A system and method for error correction of oceanographic environmental observation data

Abstract

The application discloses a kind of error correction system and method for marine environment observation data, comprising: including data acquisition, compensation modeling, error analysis, data compensation, data fusion and feedback optimization module.Data acquisition module obtains ADCP flow measurement, GPS, IMU, temperature, pressure and water quality monitoring data.Error analysis module utilizes Kalman filter and hidden Markov model to detect drift error, and based on long short-term memory network analysis environmental error.Data compensation module optimizes compensation model, improves data precision.Data fusion module uses particle filter and bayesian optimization algorithm, and fuses multi-sensor data.Feedback optimization module dynamically adjusts compensation model weight according to error correction data, improves system adaptability.The application improves the accuracy of ocean current monitoring data, and provides more reliable data support for underwater environment research.

Description

Technical Field

[0001] This invention relates to the field of marine water quality monitoring technology, and in particular to an error correction system and method for marine environmental observation data. Background Technology

[0002] Marine environmental monitoring plays a vital role in marine resource development, marine ecological protection, and navigation safety. Accurate measurement of ocean current data is crucial for applications such as marine environmental research, maritime navigation, search and rescue operations, and pollution monitoring. Currently, common ocean current monitoring equipment includes buoy monitoring stations, unmanned underwater vehicles, and underwater sensor arrays. These devices typically carry acoustic Doppler current profilers (ADCP), global positioning systems (GPS), inertial measurement units (IMUs), and water quality sensors.

[0003] However, existing ocean current monitoring systems suffer from measurement errors due to the complexity of the marine environment and sensor drift during long-term operation. For example, ADCP (Advanced Digital Current Profiler) may experience drift errors due to water turbulence, attitude changes, and instrument errors; cumulative errors in GPS and IMU (Instrument Measure) can also reduce positioning accuracy; furthermore, changes in environmental factors such as temperature and pressure can have unstable effects on water quality sensor data. Currently, commonly used error correction methods mainly rely on a single data source or simple empirical models, which cannot fully utilize multi-source data for error compensation, resulting in limited correction effectiveness.

[0004] Therefore, there is an urgent need for a system that can comprehensively utilize multi-source data and combine advanced algorithms to effectively correct drift errors and environmental errors, so as to improve the accuracy and reliability of ocean current monitoring data. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide an error correction system and method for marine environmental observation data, thereby improving the accuracy and reliability of ocean current monitoring data.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an error correction system for marine environmental observation data, applicable to water quality sensing devices on buoy monitoring stations, unmanned underwater vehicles, and underwater sensor arrays, comprising:

[0007] The data acquisition module is used to collect ADCP flow measurement data, GPS data, IMU data, temperature data, and pressure data from the water quality sensing device in real time, as well as water quality monitoring data from the water quality sensing device.

[0008] The compensation modeling module is used to acquire multiple historical training data of the water quality sensing device and construct an initial compensation model based on each of the historical training data.

[0009] The error analysis module, connected to the data acquisition module, is used to perform anomaly detection on the ADCP flow measurement data, the GPS data, and the IMU data according to the improved Kalman filter algorithm and the hidden Markov model to obtain drift error, and to perform anomaly detection on the temperature data and the pressure data according to the long short-term memory network to obtain environmental error.

[0010] The data compensation module, connected to the error analysis module, the compensation modeling module, and the data acquisition module, is used to perform error correction on the initial compensation model based on the drift error and the environmental error to obtain an optimized compensation model. The optimized compensation model is used to perform error compensation on water quality monitoring data to obtain compensated monitoring data.

[0011] The data fusion module, connected to the data compensation module, is used to fuse and optimize the compensation monitoring data of multiple sensing devices according to the particle filtering algorithm and the Bayesian optimization algorithm to obtain error correction data.

[0012] The feedback optimization module, connected to the data fusion module and the data compensation module, is used to process the compensation monitoring data and the error correction data to obtain deviation feedback data, and dynamically adjust the weight parameters of the optimization compensation model based on the deviation feedback data to obtain the adjusted optimization compensation model.

[0013] Furthermore, if the historical training data includes historical error data, historical real water quality data, and historical sensor monitoring data, then the compensation modeling module includes:

[0014] The preprocessing unit is used to preprocess the historical error data, the historical real water quality data, and the historical sensor monitoring data to obtain multiple preprocessed historical error data, multiple historical real water quality data, and multiple historical sensor monitoring data.

[0015] The introductory unit is used to introduce a CNN-LSTM hybrid model as the original model.

[0016] The training unit, connecting the input unit and the preprocessing unit, is used to take the preprocessed historical error data and historical sensor monitoring data as inputs, and the corresponding historical real water quality data as outputs, to retrain the original model and obtain the initial compensation model.

[0017] Furthermore, the error analysis module includes a drift error detection unit, which includes:

[0018] The preliminary detection subunit is used to make a preliminary estimate of the drift error based on the improved Kalman filter algorithm to obtain the preliminary error, and then analyze the time series characteristics of the preliminary error based on the hidden Markov model to predict the trend of the preliminary error.

[0019] The burst detection subunit is used to detect burst errors based on the double sliding window method and the isolated forest algorithm.

[0020] An enhanced detection subunit is used to acquire historical data of the ADCP current measurement data and the IMU data, and input them into a preset bidirectional long short-term memory network to obtain the prediction error;

[0021] The fusion detection subunit is connected to the preliminary detection subunit, the burst detection subunit, and the enhancement detection subunit, respectively, and is used to obtain the drift error by fusing the preliminary error trend, the burst error, and the prediction error.

