A Quality Control Method for Marine Meteorological Data Based on Cross-parameter Correlation Networks

Abstract

This invention provides a quality control method for marine meteorological data based on a cross-parameter correlation network, belonging to the field of marine meteorological technology. This invention constructs a cross-parameter correlation network model including correlation rules for air temperature and air pressure, linear correlation rules for humidity and air temperature, and square root correlation rules for wind speed and air pressure gradients. It employs a sliding window algorithm to calculate correlation coefficients in real time and identify correlation anomalies. A multi-dimensional anomaly detection mechanism is established, encompassing four dimensions: parameter threshold exceeding limits, spatiotemporal rate of change anomalies, probability density distribution shifts, and correlation verification failures. A fuzzy comprehensive evaluation method is used to calculate a comprehensive anomaly index and determine data anomalies. A multi-sensor cross-validation mechanism is combined to confirm the authenticity of abnormal data. Finally, a virtual sensor data reconstruction algorithm is used to reverse-calculate and repair abnormal data based on the correlation network, thus solving the technical problem of poor anomaly detection performance in existing technologies.

Description

Technical Field

[0001] This invention belongs to the field of marine meteorological technology, and more specifically, relates to a marine meteorological data quality control method based on cross-parameter correlation networks. Background Technology

[0002] In the field of marine meteorological observation, traditional data quality control techniques mainly employ methods such as single-parameter threshold testing, temporal consistency testing, and spatial consistency testing to control the quality of marine meteorological observation data. These methods are widely used in marine environmental monitoring systems such as marine meteorological stations, buoy observation networks, and ship automatic weather stations, identifying anomalous data by setting fixed parameter threshold ranges and rate of change limits. However, traditional quality control methods have significant shortcomings when dealing with the complex and variable meteorological parameters of the marine environment. These shortcomings primarily manifest in neglecting the inherent correlations between marine meteorological parameters, failing to effectively identify anomalous data that meets single-parameter threshold requirements but violates the correlation laws between parameters, and lacking the ability to dynamically adjust parameter correlations under different sea areas and seasonal conditions. In current marine meteorological data quality control practices, because traditional methods rely solely on single-dimensional anomaly judgment criteria, they struggle to accurately identify the complex nonlinear correlations between multiple meteorological parameters such as air temperature, air pressure, humidity, and wind speed in the marine environment. This leads to frequent instances of missed and false anomaly detections. In other words, existing technologies suffer from insufficient accuracy in identifying multi-parameter correlations during marine meteorological data quality control, resulting in poor anomaly detection performance. Summary of the Invention

[0003] In view of this, the present invention provides a marine meteorological data quality control method based on cross-parameter correlation networks, which can solve the technical problem in the prior art that the accuracy of multi-parameter correlation identification in the marine meteorological data quality control process is insufficient, resulting in poor abnormal data detection effect.

[0004] This invention is implemented as follows: It provides a marine meteorological data quality control method based on a cross-parameter correlation network, comprising constructing a marine meteorological parameter correlation network model to obtain correlation rules for air temperature and pressure, linear correlation rules for humidity and air temperature, and square root correlation rules for wind speed and air pressure gradients, forming a cross-parameter correlation network structure; implementing sliding window correlation coefficient calculation, calculating the correlation coefficient of each parameter in the cross-parameter correlation network structure in real time, triggering a correlation anomaly flag when the real-time correlation coefficient deviates from the standard correlation coefficient by more than a preset threshold; and establishing a multi-dimensional anomaly signal detection mechanism to detect signals exceeding parameter thresholds, anomaly signals in spatiotemporal change rates, probability density distribution shift signals, and phase... The system identifies and quantifies the correlation verification failure signal to generate corresponding intensity values; it then executes an adaptive weight optimization algorithm to calculate dynamic weight values ​​using an adaptive weight optimization model for ocean parameters; fuzzy comprehensive evaluation is applied to perform weighted fusion calculations on the intensity values ​​and dynamic weight values ​​to calculate a comprehensive anomaly index, which is determined to be data anomaly when it exceeds a preset threshold; multi-sensor cross-validation is implemented to determine the authenticity of the data anomaly and calculate sensor reliability scores; finally, anomaly data repair and reconstruction is performed using a virtual sensor data reconstruction algorithm, which solves for the estimated values ​​of the anomaly parameters through the inverse calculation equation of the correlation network, and outputs the repaired marine meteorological parameter data.

[0005] The steps involved in constructing a marine meteorological parameter correlation network model are as follows: First, an iterative regression algorithm is used to analyze and process historical marine meteorological data to obtain the correlation rules between air temperature and air pressure. Second, linear correlation rules between humidity and air temperature and square root correlation rules between wind speed and air pressure gradient are established to form a cross-parameter correlation network structure containing six sets of core parameters. In the cross-parameter correlation network structure, each parameter node is connected through a correlation coefficient.

[0006] Specifically, the cross-parameter correlation network structure is a network topology model based on the principles of marine meteorology, which establishes the interconnection between multiple meteorological parameters. It describes the coordinated change patterns of various meteorological elements in the marine environment by quantifying the mathematical relationships between different parameters.

