain the updated energy consumption model parameters;
[0167] By updating the energy consumption model parameters, adjusting the data acquisition frequency and optimizing the computational efficiency, the final operating parameters of the micro propulsion device are generated.
[0168] Specifically, in step S106, the energy consumption parameters of the micro propulsion device in low thrust mode are extracted from the preset energy consumption parameter database as 0.5W and in high thrust mode as 2.3W, and the adjustment basis is to prioritize the use of low thrust mode.
[0169] Based on the energy consumption parameters of the low thrust mode, a mode switching algorithm based on PID control is adopted to switch the working mode of the propulsion device from high thrust to low thrust, and the adjusted thrust value is 0.8N.
[0170] Real-time energy consumption data was collected at a frequency of 10Hz using a high-precision current sensor, and the current average power consumption was calculated to be 0.6W, which is lower than the preset threshold of 1.0W.
[0171] If the instantaneous power consumption exceeds 1.2W, an iterative optimization algorithm based on gradient descent is triggered to adjust the PWM duty cycle to 65% so that the optimized power consumption is stabilized at 0.7W.
[0172] The remaining battery capacity was determined to be 1200mAh using a coulomb counter sensor, and the usable energy was calculated to be 4.32Wh based on the discharge curve.
[0173] The ARIMA prediction algorithm was used to analyze the energy consumption demand for the next 2 hours, which was 3.6Wh, indicating that the current battery capacity meets the demand.
[0174] If the predicted capacity is insufficient, a dynamic programming algorithm is invoked to reallocate the thrust sequence and generate an operating scheme with a final power consumption of 0.55W.
[0175] The optimized 0.55W power consumption value was updated to the energy consumption model database, and the slope parameter k of the energy consumption curve was corrected to 0.12.
[0176] Based on the updated model parameters, the data acquisition frequency was reduced to 5Hz and sparse matrix operations were enabled, resulting in final operating parameters where the output thrust remained at 0.8N and the duty cycle at 60%.
[0177] Step S107: If the remaining battery capacity is lower than the preset threshold T3, the energy management module switches to low power mode, acquires the working status data of the buoy sensor, and uses the support vector machine algorithm to adaptively adjust the sensor data acquisition frequency to obtain the adjusted data acquisition scheme.
[0178] Furthermore, in step S107, the battery voltage value and discharge curve parameters are collected in real time by the battery capacity sensor to determine the current remaining battery capacity percentage;
[0179] If the remaining battery capacity is lower than the preset threshold T3, the energy management module will switch to low power mode, shut down non-core function circuits and reduce the operating frequency of the main control chip.
[0180] Acquire current operating status data of the buoy sensor, including environmental parameters such as sensor temperature, humidity, pressure, and acceleration, as well as internal operating parameters of the sensor;
[0181] The support vector machine algorithm was used to analyze the correlation between historical data collection patterns and environmental changes, and a mapping relationship model between environmental parameters and the optimal sampling frequency was established.
[0182] The sampling frequency of each sensor is dynamically adjusted based on the output of the support vector machine, increasing the sampling frequency in areas with drastic environmental changes and decreasing the sampling frequency in stable areas.
[0183] The data acquisition module performs data compression processing, applying differential coding and Huffman compression algorithms to reduce data storage space and transmission energy consumption.
[0184] An optimized data acquisition scheme is generated based on the adjusted sampling frequency and compression processing results, including new sampling time intervals, sampling accuracy, data processing flow and storage strategy for each sensor.
[0185] Optimized energy consumption parameters are obtained from a pre-set energy consumption parameter database to determine the adjustment basis for the micro-propulsion device;
[0186] A mode switching algorithm is used to adjust the working mode of the micro propulsion device to obtain the adjusted working state.
[0187] Specifically, in step S107, the battery capacity sensor collects the voltage value (3.7V) and the discharge curve slope parameter (-0.02V / min) at a sampling rate of 100Hz, and calculates the remaining capacity percentage (65%) by integrating with a coulomb counter.
[0188] If the detected capacity is below the threshold T3 (30%), the energy management module shuts down the GPS module and the wireless communication circuit, and reduces the main control chip frequency from 48MHz to 8MHz.
[0189] The buoy sensor uploads current temperature (25.3℃), humidity (78%RH), pressure (101.2kPa) and triaxial acceleration data (X: 0.02g, Y: 0.05g, Z: 1.02g), and simultaneously records the sensor's internal ADC sampling rate (10bit) and signal-to-noise ratio (62dB).
[0190] The support vector machine is loaded with a historical dataset of the past 24 hours (containing 500 sets of environmental parameters paired with sampling frequencies), and the classification model is trained using the RBF kernel function. The weight coefficients of each environmental parameter are output (temperature 0.35, humidity 0.15, acceleration 0.5).
[0191] Based on the model prediction results, the sampling frequency in the region with an acceleration change rate > 0.1g / s was increased to 5Hz, and the frequency in the stable region was reduced to 0.5Hz;
[0192] The data acquisition module performs differential encoding on 10 consecutive sets of temperature data (25.1℃, 25.2℃, ... 25.3℃), generates an 8-bit compressed bitstream, and then applies Huffman encoding (compression ratio 1:0.7).
[0193] The optimization scheme adjusts the temperature sampling interval from 1 second to 2 seconds, reduces the pressure sampling accuracy from 12 bits to 10 bits, and changes the storage strategy to packet transmission every 30 minutes.
[0194] The energy consumption parameter database is used to query the optimal PWM duty cycle (45%) of the thruster corresponding to the current water temperature (20℃), and the thrust parameters are generated by combining the motor efficiency curve (85%@45% load).