[0022] Furthermore, the error analysis module includes an environmental error detection unit, which includes:

[0023] The initial error screening unit is used to calculate the mean, variance, skewness, and kurtosis of the temperature and pressure data, and to remove outlier data based on the 3σ principle to obtain the screened data.

[0024] A modeling and prediction subunit is connected to the error screening subunit. It is used to acquire historical temperature data, historical pressure data, and historical environmental errors. Based on the historical temperature data, historical pressure data, and historical environmental errors, a one-way long short-term memory network is constructed. The screened data is input into the one-way long short-term memory network to predict the environmental error.

[0025] The model adjustment subunit, connected to the modeling and prediction subunit, is used to calculate the credibility of the environmental error based on Bayesian inference and dynamically adjust the weights of the unidirectional long short-term memory network based on the credibility.

[0026] Furthermore, the data compensation module includes:

[0027] An error feature extraction unit is used to time-align the drift error, the environmental error, and the water quality monitoring data, and to extract error features from the drift error and the environmental error based on principal component analysis and empirical mode decomposition methods.

[0028] An incremental learning unit, connected to the error feature extraction unit, is used to enable the initial compensation model to learn the extracted error features online and perform error correction based on the incremental learning method, thereby obtaining a correction compensation model.

[0029] The dynamic optimization unit, connected to the incremental learning unit, is used to dynamically adjust the hyperparameters of the correction compensation model according to the Bayesian optimization algorithm to obtain the optimized compensation model.

[0030] Furthermore, the data fusion module includes:

[0031] A multimodal feature extraction unit is used to perform time alignment on the compensation monitoring data of multiple sensing devices, and to extract sensing features from each of the compensation monitoring data according to the principal component analysis method and the deep learning feature extraction algorithm.

[0032] The weight adjustment unit, connected to the multimodal feature extraction unit, is used to calculate the feature importance of each of the sensing features based on information entropy, and dynamically adjust the weight of the corresponding compensation monitoring data of the sensing features based on the feature importance and the Bayesian optimization algorithm.

[0033] The weighted fusion unit, connected to the weight adjustment unit, is used to perform weighted fusion based on the adjusted weights and the corresponding compensation monitoring data to obtain preliminary corrected data;

[0034] The timing optimization unit, connected to the weighted fusion unit, is used to perform timing optimization on the preliminary correction data according to the particle filter algorithm to obtain error correction data.

[0035] Furthermore, the feedback optimization module includes:

[0036] The deviation calculation unit is used to calculate the deviation feedback data by processing the compensation monitoring data and the error correction data based on the Euclidean distance.

[0037] The compensation processing unit is connected to the deviation calculation unit and is used to process the deviation feedback data to obtain the compensation accuracy improvement amount.

[0038] The model optimization unit, connected to the compensation processing unit, is used to dynamically adjust the weight parameters of the optimized compensation model by adjusting the amount of compensation accuracy improvement, so as to obtain the adjusted optimized compensation model.

[0039] Furthermore, the calculation formula for the deviation feedback data is configured as follows:

[0040] ;

[0041] in, Used to represent the deviation feedback data. Used to represent the compensation monitoring data Used to represent the error correction data. This is used to represent the total number of the deviation feedback data.

[0042] An error correction method for marine environmental observation data, applied to the aforementioned error correction system for marine environmental observation data, includes:

[0043] Step S1: The data acquisition module collects ADCP flow measurement data, GPS data, IMU data, temperature data, and pressure data from the water quality sensing device in real time, as well as water quality monitoring data from the water quality sensing device.

[0044] Step S2: The compensation modeling module acquires multiple historical training data of the water quality sensing device and constructs an initial compensation model based on each of the historical training data.

[0045] Step S3: The error analysis module performs anomaly detection on the ADCP flow measurement data, the GPS data, and the IMU data based on the improved Kalman filter algorithm and the hidden Markov model to obtain drift error, and performs anomaly detection on the temperature data and the pressure data based on the long short-term memory network to obtain environmental error.

[0046] Step S4: The data compensation module corrects the initial compensation model based on the drift error and the environmental error to obtain an optimized compensation model. The optimized compensation model is used to compensate for the error in the water quality monitoring data to obtain compensated monitoring data.

[0047] Step S5: The feedback optimization module processes the compensation monitoring data and the error correction data to obtain deviation feedback data, and dynamically adjusts the weight parameters of the optimization compensation model based on the deviation feedback data to obtain the adjusted optimization compensation model.

[0048] The beneficial effects of this invention are:

[0049] This invention acquires ADCP flow measurement data, GPS data, IMU data, temperature data, pressure data, and water quality monitoring data, and comprehensively utilizes information from multiple sensors to effectively reduce errors that may be caused by a single data source and improve the integrity and reliability of the data.

[0050] This invention also utilizes Kalman filtering and Hidden Markov Model to detect drift errors, and combines Long Short-Term Memory Networks to analyze environmental errors, which can accurately identify data anomalies caused by sensor drift, environmental changes, etc., and improve the accuracy of error detection.

[0051] This invention dynamically optimizes the compensation model based on drift error and environmental error to ensure the real-time performance and effectiveness of error correction, making ocean current monitoring data more accurate and stable. Then, the compensation model is dynamically adjusted based on the error correction data to achieve adaptive optimization of the compensation model, thereby improving the applicability and long-term stability of the system under different environments.