[0007] Specifically, the temperature and air pressure correlation rule is obtained by using least squares regression analysis and polynomial fitting algorithm to iteratively calculate the piecewise negative correlation mathematical relationship between temperature and air pressure from a large amount of historical marine meteorological observation data. The final correlation coefficient and applicable temperature range are determined through more than 1,000 iterations of optimization.

[0008] Specifically, the linear correlation rule between humidity and temperature is obtained by statistically analyzing the observational data of relative humidity and temperature in the marine environment using a linear regression algorithm. The correlation coefficient is determined by the Pearson correlation coefficient calculation method.

[0009] Specifically, the square root correlation rule of wind speed and air pressure gradient is based on the principle of fluid mechanics to establish the square root function relationship between wind speed and air pressure gradient. The function parameters are determined by nonlinear fitting analysis of measured wind speed data and air pressure gradient data from the same period.

[0010] Specifically, the step of implementing the sliding window correlation coefficient calculation involves using a sliding window algorithm with a time window length of 30 minutes to calculate the correlation coefficient of each parameter in the cross-parameter correlation network structure in real time. When the calculated real-time correlation coefficient deviates by more than 25% from the standard correlation coefficient corresponding to the temperature and pressure correlation rule, the humidity and temperature linear correlation rule, and the wind speed and pressure gradient square root correlation rule, a correlation anomaly flag is triggered.

[0011] The steps of establishing a multi-dimensional abnormal signal detection mechanism specifically involve identifying and quantifying four dimensions: parameter threshold exceeding limit signal, spatiotemporal change rate abnormal signal, probability density distribution shift signal, and correlation verification failure signal, to generate corresponding parameter threshold exceeding limit intensity values, spatiotemporal change rate abnormal intensity values, probability density distribution shift intensity values, and correlation verification failure intensity values.

[0012] Specifically, the step of executing the adaptive weight optimization algorithm involves calculating dynamic weight values ​​for the intensity of parameter threshold exceeding limits, the intensity of spatiotemporal change rate anomalies, the intensity of probability density distribution shifts, and the intensity of correlation verification failures through the ocean parameter adaptive weight optimization model. The ocean parameter adaptive weight optimization model adjusts the hierarchical fusion weight parameters of the model based on the current sea area depth data, geographic latitude data, and seasonal temperature difference data.

[0013] The specific steps of applying fuzzy comprehensive evaluation calculation are as follows: weighted fusion calculation is performed on the parameter threshold exceeding intensity value, spatiotemporal change rate anomaly intensity value, probability density distribution offset intensity value, and correlation verification failure intensity value with dynamic weight value; the comprehensive anomaly index is calculated according to the fuzzy comprehensive evaluation calculation rules; when the comprehensive anomaly index exceeds the preset threshold of 0.7, it is judged as data anomaly.

[0014] The steps of implementing multi-sensor cross-validation processing specifically involve initiating a multi-sensor cross-validation mechanism for identified data anomalies, determining the authenticity of the data anomalies through comparative analysis of redundant sensor observation data, and simultaneously calling the sensor health assessment function to process and calculate the sensor reliability score based on the sensor output data variance, data drift trend coefficient, cross-validation consistency index, sensor operating time, and environmental corrosion degree assessment value.

[0015] The step of performing abnormal data repair and reconstruction processing involves using a virtual sensor data reconstruction algorithm to reconstruct confirmed data anomalies. This algorithm inputs normal parameter values ​​from a cross-parameter correlation network structure, along with correlation rules for air temperature and pressure, linear correlation rules for humidity and air temperature, and square root correlation rules for wind speed and air pressure gradients. The algorithm then solves for estimated values ​​of the abnormal parameters by inversely calculating the correlation network equations, and outputs the repaired marine meteorological parameter data to complete the quality control process.

[0016] The ocean parameter adaptive weight optimization model is specifically a multi-layer neural network model based on a hierarchical attention network architecture, which includes four main components: an input layer, a feature extraction layer, an attention mechanism layer, and an output layer. The attention mechanism layer uses a multi-level attention structure to assign weights to features of different sea areas and seasonal features.

[0017] The sensor health assessment function is used to calculate the operational reliability score of the marine meteorological sensor. The inputs include the sensor output data variance, data drift trend coefficient, cross-validation consistency index, sensor operating time, and environmental corrosion degree assessment value. The output is a sensor reliability score between 0 and 1. When the sensor reliability score is lower than 0.6, it indicates that the sensor needs maintenance and calibration.

[0018] Specifically, the virtual sensor data reconstruction algorithm is a data reconstruction algorithm that uses mathematical operations to calculate the values ​​of missing or abnormal parameters based on the correlation between parameters. The normal parameter values ​​are specifically the reliable marine meteorological parameter observation values ​​that have been verified through quality control in the cross-parameter correlation network structure.

[0019] Specifically, the correlation network inverse calculation equation is a set of mathematical equations established based on known normal parameter values ​​and correlation rules to solve for the estimated values ​​of abnormal parameters. The estimated values ​​of the abnormal parameters are specifically the inferred parameter values ​​obtained by solving the correlation network inverse calculation equation to replace the abnormal data.