[0195] The mode switching algorithm predicts the energy consumption (0.8Wh) in the next 5 minutes based on Kalman filtering, and switches the thruster operating mode from continuous operation to pulse mode (turn on at 200ms / turn off at 800ms).
[0196] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this application.
Claims
1. An automated comparative testing and maintenance method for buoy observation data, characterized in that, include: Acquire real-time buoy position data and surrounding ocean current velocity vector data, extract ocean current spatiotemporal characteristics, and predict ocean current change trends; The Kalman filter algorithm is used to fuse real-time buoy position and ocean current prediction data to calculate the position deviation in the future time period. If the position deviation exceeds the threshold T1, perform sensor calibration and buoy structural integrity detection, and optimize the thrust distribution scheme of the micro propulsion device through a genetic algorithm to generate thrust control parameters. Adjust the buoy position to the target monitoring area based on the thrust control parameters, and determine whether the position deviation after adjustment is lower than the threshold T2; Collect real-time energy consumption data of the propulsion device, combine dynamic programming algorithm to optimize thrust distribution and energy consumption balance, and generate optimized energy consumption parameters; Adjust the operating mode of the propulsion device according to energy consumption parameters, and determine whether the remaining battery capacity meets future operating requirements; If the remaining battery capacity is lower than the threshold T3, switch to low power mode and use the support vector machine algorithm to adaptively adjust the sensor data acquisition frequency to generate an optimized data acquisition scheme.
2. The automated comparative testing and maintenance method for buoy observation data according to claim 1, characterized in that, The predicted ocean current trend includes: A 3D convolutional neural network ConvLSTM structure is adopted, with a convolutional kernel size of 3×3×5, to process the spatial distribution of ocean current velocity vectors and extract vortex intensity and gradient change features. The flow velocity periodicity and trend are analyzed and spatiotemporal characteristics are output by using a temporal convolutional layer TCN network. The expansion factor of the TCN network is set to 2. After fusing the aforementioned spatiotemporal features, the attention mechanism Transformer encoder is used to calculate the ocean current trend weights for the next 6 hours, and the current velocity changes for the next 24 hours are predicted based on the ARIMA model. The prediction error threshold of the ARIMA model is ≤0.05m / s.
3. The automated comparative testing and maintenance method for buoy observation data according to claim 1, characterized in that, The state equation of the Kalman filter algorithm is based on the buoy motion model and the ocean current dynamics model. It iteratively updates the position prediction value. If the deviation between the predicted value and the real-time position exceeds 0.1km, the Kalman gain is adjusted, the optimized prediction value is recalculated, and the ocean current model parameters are updated by the least squares method.
4. The automated comparative testing and maintenance method for buoy observation data according to claim 1, characterized in that, The optimization process of the genetic algorithm includes: Using the deviation vector direction angle θ and magnitude M as inputs, the population size is set to 50, the number of iterations is set to 100, and the fitness function is to minimize energy consumption; The output thrust distribution matrix P contains a thrust intensity accuracy of 0.1N and a directional accuracy of 1° for each thruster, and the adjusted deviation is verified to be lower than T2 by a laser rangefinder.
5. The automated comparative testing and maintenance method for buoy observation data according to claim 1, characterized in that, The dynamic programming algorithm adopts the Bellman-Ford algorithm, with a thrust control accuracy of 0.1N as a constraint, to calculate the Pareto front optimization path and output preliminary optimization parameters that reduce energy consumption by 8%. If the initial optimized parameters meet the condition of energy consumption fluctuation threshold ±2W, then adjust the thrust distribution coefficient, and adjust the thrust distribution coefficient step size by 0.
05.
6. The automated comparative testing and maintenance method for buoy observation data according to claim 1, characterized in that, The determination of the remaining battery capacity includes: Available energy is obtained by using a coulomb counter sensor, and energy consumption demand for the next 2 hours is predicted by combining it with the ARIMA algorithm. If the predicted energy consumption exceeds the remaining capacity, a dynamic programming algorithm is invoked to reallocate the thrust sequence and generate an operating scheme with a power consumption reduction of ≥10%.
7. The automated comparative testing and maintenance method for buoy observation data according to claim 1, characterized in that, The specific operation of the support vector machine algorithm is as follows: The historical dataset, which includes environmental parameters paired with sampling frequencies, is loaded, and a classification model is trained using the RBF kernel function. Output the weighting coefficients for each environmental parameter: temperature 0.35, humidity 0.15, acceleration 0.
5. Dynamically adjust the sampling frequency: increase to 5Hz when the rate of change of acceleration is >0.1g / s, and decrease to 0.5Hz in the stable region.
8. The automated comparative testing and maintenance method for buoy observation data according to claim 1, characterized in that, The low-power mode includes: Turn off the GPS module and wireless communication circuit, and reduce the frequency of the main control chip from 48MHz to 8MHz; The sensor data is compressed using differential coding and Huffman compression algorithm with a compression ratio of ≥1:0.7, and the storage strategy is adjusted to be transmitted in packets every 30 minutes.
9. The automated comparative testing and maintenance method for buoy observation data according to claim 4, characterized in that, The verification of the thrust allocation matrix P includes: The strength of the buoy's outer shell weld was tested using an ultrasonic scanner; the strength was ≥150MPa and the deformation of the fixing device was ≤0.5mm. If the safety factor of the finite element analysis is ≥2.0, then the drive thruster will perform position adjustment, and the deviation threshold T2 will be set to 0.5m.
Citation Information
Patent Citations
Integrated acoustic positioning and tracking control method for autonomous underwater vehicle
CN112947068A
Acoustic survey technique
US5995445A