[0052] In addition, this invention employs particle filtering algorithm and Bayesian optimization algorithm to fuse and optimize data from multiple sensing devices, reducing measurement deviations between sensors and improving the reliability and accuracy of the final error correction data;

[0053] In summary, this invention effectively improves the accuracy and reliability of ocean current monitoring data, providing higher quality data support for marine environmental research, navigation safety, pollution monitoring, and underwater operations. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the structure of the ocean current drift error correction system in this invention;

[0055] Figure 2 This is a schematic diagram of the compensation modeling module in this invention;

[0056] Figure 3 This is a schematic diagram of the error analysis module in this invention;

[0057] Figure 4 This is a schematic diagram of the data compensation module in this invention;

[0058] Figure 5 This is a schematic diagram of the data fusion module in this invention;

[0059] Figure 6 This is a schematic diagram of the feedback optimization module in this invention;

[0060] Figure 7 This is a flowchart of the steps in the ocean current drift error correction method of the present invention.

[0061] Figure labeling: 1. Data acquisition module; 2. Compensation modeling module; 21. Preprocessing unit; 22. Input unit; 23. Training unit; 3. Error analysis module; 31. Drift error detection unit; 311. Preliminary detection subunit; 312. Burst detection subunit; 313. Enhanced detection subunit; 314. Fusion detection subunit; 32. Environmental error detection unit; 321. Initial error screening subunit; 322. Modeling and prediction subunit; 323. Model adjustment subunit; 4. Data compensation module; 41. Error feature extraction unit; 42. Incremental learning unit; 43. Dynamic optimization unit; 5. Data fusion module; 51. Multimodal feature extraction unit; 52. Weight adjustment unit; 53. Weighted fusion unit; 54. Temporal optimization unit; 6. Feedback optimization module; 61. Deviation calculation unit; 62. Compensation processing unit; 63. Model optimization unit. Detailed Implementation

[0062] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom surface," "top surface," "inner," and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0063] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides an error correction system for marine environmental observation data. This system effectively improves the accuracy and reliability of ocean current monitoring data and is applicable to water quality sensing devices on buoy monitoring stations, unmanned underwater vehicles, and underwater sensor arrays. The system includes:

[0064] Data acquisition module 1 is used to collect ADCP flow measurement data, GPS data, IMU data, temperature data, pressure data, and water quality monitoring data from the water quality sensing device in real time.

[0065] Compensation modeling module 2 is used to acquire multiple historical training data from water quality sensing devices and build an initial compensation model based on each historical training data.

[0066] Error analysis module 3 is connected to data acquisition module 1. It is used to detect anomalies in ADCP flow measurement data, GPS data, and IMU data based on the improved Kalman filter algorithm and hidden Markov model to obtain drift error, and to detect anomalies in temperature data and pressure data based on long short-term memory network to obtain environmental error.

[0067] Data compensation module 4 connects error analysis module 3, compensation modeling module 2 and data acquisition module 1. It is used to correct the initial compensation model based on drift error and environmental error to obtain an optimized compensation model. The optimized compensation model is used to compensate for errors in water quality monitoring data to obtain compensated monitoring data.

[0068] The data fusion module 5 is connected to the data compensation module 4 and is used to fuse and optimize the compensation monitoring data of multiple sensing devices according to the particle filtering algorithm and the Bayesian optimization algorithm to obtain error correction data.

[0069] The feedback optimization module 6 connects the data fusion module 5 and the data compensation module 4. It is used to process the compensation monitoring data and error correction data to obtain deviation feedback data, and dynamically adjust the weight parameters of the optimization compensation model based on the deviation feedback data to obtain the adjusted optimization compensation model.

[0070] Working principle of Example 1:

[0071] Data acquisition module 1, deployed on buoy monitoring stations, unmanned underwater vehicles, and underwater sensor arrays, is responsible for real-time acquisition of ADCP current measurement data, GPS data, IMU data, temperature data, pressure data, and water quality monitoring data. ADCP is used to measure ocean current velocity and direction, GPS provides position reference, IMU monitors changes in device attitude, and temperature and pressure sensors record environmental parameters.

[0072] Compensation modeling module 2 trains an initial compensation model based on long-term historical data. In a laboratory environment, using a large amount of historical training data, a compensation model based on a deep neural network is employed to give it preliminary error correction capabilities.

[0073] Error analysis module 3 employs Kalman filtering and Hidden Markov Models (HMMs) to detect errors in ADCP, GPS, and IMU data. GPS signals may be affected by waves or signal obstruction, causing temporary drift. This embodiment uses an HMM to identify abnormal states and combines it with KF smoothing. Furthermore, in this embodiment, a Long Short-Term Memory (LSTM) network is used to analyze temperature and pressure data, identifying the impact of environmental factors on sensor readings, such as the effect of deep-sea high pressure on ADCP flow measurement accuracy.

[0074] Based on the error analysis results, the data compensation module 4 dynamically optimizes the initial compensation model. For example, to address ADCP drift error, the system uses a Bayesian optimization method to adjust the compensation model parameters, making the compensation results more accurate.

[0075] The data fusion module 5 employs particle filtering and Bayesian optimization algorithms to fuse and optimize data from multiple sensors. For example, underwater sensors at different depths may measure slightly inaccurate flow velocity data. The system uses a particle filtering algorithm to perform weighted averaging, making the final error-corrected data more stable.

[0076] The feedback optimization module 6 uses the error correction data after data fusion to dynamically adjust the weight parameters of the optimization compensation model. For example, if the system detects a drift trend in IMU data, it adaptively reduces the impact of IMU data on the final calculation result and increases the weight of GPS and ADCP data.

[0077] In actual testing, this embodiment was deployed on a buoy monitoring station and an AUV in a certain sea area. By comparing the ocean current data corrected by this system with high-precision reference measurement data (such as shipborne high-precision ADCP measurement data), it was found that this system can effectively reduce drift error, improve data accuracy by about 30%, and reduce current measurement error to within ±0.05 m / s, which is a significant improvement compared with the traditional single-sensor error correction method (error of about ±0.15 m / s).