[0020] This invention addresses the technical problem of poor anomaly detection caused by insufficient accuracy in identifying multi-parameter correlations during marine meteorological data quality control. It constructs a cross-parameter correlation network model and combines it with a multi-dimensional anomaly detection mechanism and an adaptive weight optimization algorithm. Specifically, it establishes a cross-parameter correlation network using correlation rules based on historical data analysis, including temperature-pressure correlation rules, humidity-temperature linear correlation rules, and wind speed-pressure gradient square root correlation rules. A sliding window algorithm is used to monitor changes in correlation coefficients between parameters in real time. A comprehensive analysis of four dimensions—parameter threshold exceeding limits, spatiotemporal rate of change anomalies, probability density distribution shifts, and correlation verification failures—effectively improves the accuracy and reliability of anomaly data identification. Furthermore, this invention uses an adaptive weight optimization model for marine parameters to dynamically adjust the detection weights of each dimension based on environmental characteristics such as sea depth, latitude, and seasonal temperature differences. A fuzzy comprehensive evaluation method is employed to calculate a comprehensive anomaly index, overcoming the technical deficiency of inaccurate parameter correlation identification in complex marine environments. In summary, this invention solves the technical problem mentioned in the background art of poor anomaly detection caused by insufficient accuracy in identifying multi-parameter correlations during marine meteorological data quality control. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention.

[0022] Figure 2 This is a time series variation diagram of marine meteorological parameters in Example 2.

[0023] Figure 3 This is a schematic diagram of the cross-parameter correlation network structure in Example 2.

[0024] Figure 4 This is a statistical chart of the multi-dimensional anomaly detection results in Example 2. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0026] like Figure 1 The diagram shown is a flowchart of a marine meteorological data quality control method based on a cross-parameter correlation network provided by the present invention. This method includes the following steps:

[0027] S01. Construct a marine meteorological parameter correlation network model, and use an iterative regression algorithm to analyze and process historical marine meteorological data to obtain the correlation rules of air temperature and air pressure. At the same time, establish the linear correlation rules of humidity and air temperature and the square root correlation rules of wind speed and air pressure gradient, forming a cross-parameter correlation network structure containing 6 sets of core parameters. In the cross-parameter correlation network structure, each parameter node is connected by a correlation coefficient.

[0028] S02. Implement sliding window correlation coefficient calculation. Use a sliding window algorithm with a time window length of 30 minutes to calculate the correlation coefficient of each parameter in the cross-parameter correlation network structure in real time. When the calculated real-time correlation coefficient deviates from the standard correlation coefficient in the temperature and pressure correlation rule, humidity and temperature linear correlation rule and wind speed and pressure gradient square root correlation rule by more than 25%, trigger the correlation anomaly flag.

[0029] S03. Establish a multi-dimensional abnormal signal detection mechanism to identify and quantify the four dimensions of parameter threshold over-limit signal, spatiotemporal change rate abnormal signal, probability density distribution offset signal and correlation verification failure signal respectively, and generate corresponding parameter threshold over-limit intensity value, spatiotemporal change rate abnormal intensity value, probability density distribution offset intensity value and correlation verification failure intensity value.

[0030] S04. Execute the adaptive weight optimization algorithm. Calculate dynamic weight values ​​for the parameter threshold exceeding intensity, spatiotemporal change rate anomaly intensity, probability density distribution shift intensity, and correlation verification failure intensity values ​​respectively through the marine parameter adaptive weight optimization model. The marine parameter adaptive weight optimization model adjusts the hierarchical fusion weight parameters of the model based on the current sea area water depth data, geographic latitude data, and seasonal temperature difference data.

[0031] S05. Apply fuzzy comprehensive evaluation calculation and processing, perform weighted fusion calculation on the parameter threshold over-limit intensity value, spatiotemporal change rate anomaly intensity value, probability density distribution offset intensity value and correlation verification failure intensity value with the dynamic weight value, calculate the comprehensive anomaly index through fuzzy comprehensive evaluation calculation rules, and determine the data anomaly when the comprehensive anomaly index exceeds the preset threshold of 0.7.

[0032] S06. Implement multi-sensor cross-validation processing. Initiate multi-sensor cross-validation mechanism for identified data anomalies. Determine the authenticity of data anomalies through comparative analysis of redundant sensor observation data. At the same time, call the sensor health assessment function to process and calculate sensor reliability score based on sensor output data variance, data drift trend coefficient, cross-validation consistency index, sensor working time, and environmental corrosion degree assessment value.

[0033] S07. Perform abnormal data repair and reconstruction processing. For confirmed data anomalies, a virtual sensor data reconstruction algorithm is used. Input the normal parameter values ​​in the cross-parameter correlation network structure and the correlation rules of air temperature and air pressure, linear correlation rules of humidity and air temperature, and square root correlation rules of wind speed and air pressure gradient. Solve the estimated values ​​of abnormal parameters through the inverse calculation equation of the correlation network, and output the repaired marine meteorological parameter data to complete the quality control process.