[0078] This embodiment can be widely applied to fields such as marine environmental monitoring, marine scientific research, maritime traffic and navigation safety, underwater archaeology, and marine resource development, providing reliable technical support for accurate ocean current data.

[0079] In summary, this embodiment proposes an ocean current drift error correction system based on multi-source data fusion, which can significantly improve the accuracy and reliability of ocean current monitoring data and has better adaptability and real-time performance compared with traditional error compensation methods.

[0080] Preferably, historical training data includes historical error data, historical real water quality data, and historical sensor monitoring data, for reference. Figure 2 The compensation modeling module 2 includes:

[0081] Preprocessing unit 21 is used to preprocess the historical error data, historical real water quality data, and historical sensor monitoring data to obtain multiple historical error data, multiple historical real water quality data, and multiple historical sensor monitoring data after preprocessing.

[0082] Unit 22 is used to introduce a CNN-LSTM hybrid model as the original model;

[0083] Training unit 23, which connects input unit 22 and preprocessing unit 21, is used to take the preprocessed historical error data and historical sensor monitoring data as inputs and the corresponding historical real water quality data as outputs to retrain the original model and obtain the initial compensation model.

[0084] Specifically, in this embodiment, the preprocessing unit 21 performs data cleaning and format conversion on historical error data, historical real water quality data, and historical sensor monitoring data to improve the stability and accuracy of model training. The main processing flow of the preprocessing unit 21 is as follows:

[0085] Data denoising: The Savitzky-Golay filtering method is used to smooth the original data and remove high-frequency noise;

[0086] Data standardization: All input data (historical error data, historical sensor monitoring data) are standardized to make the data conform to the input requirements of the CNN-LSTM model;

[0087] Data alignment: Historical error data, historical real water quality data, and historical sensor monitoring data are synchronized based on timestamps to ensure the temporal consistency of data input;

[0088] Data splitting: The data is divided into training and test sets according to the ratio of 80% training and 20% testing.

[0089] The CNN-LSTM hybrid model consists of a CNN part and an LSTM part. The CNN part extracts features from historical sensor monitoring data and historical error data to identify the spatial correlation of the data. The LSTM part processes time-series data, captures the patterns of historical error data changes over time, and improves the model's predictive ability for time-series data.

[0090] The combination of CNN-LSTM hybrid models is as follows: First, CNN is used to extract local features of the data (e.g., the spatial relationship between historical error data and historical sensor monitoring data); then, the features output by CNN are passed to LSTM for time series modeling to learn the error variation pattern; finally, the fully connected layers are combined to output the final error compensation result.

[0091] The training process for training unit 23 of the model includes:

[0092] Model initialization: Load the CNN-LSTM structure and set parameters such as the number of network layers, convolutional kernel size, and number of LSTM units.

[0093] Forward propagation: Input historical error data and historical sensor monitoring data, extract features through CNN and then feed them into LSTM, finally outputting predicted water quality data through a fully connected layer.

[0094] Loss calculation: The mean squared error (MSE) is used as the loss function to calculate the error between the model output value and the historical real water quality data.

[0095] Backpropagation and optimization: The Adam optimizer is used to update the weights of the CNN and LSTM to minimize the loss function.

[0096] Model training: Iterative training until convergence is achieved, and an initial compensation model is obtained, which can be used for error compensation.

[0097] The initial compensation model was finally tested at a buoy monitoring station in a certain sea area:

[0098] Six months of historical error data, historical sensor monitoring data, and historical real water quality data were collected, totaling 10,000 sets of data for training and testing.

[0099] Model performance: MSE error decreased by 40%, demonstrating that CNN-LSTM has better temporal modeling capabilities compared to traditional BP neural network models; water quality monitoring data error decreased by 25%, effectively improving the accuracy of water quality data; computational efficiency increased by 30%, with CNN dimensionality reduction and feature extraction improving the efficiency of LSTM in processing time-series data. This example demonstrates that the initial compensation model trained based on CNN-LSTM can effectively reduce errors and improve the reliability of marine water quality monitoring.

[0100] Preferred, Reference Figure 3 The error analysis module 3 includes a drift error detection unit 31, which includes:

[0101] The preliminary detection subunit 311 is used to make a preliminary estimate of the drift error based on the improved Kalman filter algorithm to obtain the preliminary error, and then analyze the time series characteristics of the preliminary error based on the hidden Markov model to predict the trend of the preliminary error.

[0102] The burst detection subunit 312 is used to detect burst errors based on the double sliding window method and the isolated forest algorithm.

[0103] The enhanced detection subunit 313 is used to acquire historical data of ADCP current measurement data and IMU data, and input them into a preset bidirectional long short-term memory network to obtain the prediction error;

[0104] The fusion detection subunit 314 is connected to the preliminary detection subunit 311, the burst detection subunit 312 and the enhancement detection subunit 313 respectively, and is used to obtain the drift error by fusing the preliminary error trend, the burst error and the prediction error.

[0105] Specifically, in this embodiment, improved Kalman filter (EKF), hidden Markov model (HMM), double sliding window method, isolation forest algorithm, and bidirectional long short-term memory network (Bi-LSTM) are used to detect ocean current drift error.

[0106] The preliminary detection subunit 311 uses an improved Kalman filter to fuse ADCP data and IMU data, estimate drift error, and output preliminary error value.

[0107] Then, a Hidden Markov Model (HMM) is used to calculate the time series characteristics of the initial error, such as autocorrelation and trend change rate; the future trend of the error is also predicted by the HMM to obtain the initial error trend.

[0108] The burst detection subunit 312 uses a dual sliding window method to set a short-term window and a long-term window, and calculates the mean and standard deviation of the drift error respectively; when the mean of the short-term window exceeds the 3σ threshold of the long-term window, the burst error point is marked.

[0109] The burst detection subunit 312 also uses the isolated forest algorithm to train an unsupervised anomaly detection model to identify abnormal drift errors; at the same time, it combines the sliding window method to remove normal errors and retain burst error points.

[0110] Enhanced detection subunit 313 employs a bidirectional long short-term memory network for time series prediction, a process that includes:

[0111] Train the Bi-LSTM model by inputting historical ADCP and IMU data;

[0112] Predict the trend of error changes within a short time window to obtain prediction error data.

[0113] The fusion detection subunit 314 calculates the final drift error based on a weighted fusion method, and the calculation formula is configured as follows:

[0114] in, Indicates the final drift error. Indicates the initial error trend. Indicates sudden error. Indicates the prediction error. The first, second, and third fusion coefficients, which are preset, are dynamically adjusted by a Bayesian optimization algorithm to ensure that the weights for different errors are optimal.

[0115] Preferably, the error analysis module 3 includes an environmental error detection unit 32, which includes:

[0116] The error screening unit 321 is used to calculate the mean, variance, skewness and kurtosis of temperature and pressure data, and to obtain the screening data after removing outlier data based on the 3σ principle;

[0117] The modeling and prediction subunit 322 is connected to the error screening subunit 321. It is used to acquire historical temperature data, historical pressure data and historical environmental errors, and to construct a one-way long short-term memory network based on historical temperature data, historical pressure data and historical environmental errors. The screened data is input into the one-way long short-term memory network to predict the environmental error.

[0118] Model adjustment subunit 323 is connected to modeling prediction subunit 322, and is used to calculate the confidence level of environmental error based on Bayesian inference, and dynamically adjust the weights of the unidirectional long short-term memory network based on the confidence level.

[0119] Specifically, in this embodiment, the error screening unit 321 calculates the mean, variance, skewness, and kurtosis of the temperature and pressure data, analyzes the data distribution, and then removes outliers based on the 3σ principle: if the data exceeds the mean ± 3σ, it is considered abnormal, removed, and the abnormal data points are recorded. The output of the error screening unit 321 is the sieved data after removing outliers.

[0120] The modeling and prediction subunit 322 trains a unidirectional long short-term memory network. The model adjustment subunit 323 first calculates the confidence level of the environmental error based on Bayesian inference: it calculates the confidence interval of the prediction error and sets a confidence threshold; then it dynamically adjusts the LSTM weights: if the confidence level is low, it increases the weights of historical data to make the model more sensitive to long-term trends; if the confidence level is high, it maintains the current model parameters to improve real-time prediction capabilities. The output of the model adjustment subunit 323 is the adjusted predicted value of the environmental error.

[0121] The one-way long short-term memory network was finally tested at a buoy monitoring station in a certain sea area:

[0122] Data source: 20,000 sets of ADCP, GPS, IMU, temperature, and pressure data collected over 6 months;

[0123] Detection results: Drift error detection accuracy improved by 35% (Isolated Forest algorithm effectively reduces false detection rate);

[0124] Environmental error detection error reduced by 20% (Bayesian optimization improves the reliability of LSTM predictions);

[0125] The overall error compensation accuracy is improved by 30%, making the water quality monitoring data more reliable.

[0126] This embodiment demonstrates that the error detection method integrating multiple algorithms can effectively improve the reliability of marine environmental monitoring.

[0127] Preferred, Reference Figure 4 The data compensation module 4 includes:

[0128] Error feature extraction unit 41 is used to time-align drift error, environmental error and water quality monitoring data, and extract error features of drift error and environmental error according to principal component analysis method and empirical mode decomposition method;

[0129] Incremental learning unit 42 is connected to error feature extraction unit 41, and is used to enable the initial compensation model to learn the extracted error features online and perform error correction based on the incremental learning method to obtain the correction compensation model.

[0130] The dynamic optimization unit 43 is connected to the incremental learning unit 42 and is used to dynamically adjust the hyperparameters of the correction compensation model according to the Bayesian optimization algorithm to obtain the optimized compensation model.

[0131] Specifically, this embodiment is used for time alignment error data, and the error features are extracted by principal component analysis and empirical mode decomposition to ensure that the initial compensation model can make full use of the error modes.

[0132] Error feature extraction unit 41 uses linear interpolation to align the time of different data sources, ensuring that all data have the same time dimension. Then, dynamic time warping (DTW) is used to match the time series data, improving the alignment accuracy of data with different sampling rates.

[0133] The error feature extraction unit 41 extracts the main error information by calculating the feature vectors of drift error and environmental error, thereby reducing data dimensionality and improving computational efficiency.

[0134] The error feature extraction unit 41 also decomposes drift error and environmental error into intrinsic mode functions (IMF) of different frequencies, removes high-frequency noise, and improves the interpretability of error modes.

[0135] Incremental learning unit 42 is based on the incremental learning method, which enables the error compensation model to have online learning capabilities and continuously optimize the error correction effect.

[0136] The incremental learning unit 42 allows it to dynamically update its weights; each time the initial compensation model receives new error feature data, it is fine-tuned using parameter transfer methods instead of retraining the entire model.

[0137] The incrementally learned model is compared with the initial compensation model. If the new model performs better, it is updated to the correction compensation model. The sliding window method is used to limit the time range of model memory to avoid the model overfitting to historical data.