[0034] Among them, the cross-parameter correlation network structure refers to a network topology model based on the principles of marine meteorology, which establishes the interconnection between multiple meteorological parameters. It describes the coordinated change patterns of various meteorological elements in the marine environment by quantifying the mathematical relationships between different parameters. The temperature-pressure correlation rule refers to the piecewise negative correlation between temperature and pressure obtained through iterative calculations using least squares regression analysis and polynomial fitting algorithms on a large amount of historical marine meteorological observation data. When the temperature is above 25℃, the correlation coefficient is negative and the absolute value fluctuates within the range of 0.5 to 0.7. The final correlation coefficient and applicable temperature range are determined through more than 1000 iterations of optimization. The humidity-temperature linear correlation rule refers to the linear correlation obtained by statistical analysis of observational data on relative humidity and temperature in the marine environment using a linear regression algorithm. The correlation coefficient is determined using the Pearson correlation coefficient calculation method.

[0035] The wind speed and pressure gradient square root correlation rule refers to the square root function relationship between wind speed and pressure gradient established based on fluid dynamics principles. The function parameters are determined through nonlinear fitting analysis of measured wind speed data and concurrent pressure gradient data. The sliding window algorithm is a data processing method that uses a fixed-length data window to continuously move across the time series, performing statistical analysis and calculations on the data within the window. The real-time correlation coefficient is the correlation value between parameters calculated within the current time window using the sliding window algorithm. The standard correlation coefficient is a pre-determined benchmark value for correlation among the air-temperature-pressure correlation rule, the humidity-air-humidity linear correlation rule, and the wind speed and pressure gradient square root correlation rule. The correlation anomaly indicator is an abnormal alarm signal generated when the real-time correlation coefficient deviates from the standard correlation coefficient by more than a set threshold. The parameter threshold exceeding the limit signal is an abnormal prompt signal generated when the marine meteorological parameter value exceeds the normal range of change. The spatiotemporal change rate anomaly signal is an abnormal prompt signal generated when the rate of change of marine meteorological parameters in the time or spatial dimensions exceeds a reasonable range. The probability density distribution offset signal is an abnormal prompt signal generated when the statistical distribution characteristics of marine meteorological parameters deviate from the historical normal distribution pattern.

[0036] Among them, the correlation verification failure signal refers to the abnormal prompt signal generated when the correlation between parameters does not conform to the preset rules. The parameter threshold exceeding the limit intensity value refers to the abnormal intensity assessment value obtained by quantitatively analyzing the degree of parameter threshold exceeding the limit. The spatiotemporal change rate abnormal intensity value refers to the abnormal intensity assessment value obtained by quantitatively analyzing the degree of deviation of the spatiotemporal change rate. The probability density distribution offset intensity value refers to the abnormal intensity assessment value obtained by quantitatively analyzing the degree of probability density distribution offset. The correlation verification failure intensity value refers to the abnormal intensity assessment value obtained by quantitatively analyzing the degree of correlation verification failure. The ocean parameter adaptive weight optimization model refers to a multi-layer neural network model based on a hierarchical attention network architecture, used to dynamically calculate the weight allocation of each dimension of anomaly detection according to sea area characteristics and seasonal characteristics. Current sea area water depth data refers to the seabed depth measurement value at the quality control processing location. Geographic latitude data refers to the geographic latitude coordinate value of the quality control processing location. Seasonal temperature difference data refers to the historical average temperature change amplitude value corresponding to the season of the quality control processing time. Hierarchical fusion weight parameter refers to the weight allocation coefficient when fusing features of different abstract hierarchical levels in the ocean parameter adaptive weight optimization model. Dynamic weight values ​​refer to the weight coefficients output by the adaptive weight optimization model for ocean parameters, used for weighted fusion of multi-dimensional anomaly signals. Fuzzy comprehensive evaluation operation rules refer to the mathematical operation methods that use fuzzy mathematics theory to establish a fuzzy relation matrix and calculate the comprehensive evaluation results through matrix operations and weight vectors. The comprehensive anomaly index refers to the numerical index, calculated using the fuzzy comprehensive evaluation operation rules, characterizing the overall degree of anomaly in marine meteorological data. Data anomaly refers to the state judgment result of marine meteorological observation data deviating from normal variation patterns.

[0037] The multi-sensor cross-validation mechanism refers to a verification method that uses observation data from multiple independent sensors for cross-comparison and verification to determine the authenticity of abnormal data. Redundant sensor observation data refers to marine meteorological observation data collected by multiple sensors deployed in the same or adjacent locations for cross-validation. The sensor health assessment function is a mathematical function model used to evaluate the sensor's operating status and data reliability. Sensor output data variance refers to the statistical variance of the sensor's output data under stable environmental conditions. The data drift trend coefficient refers to the degree of systematic shift in sensor output data over time. The cross-validation consistency index is a quantitative evaluation index of the consistency between observation data from multiple sensors. Sensor operating time refers to the cumulative operating time of the sensor from deployment to the current moment. The environmental corrosion assessment value is a quantitative evaluation value of the degree of corrosion impact of the marine environment on sensor equipment. The sensor reliability score is a score value representing the reliability of the sensor's operating status, calculated and output by the sensor health assessment function. The virtual sensor data reconstruction algorithm is a data reconstruction algorithm that uses mathematical operations to infer missing or abnormal parameter values ​​based on the correlation between parameters. Normal parameter values ​​refer to reliable marine meteorological parameter observation values ​​that have passed quality control verification in the cross-parameter correlation network structure. The inverse computation equation of the correlation network refers to a system of mathematical equations established based on known normal parameter values ​​and correlation rules to solve for estimated values ​​of outlier parameters. The estimated values ​​of outlier parameters are the inferred parameter values ​​obtained by solving the inverse computation equation of the correlation network, used to replace outlier data. The repaired marine meteorological parameter data refers to the complete marine meteorological dataset output after processing by the virtual sensor data reconstruction algorithm. The quality control process refers to the complete quality control process for marine meteorological data from anomaly detection to data repair.

[0038] The sensor health assessment function is used to calculate the operational reliability score of the marine meteorological sensor. The inputs include the sensor output data variance, data drift trend coefficient, cross-validation consistency index, sensor operating time, and environmental corrosion degree assessment value. The output is the sensor reliability score between 0 and 1. When the sensor reliability score is lower than 0.6, it indicates that the sensor needs maintenance and calibration.

[0039] The specific structure of the ocean parameter adaptive weight optimization model is a multi-layer neural network model based on a hierarchical attention network architecture, which includes four main components: an input layer, a feature extraction layer, an attention mechanism layer, and an output layer. The attention mechanism layer uses a multi-level attention structure to assign weights to different sea area features and seasonal features. The hierarchical fusion weight parameters of the model are dynamically adjusted according to three key parameters: the current sea area water depth data, the geographical latitude data, and the seasonal temperature difference data. The weight allocation ratio of the current sea area water depth data affecting the stability of the marine environment, the weight allocation ratio of the geographical latitude data affecting climate features, and the weight allocation ratio of the seasonal temperature difference data affecting time changes are all specified. The steps for establishing the training dataset for the ocean parameter adaptive weight optimization model specifically include: collecting ocean meteorological observation data from different sea areas and seasons worldwide as the basic data source; performing quality screening and outlier removal on the collected raw data to form a clean dataset; classifying and labeling the data according to sea area geographical location, water depth information, and seasonal time to establish a multi-dimensional label system; using the outlier data identification results labeled by ocean meteorological experts as the target labels for supervised learning; dividing the dataset into training, validation, and test sets in an 8:1:1 ratio; and standardizing all input features to ensure the stability of model training. The training steps for the ocean parameter adaptive weight optimization model specifically include: iteratively updating the model parameters using a stochastic gradient descent optimization algorithm; setting the learning rate to 0.001 and dynamically adjusting the learning rate using a cosine annealing scheduling strategy; using the cross-entropy loss function to measure the difference between the model's predicted weights and the true weights; setting the batch size to 64 and performing a complete training process of 200 training cycles; evaluating the model performance using a validation set and saving the optimal model parameters after each training cycle; employing an early stopping mechanism during training to prevent overfitting; and finally verifying the model's generalization ability and practical application effect using a test set.

[0040] The hierarchical weight adjustment function is used to adjust the hierarchical fusion weight parameters of the ocean parameter adaptive weight optimization model. The hierarchical weight adjustment function calculates a comprehensive adjustment factor value based on four data points: current sea area depth data, geographic latitude data, seasonal temperature difference data, and data quality assessment score. When the comprehensive adjustment factor value is in the range of 0 to 0.25, a weight allocation strategy dominated by deep features is used to adjust the hierarchical fusion weight parameters of the model. When the comprehensive adjustment factor value is in the range of 0.25 to 0.5, a weight allocation strategy balancing mid-level features is used to adjust the hierarchical fusion weight parameters of the model. When the comprehensive adjustment factor value is in the range of 0.5 to 0.75, a weight allocation strategy prioritizing shallow features is used to adjust the hierarchical fusion weight parameters of the model. When the comprehensive adjustment factor value is in the range of 0.75 to 1.0, a weight allocation strategy enhancing surface features is used to adjust the hierarchical fusion weight parameters of the model.

[0041] The specific implementation methods of the above steps are described in detail below.

[0042] The specific implementation of step S01 involves first using an iterative regression algorithm to statistically analyze at least five years of historical marine meteorological data. A piecewise correlation mathematical model between air temperature and air pressure is established using least squares regression analysis. When the air temperature is above 25℃, the correlation coefficient is set to a negative value, with the absolute value controlled within the range of 0.5 to 0.7. The final correlation coefficient and applicable temperature range are determined through more than 1000 iterations of optimization. The purpose of this step is to establish a reliable benchmark rule for the correlation between air temperature and air pressure. Next, the Pearson correlation coefficient calculation method is used to perform linear regression analysis on the observed data of relative humidity and air temperature, establishing a linear correlation rule between humidity and air temperature. The correlation coefficient threshold is set between 0.6 and 0.8. Then, based on fluid dynamics principles, a nonlinear fitting algorithm is used to establish a square root function relationship between wind speed and air pressure gradient. The function parameters are determined through statistical analysis of measured wind speed data and concurrent air pressure gradient data, with the correlation coefficient threshold set between 0.4 and 0.6. Finally, six core marine meteorological parameters, including air temperature, air pressure, humidity, wind speed, sea temperature, and salinity, were constructed into a network topology. The parameter nodes were connected through quantified correlation coefficients to form a complete cross-parameter correlation network model.