[0138] The dynamic optimization unit 43 uses Bayesian optimization algorithm to dynamically adjust the hyperparameters of the correction and compensation model, thereby further improving the accuracy of error compensation.

[0139] The steps involved in Bayesian optimization hyperparameter tuning are as follows:

[0140] Define the hyperparameter search space (such as learning rate, number of neural network layers, regularization parameters, etc.).

[0141] Gaussian process regression (GPR) is used to estimate the effects of different hyperparameter configurations.

[0142] A Bayesian optimization strategy is employed: During the hyperparameter search process, the Expectation Improvement (EI) criterion is introduced to prioritize high-potential hyperparameter combinations; the contribution of hyperparameters to the error compensation effect is calculated, and the hyperparameter configuration is gradually optimized. Iterative optimization continues until the optimal parameter combination is found, and the output is the final optimized compensation model.

[0143] The optimized compensation model was tested at a buoy monitoring station in a certain sea area:

[0144] Data source: 15,000 sets of ADCP, GPS, IMU, temperature, pressure and water quality monitoring data collected over 6 months.

[0145] Error compensation effect: After extracting error features using principal component analysis and empirical mode decomposition methods, the error is reduced by 25%; incremental learning method improves the adaptive ability of the error compensation model by 30%; Bayesian optimization further improves the error compensation accuracy by 20%.

[0146] This embodiment demonstrates that combining principal component analysis, empirical mode decomposition, incremental learning, and Bayesian optimization for data compensation can effectively improve error correction accuracy and enhance the reliability of water quality monitoring data.

[0147] Preferred, Reference Figure 5 The data fusion module 5 includes:

[0148] The multimodal feature extraction unit 51 is used to perform time alignment on the compensation monitoring data of multiple sensing devices, and to extract sensing features from each compensation monitoring data according to the principal component analysis method and the deep learning feature extraction algorithm.

[0149] The weight adjustment unit 52 is connected to the multimodal feature extraction unit 51. It is used to calculate the feature importance of each sensing feature based on the information entropy, and dynamically adjust the weight of the corresponding compensation monitoring data of the sensing feature based on the feature importance and the Bayesian optimization algorithm.

[0150] The weighted fusion unit 53 is connected to the weight adjustment unit 52 and is used to perform weighted fusion based on the adjusted weights and the corresponding compensation monitoring data to obtain preliminary corrected data.

[0151] The timing optimization unit 54 is connected to the weighted fusion unit 53 and is used to perform timing optimization on the preliminary correction data according to the particle filter algorithm to obtain error correction data.

[0152] Specifically, in this embodiment, the multimodal feature extraction unit 51 uses a linear interpolation method to align the compensation monitoring data of multiple sensing devices at the same time point, while calculating the time deviation between different compensation monitoring data; finally, nonlinear alignment is performed to improve data consistency.

[0153] The multimodal feature extraction unit 51 calculates the principal components of the data through principal component analysis and extracts the most representative sensing features; then, through dimensionality reduction, it reduces redundancy and improves computational efficiency.

[0154] The multimodal feature extraction unit 51 also uses a convolutional neural network (CNN) to extract spatial features; at the same time, it uses a bidirectional long short-term memory network (Bi-LSTM) to identify time series features, thereby improving the ability to model time series information.

[0155] The weight adjustment unit 52 dynamically adjusts the importance weights of sensing features based on information entropy calculation and Bayesian optimization: The processing steps for calculating feature importance are: calculating the information entropy of each sensing feature; features with high entropy values ​​indicate high information richness and are assigned higher weights.

[0156] Bayesian optimization for dynamically adjusting weights: Define a weight search space and use Gaussian process regression (GPR) to estimate the effects of different weight combinations; use the expected improvement (EI) criterion to optimize the weights and minimize the error.

[0157] The weighted fusion unit 53 uses a weighted fusion method to fuse data according to the adjusted weights, thereby improving the accuracy of error correction. The timing optimization unit 54 uses a particle filter algorithm to optimize the timing, thereby improving the stability of the fused data.

[0158] The processing steps for particle filter timing optimization include:

[0159] Initialize the particle set: Set multiple particles to represent different error correction possibilities; randomly initialize the weight of each particle using a Gaussian distribution.

[0160] Particle state update: Calculate the residual between the observed and predicted values; update the particle weights so that particles with smaller errors receive higher weights.

[0161] Resampling: Low-weight particles are eliminated, and high-weight particles are copied to reduce sampling error.

[0162] The output is the final error correction data.

[0163] Preferred, Reference Figure 6 Feedback optimization module 6 includes:

[0164] The deviation calculation unit 61 is used to calculate the deviation feedback data by processing the compensation monitoring data and error correction data based on the Euclidean distance.

[0165] The compensation processing unit 62 is connected to the deviation calculation unit 61 and is used to process the deviation feedback data to obtain the compensation accuracy improvement amount.

[0166] The model optimization unit 63 is connected to the compensation processing unit 62. It is used to dynamically adjust the weight parameters of the learning and optimization compensation model to obtain the adjusted optimized compensation model by adjusting the amount of compensation accuracy improvement.

[0167] Preferably, the formula for calculating the deviation feedback data is configured as follows:

[0168] ;

[0169] in, Used to represent deviation feedback data. Used to represent compensation monitoring data, Used to represent error correction data. Used to represent the total number of deviation feedback data.

[0170] Specifically, in this embodiment, the compensation processing unit 62 is used to calculate the improvement amount of error compensation based on the deviation feedback data, so as to ensure that the compensation model can be continuously optimized.