[0043] The specific implementation of step S02 involves using a sliding window algorithm with a time window length of 30 minutes to perform real-time correlation analysis on each parameter in the cross-parameter correlation network structure. The sliding window moves 5 minutes at a time interval to ensure the continuity and real-time nature of data processing. Within each time window, the Pearson correlation coefficient calculation method and the Spearman rank correlation coefficient calculation method are used to perform correlation analysis on the parameter pairs to obtain real-time correlation coefficient values. The calculated real-time correlation coefficients are compared with the standard correlation coefficients corresponding to the temperature-pressure correlation rules, humidity-temperature linear correlation rules, and wind speed-pressure gradient square root correlation rules established in step S01. When the deviation exceeds 25%, a correlation anomaly identification signal is triggered. The purpose of this step is to identify potential data anomalies by monitoring the changes in correlation between parameters in real time. The 25% deviation threshold is a reasonable limit determined based on the natural variation range of the marine environment and measurement errors.

[0044] The specific implementation of step S03 involves establishing a four-dimensional anomaly signal detection mechanism to identify and quantify different types of data anomalies. First, a parameter threshold exceeding detection mechanism is established. Based on historical statistical data, reasonable variation ranges for each marine meteorological parameter are determined: temperature range is set to -5℃ to 45℃, air pressure range to 980 hPa to 1040 hPa, and wind speed range to 0 to 50 m / s. When parameter values ​​exceed these ranges, the degree of exceeding the limits is calculated, and a parameter threshold exceeding intensity value is generated. Second, a spatiotemporal change rate anomaly detection mechanism is established. Anomalies are identified by calculating the rate of change of parameters in the time dimension and the gradient change in the spatial dimension. The time change rate threshold is set to no more than 10% of the parameter average per hour, and the spatial gradient change threshold is set to no more than twice the parameter standard deviation per kilometer. Finally, a probability density distribution shift detection mechanism is established. The chi-square test and Kolmogorov-Smirnov test are used to compare the current data distribution with the historical normal distribution. When the test statistic exceeds a significance level of 0.05, it is determined to be a distribution shift anomaly. Finally, a correlation verification failure detection mechanism is established. When the correlation between parameters does not conform to the correlation rules established in step S01, a correlation verification failure signal is generated. The purpose of this step is to comprehensively detect data anomalies from multiple perspectives and improve the accuracy and reliability of anomaly identification.

[0045] The specific implementation of step S04 involves calculating the dynamic weight values ​​of anomalous signals in each dimension using an adaptive weight optimization model for ocean parameters. This model is built upon a hierarchical attention network architecture and comprises four main components: an input layer, a feature extraction layer, an attention mechanism layer, and an output layer. The input layer receives three key environmental parameters: current sea area depth data, geographic latitude data, and seasonal temperature difference data. The feature extraction layer uses a multilayer perceptron network to perform nonlinear transformation and abstraction of the input features. The attention mechanism layer uses a multi-level attention structure to calculate the weight allocation for different sea area features and seasonal features. The model adjusts the weight allocation ratio for ocean environmental stability based on the current sea area depth data, setting the weight for relatively stable deep-sea environments at 0.7 and for shallow-sea environments with greater environmental changes at 0.3. It also adjusts the weight allocation ratio for climate features based on geographic latitude data, setting the weight for drastic climate change in low-latitude regions at 0.8 and for relatively stable climate in high-latitude regions at 0.4. Finally, it adjusts the weight allocation ratio for temporal changes based on seasonal temperature difference data, setting the weight for seasons with large temperature differences at 0.9 and for seasons with small temperature differences at 0.5. The purpose of this step is to dynamically adjust the weight allocation of anomaly detection based on the specific marine environmental characteristics, thereby improving the adaptability and accuracy of the quality control method.

[0046] The specific implementation of step S05 involves using a fuzzy comprehensive evaluation calculation method to perform weighted fusion processing on multi-dimensional abnormal signals. First, a fuzzy relation matrix is ​​established, using the intensity values ​​of parameter threshold exceeding limits, spatiotemporal change rate anomalies, probability density distribution shifts, and correlation verification failures as evaluation factors. A triangular membership function is used to convert each intensity value into a fuzzy value between 0 and 1. Then, the dynamic weight values ​​calculated in step S04 are used to construct a weight vector. The comprehensive evaluation result is calculated through matrix operations and the weight vector. Specifically, the calculation process involves multiplying the anomaly intensity values ​​of each dimension with their corresponding dynamic weight values, and then summing all the product results to obtain a comprehensive anomaly index. When the comprehensive anomaly index exceeds a preset threshold of 0.7, it is determined to be a data anomaly. This threshold is the optimal judgment boundary determined based on statistical analysis of numerous practical application cases. The purpose of this step is to comprehensively consider anomaly information from multiple dimensions, avoiding the limitations of single-dimensional judgment and improving the accuracy and robustness of anomaly identification.