[0171] The compensation processing unit 62 uses a moving average filter to calculate the nearest... Average error over time steps:

[0172]

[0173] in, Indicates the average error. Indicates the first Individual deviation feedback data.

[0174] Simultaneously, the compensation processing unit 62 sets the compensation accuracy adjustment rules: if If the value exceeds the preset compensation threshold, the compensation amount will be increased; if... If the compensation amount is below the preset compensation threshold, the compensation amount will be reduced.

[0175] The compensation processing unit 62 uses an exponential weighting method to calculate the improvement in compensation accuracy:

[0176]

[0177] in: The smoothing coefficient controls the impact of historical data on the current compensation. This represents the current improvement in compensation accuracy.

[0178] The model optimization unit 63 dynamically adjusts the weight parameters of the error compensation model based on the improvement in compensation accuracy, so that it can continuously adapt to new error information.

[0179] The specific steps for adjusting the weight parameters of the optimization compensation model in model optimization unit 63 are as follows:

[0180] Dynamically adjust model weights based on Bayesian optimization:

[0181] Set model weight parameters

[0182] The goal of weight adjustment is to find the optimal weight combination that minimizes the error.

[0183]

[0184] In this process, Gaussian process regression (GPR) is used to estimate the error compensation effect of different weight combinations; at the same time, the expectation boosting strategy is used to adjust the weights and improve the compensation accuracy.

[0185] During the process of dynamically updating the compensation model in model optimization unit 63: if the error of the optimized compensation model decreases by more than the threshold, the model parameters are updated; if the error does not improve significantly, the original model weights are maintained.

[0186] The final output of the model optimization unit 63 is the adjusted optimized compensation model.

[0187] An error correction method for marine environmental observation data is proposed, applied to the aforementioned error correction system for marine environmental observation data, with reference to... Figure 7 ,include:

[0188] Step S1: Data acquisition module 1 collects ADCP flow measurement data, GPS data, IMU data, temperature data, pressure data, and water quality monitoring data from the water quality sensing device in real time.

[0189] Step S2: The compensation modeling module 2 acquires multiple historical training data from the water quality sensing device and constructs an initial compensation model based on each historical training data.

[0190] Step S3: Error analysis module 3 performs anomaly detection on ADCP flow measurement data, GPS data, and IMU data based on the improved Kalman filter algorithm and hidden Markov model to obtain drift error, and performs anomaly detection on temperature data and pressure data based on long short-term memory network to obtain environmental error.

[0191] Step S4: The data compensation module 4 corrects the initial compensation model based on drift error and environmental error to obtain an optimized compensation model. The optimized compensation model is used to compensate for errors in water quality monitoring data to obtain compensated monitoring data.

[0192] In step S5, the feedback optimization module 6 processes the compensation monitoring data and error correction data to obtain deviation feedback data, and dynamically adjusts the weight parameters of the optimization compensation model based on the deviation feedback data to obtain the adjusted optimization compensation model.

[0193] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. An error correction system for oceanographic observation data, suitable for water quality sensing equipment on a buoy monitoring station, an unmanned underwater vehicle, and an underwater sensor array, characterized in that, The method comprises the following steps: A data acquisition module (1) is used to acquire ADCP flow measurement data, GPS data, IMU data, temperature data, and pressure data on the water quality sensing device in real time, and to acquire water quality monitoring data of the water quality sensing device; A compensation modeling module (2) is used to obtain a plurality of historical training data of the water quality sensing device, and to construct an initial compensation model based on each historical training data; An error analysis module (3) is connected to the data acquisition module (1) and is used to detect drift errors of the ADCP flow measurement data, the GPS data, and the IMU data according to an improved Kalman filtering algorithm and a hidden Markov model, and to detect environmental errors of the temperature data and the pressure data according to a long short-term memory network; A data compensation module (4) is connected to the data acquisition module (1), the compensation modeling module (2), and the error analysis module (3), and is used to correct errors of the initial compensation model according to the drift errors and the environmental errors to obtain an optimized compensation model, wherein the optimized compensation model is used to compensate errors of water quality monitoring data to obtain compensated monitoring data; A data fusion module (5) is connected to the data compensation module (4) and is used to fuse and optimize the compensated monitoring data of a plurality of sensing devices according to a particle filtering algorithm and a Bayesian optimization algorithm to obtain error correction data; A feedback optimization module (6) is connected to the data fusion module (5) and the data compensation module (4), and is used to process deviation feedback data from the compensated monitoring data and the error correction data, to dynamically adjust weight parameters of the optimized compensation model based on the deviation feedback data, and to obtain an adjusted optimized compensation model.

2. The error correction system for ocean environment observation data according to claim 1, characterized in that: The historical training data comprises historical error data, historical real water quality data, and historical sensing monitoring data, and the compensation modeling module (2) comprises: A preprocessing unit (21) is used to preprocess each historical error data, each historical real water quality data, and each historical sensing monitoring data to obtain a plurality of preprocessed historical error data, a plurality of preprocessed historical real water quality data, and a plurality of preprocessed historical sensing monitoring data; An introduction unit (22) is used to introduce a CNN-LSTM hybrid model as an original model; A training unit (23) is connected to the introduction unit (22) and the preprocessing unit (21), and is used to retrain the original model by taking the preprocessed historical error data and the preprocessed historical sensing monitoring data as inputs and taking the corresponding historical real water quality data as outputs to obtain the initial compensation model.