[0047] The specific implementation of step S06 involves initiating a multi-sensor cross-validation mechanism for identified data anomalies. This involves comparing and analyzing observation data from multiple redundant sensors deployed in the same or adjacent locations. First, an analysis of variance (ANOVA) method is used to compare the consistency of observation data from different sensors. A sensor malfunction is identified when the standard deviation of the data difference between sensors exceeds twice the measurement accuracy. Then, a sensor health assessment function is called to evaluate the sensor's operating status. This function takes five parameters as input: sensor output data variance, data drift trend coefficient, cross-validation consistency index, sensor operating time, and environmental corrosion assessment value. The data variance threshold is set to 1.5 times the normal operating value, the data drift trend coefficient threshold is set to no more than 0.1% per month, the cross-validation consistency index threshold is set to above 0.8, and reliability begins to decline when the sensor's operating time exceeds 80% of its design life. The environmental corrosion level is assessed based on a comprehensive evaluation of seawater salinity and temperature. A weighted average method is used to calculate the sensor reliability score. A score below 0.6 indicates that the sensor requires maintenance and calibration. The purpose of this step is to determine the authenticity of abnormal data through multi-sensor data cross-validation, avoiding misjudgments caused by sensor malfunctions and improving the reliability of quality control results.

[0048] The specific implementation of step S07 involves repairing the confirmed anomalous data using a virtual sensor data reconstruction algorithm. This algorithm establishes a mathematical model based on the parameter correlation relationships in a cross-parameter correlation network structure. First, normal parameter values ​​related to the anomalous parameters are extracted from the cross-parameter correlation network structure. Then, based on the correlation rules established in step S01 (temperature and pressure correlation, humidity and temperature linear correlation, and wind speed and pressure gradient square root correlation), a set of inverse calculation equations for the correlation network is constructed. The least squares method and the Newton-Raphson iterative method are used to solve the equations to obtain estimated values ​​of the anomalous parameters. To ensure the rationality of the reconstructed data, the estimation results are constrained and checked, requiring that the reconstructed values ​​be within the historical statistical range and that their correlation with the relevant parameters conforms to physical laws. Finally, the reconstructed parameter values ​​replace the original anomalous data, outputting a repaired complete marine meteorological parameter dataset. The purpose of this step is to deduce reasonable values ​​of anomalous parameters through the inherent correlation relationships between parameters, ensuring the integrity and continuity of the data and providing a reliable data foundation for subsequent marine meteorological analysis.

[0049] The detailed structure of the ocean parameter adaptive weight optimization model comprises four main components: an input layer, a feature extraction layer, an attention mechanism layer, and an output layer. The input layer receives three key environmental parameters: current sea area depth, geographic latitude, and seasonal temperature range. Each parameter is first normalized to a value between 0 and 1. The feature extraction layer employs a three-layer fully connected neural network structure: the first layer contains 64 neurons, the second layer contains 32 neurons, and the third layer contains 16 neurons. ReLU activation is used for non-linear transformation between each layer. The attention mechanism layer uses a multi-head attention mechanism structure with four attention heads, each with a dimension of 16. It calculates the association weights between different input features through a self-attention mechanism. The output layer uses a softmax activation function to output the weight allocation coefficients for anomaly detection in four dimensions. The weights sum to 1, and each weight value is between 0 and 1.

[0050] The detailed steps for establishing the training dataset for this model first involved collecting marine meteorological observation data from different sea areas and seasons worldwide as the basic data source. Data sources included buoy observation stations, oceanographic survey vessels, satellite remote sensing data, and coastal observation stations, covering a time span of at least 10 years. Next, the collected raw data underwent quality screening and outlier removal. The principle was to remove obviously anomalous observations and use time series continuity tests to remove records with excessive missing data, resulting in a clean base dataset. Then, based on geographical location, the data was categorized into three types: tropical, temperate, and polar seas; based on water depth, into deep, mid-deep, and shallow seas; and based on season, into four types: spring, summer, autumn, and winter, establishing a multi-dimensional labeling system. Next, marine meteorological experts were invited to manually annotate the data, including anomaly identification results and corresponding weight allocation suggestions, serving as target labels for supervised learning. Finally, the dataset was randomly divided into training, validation, and test sets in an 8:1:1 ratio, and all input features were standardized to ensure the stability of model training.

[0051] The reason why the adaptive weight optimization model for marine parameters is suitable for solving the technical problem of this invention is that the complexity and variability of the marine environment require quality control methods to have dynamic adaptability. Traditional fixed-weight quality control methods cannot adapt to the special characteristics of different marine environments and seasonal changes, and are prone to misjudgment or omission. The adaptive weight optimization model of this invention automatically learns the optimal weight allocation strategy under different environmental conditions through deep learning technology, and can adjust the weight ratio of anomaly detection in each dimension in real time according to the current water depth, latitude and seasonal characteristics of the sea area. Compared with existing fixed-weight methods based on expert experience, this model has stronger generalization ability and adaptability, and can handle complex and variable marine environmental conditions. Compared with dynamic weight adjustment technology based on simple statistical methods, this model can more accurately capture the complex correlation between different environmental factors through attention mechanism, and achieve more precise weight allocation. The advantage of this model is that it can automatically learn the optimal weight allocation pattern from a large amount of historical data, avoid the subjectivity and limitations of manual weight setting, improve the objectivity and accuracy of quality control methods, and provide a more intelligent and precise solution for marine meteorological data quality control.