3. The error correction system for ocean environment observation data according to claim 1, characterized in that: The error analysis module (3) comprises a drift error detection unit (31), and the drift error detection unit (31) comprises: A preliminary detection subunit (311) is used to preliminarily estimate drift errors according to an improved Kalman filtering algorithm to obtain preliminary errors, and to analyze time sequence characteristics of the preliminary errors according to a hidden Markov model and to predict preliminary error trends; a burst detection subunit (312) configured to detect burst errors according to a double sliding window method and an isolation forest algorithm; an enhancement detection subunit (313) configured to obtain historical data of the ADCP flow measurement data and the IMU data, and input the historical data into a preset bidirectional long short-term memory network to obtain prediction errors; a fusion detection subunit (314) connected to the preliminary detection subunit (311), the burst detection subunit (312), and the enhancement detection subunit (313), and configured to fuse the preliminary error trend, the burst errors, and the prediction errors to obtain the drift errors.

4. The error correction system for ocean environment observation data according to claim 1, wherein: The error analysis module (3) comprises an environmental error detection unit (32), which comprises: an error preliminary screening subunit (321) configured to calculate the mean, variance, skewness, and kurtosis of the temperature and pressure data, and obtain screened data by removing abnormal data based on a 3σ principle; a modeling and prediction subunit (322) connected to the error preliminary screening subunit (321), and configured to obtain historical temperature data, historical pressure data, and historical environmental errors, construct a unidirectional long short-term memory network based on the historical temperature data, the historical pressure data, and the historical environmental errors, and input the screened data into the unidirectional long short-term memory network to predict the environmental errors; a model adjustment subunit (323) connected to the modeling and prediction subunit (322), and configured to calculate the reliability of the environmental errors according to Bayesian inference, and dynamically adjust the weights of the unidirectional long short-term memory network based on the reliability.

5. The error correction system for ocean environment observation data according to claim 1, wherein: The data compensation module (4) comprises: an error feature extraction unit (41) configured to time-align the drift errors, the environmental errors, and the water quality monitoring data, and extract error features of the drift errors and the environmental errors according to a principal component analysis method and an empirical mode decomposition method; an incremental learning unit (42) connected to the error feature extraction unit (41), and configured to extract the error features and correct errors based on an incremental learning method, so as to obtain a corrected compensation model; a dynamic optimization unit (43) connected to the incremental learning unit (42), and configured to dynamically adjust hyperparameters of the corrected compensation model according to a Bayesian optimization algorithm, so as to obtain an optimized compensation model.

6. The error correction system for ocean environment observation data according to claim 1, wherein: The data fusion module (5) comprises: a multi-modal feature extraction unit (51) configured to time-align the compensation monitoring data of a plurality of the sensing devices, and extract sensing features of each of the compensation monitoring data according to a principal component analysis method and a deep learning feature extraction algorithm; a weight adjustment unit (52) connected to the multi-modal feature extraction unit (51), and configured to calculate feature importance of each of the sensing features according to information entropy, and dynamically adjust weights of the compensation monitoring data corresponding to the sensing features according to the feature importance and a Bayesian optimization algorithm; and a data fusion unit (53) connected to the weight adjustment unit (52), and configured to fuse the compensation monitoring data of the sensing devices according to the weights, so as to obtain fused data. A weighting fusion unit (53) connected to the weight adjusting unit (52) is configured to perform weighting fusion on the adjusted weights and the corresponding compensation monitoring data to obtain preliminary correction data; A time sequence optimization unit (54) connected to the weighting fusion unit (53) is configured to perform time sequence optimization on the preliminary correction data according to a particle filtering algorithm to obtain error correction data.

7. The error correction system for ocean environment observation data according to claim 1, wherein: The feedback optimization module (6) comprises: A bias calculation unit (61) configured to calculate bias feedback data from the compensation monitoring data and the error correction data according to a Euclidean distance; A compensation processing unit (62) connected to the bias calculation unit (61) and configured to obtain a compensation accuracy improvement amount from the bias feedback data; A model optimization unit (63) connected to the compensation processing unit (62) and configured to dynamically adjust the weight parameters of the optimized compensation model according to the compensation accuracy improvement amount to obtain an adjusted optimized compensation model.

8. The error correction system for ocean environment observation data according to claim 7, characterized in that: The calculation formula of the bias feedback data is configured as: ； wherein, for representing the deviation feedback data, for representing the compensation monitoring data, for representing the error correction data, for representing the total number of deviation feedback data.

9. A method for error correction of marine environmental observation data, applied to the error correction system for marine environmental observation data according to any one of claims 1-8, characterized in that, Comprises: Step S1, the data acquisition module (1) acquires the ADCP flow measurement data, GPS data, IMU data, temperature data and pressure data on the water quality sensing device in real time, and acquires the water quality monitoring data of the water quality sensing device; Step S2, the compensation modeling module (2) obtains a plurality of historical training data of the water quality sensing device, and constructs an initial compensation model based on each historical training data; Step S3, the error analysis module (3) performs abnormal detection on the ADCP flow measurement data, the GPS data and the IMU data according to an improved Kalman filtering algorithm and a hidden Markov model to obtain drift error, and performs abnormal detection on the temperature data and the pressure data according to a long short-term memory network to obtain environmental error; Step S4, the data compensation module (4) corrects errors of the initial compensation model according to the drift error and the environmental error to obtain an optimized compensation model, and the optimized compensation model is used for error compensation of water quality monitoring data to obtain compensation monitoring data; Step S4a, the data fusion module (5) performs fusion optimization on the compensation monitoring data of a plurality of sensing devices according to a particle filtering algorithm and a Bayesian optimization algorithm to obtain error correction data; Step S5, the feedback optimization module (6) processes bias feedback data from the compensation monitoring data and the error correction data, and dynamically adjusts the weight parameters of the optimized compensation model based on the bias feedback data to obtain an adjusted optimized compensation model.

Images