[0052] The key technical ideas of this invention mainly include four aspects: cross-parameter correlation network construction, multi-dimensional anomaly signal detection mechanism, marine parameter adaptive weight optimization model, and virtual sensor data reconstruction algorithm. The cross-parameter correlation network construction technology overcomes the limitations of traditional single-parameter quality control methods by establishing a mathematical correlation model between marine meteorological parameters. It can improve the accuracy of anomaly identification by utilizing the inherent physical relationships between parameters, and has stronger logical consistency and reliability compared to existing independent parameter verification methods. The multi-dimensional anomaly signal detection mechanism comprehensively analyzes four dimensions: parameter threshold exceeding limits, spatiotemporal rate of change anomalies, probability density distribution shift, and correlation verification failure. Compared to traditional single-dimensional or dual-dimensional detection methods, it can more comprehensively identify various types of data anomalies, effectively reducing the false negative and false positive rates. The marine parameter adaptive weight optimization model uses deep learning technology to achieve dynamic weight adjustment. Compared to fixed weight methods, it can optimize the detection strategy according to specific environmental conditions, improving the adaptability and accuracy of the quality control method under different sea areas and seasonal conditions. The virtual sensor data reconstruction algorithm repairs anomaly data based on the parameter correlation network, maintaining the physical consistency and logical rationality of the data compared to simple interpolation or replacement methods. The synergistic effect of these technological approaches has formed a complete intelligent marine meteorological data quality control system. Through the organic combination of parameter correlation analysis, multi-dimensional detection, dynamic weight optimization and intelligent data reconstruction, it has achieved a technological leap from passive detection to proactive prevention, from single judgment to comprehensive evaluation, and from fixed strategies to adaptive optimization compared with existing technologies, significantly improving the intelligence level and practical effect of marine meteorological data quality control.

[0053] It should be noted that this invention also solves the following technical problem: Existing technologies suffer from inaccurate health status assessments of marine meteorological sensors, leading to difficulties in determining data reliability. Traditional sensor status monitoring methods primarily rely on single data variance analysis or simple threshold comparisons to determine the sensor's operating status. This method cannot comprehensively reflect the actual health level of the sensor in complex marine environments. Especially when facing harsh conditions such as high salinity, high humidity, and strong corrosiveness in the marine environment, the sensor's performance gradually degrades with prolonged operation, and traditional methods struggle to accurately assess this gradual performance degradation process. This invention establishes a sensor health assessment function that comprehensively considers five key parameters: sensor output data variance, data drift trend coefficient, cross-validation consistency index, sensor operating time, and environmental corrosion assessment value. It employs a multi-dimensional fusion approach to calculate the sensor reliability score, accurately reflecting the sensor's true operating status in the marine environment. This provides reliable sensor status information for data quality control and effectively solves the technical problem of inaccurate sensor health status assessment.

[0054] Furthermore, this invention addresses the technical problem in existing technologies where the accuracy of abnormal marine meteorological data repair is low, leading to a lack of data integrity and continuity. Traditional data repair methods typically employ simple interpolation algorithms or mean-filling methods to process missing or abnormal marine meteorological data. These methods do not fully consider the physical correlations between marine meteorological parameters, and the repaired data often does not conform to the actual changing patterns of the marine environment, resulting in a decline in data quality. This invention, through a virtual sensor data reconstruction algorithm, uses the normal parameter values ​​in a cross-parameter correlation network and established correlation rules to solve for the estimated values ​​of abnormal parameters using the inverse calculation equation of the correlation network. This repair method based on the physical correlations between parameters ensures that the repaired data conforms to the principles of marine meteorology, maintains the physical consistency and logical rationality of the data, significantly improves the accuracy and reliability of abnormal data repair, and effectively guarantees the integrity and continuity of marine meteorological data.

[0055] Specifically, the principle of this invention is as follows: This invention can solve the technical problem of insufficient accuracy in identifying multi-parameter correlations in marine meteorological data quality control. Its fundamental principle lies in establishing a multi-dimensional anomaly detection system based on a cross-parameter correlation network. This system fully considers the inherent physical correlations and dynamic changes among marine meteorological parameters. First, it establishes correlation rules for air temperature and air pressure, linear correlation rules for humidity and air temperature, and square root correlation rules for wind speed and air pressure gradients through iterative regression analysis of historical marine meteorological data. These rules reflect the objective physical relationships between meteorological elements in the marine environment, providing a scientific benchmark for anomaly detection. Second, it uses a sliding window algorithm to calculate the correlation coefficients between each parameter in real time. Anomalies are determined by the degree of deviation from the standard correlation coefficient. This dynamic monitoring mechanism can promptly detect abnormal data that violates the normal correlation between parameters. Third, it establishes a multi-dimensional anomaly detection mechanism encompassing four dimensions: parameter threshold exceeding limits, spatiotemporal rate of change anomalies, probability density distribution shifts, and correlation verification failures. By quantifying the anomaly intensity values ​​of each dimension, it comprehensively assesses the degree of data anomaly, avoiding the limitations of single-dimensional judgment. Finally, an adaptive weight optimization model for marine parameters based on a hierarchical attention network architecture was adopted. This model dynamically adjusts the weight allocation of each dimension according to environmental characteristics such as water depth, latitude, and seasonal temperature differences. A comprehensive anomaly index is calculated using fuzzy comprehensive evaluation rules, enabling adaptive identification of parameter correlations under different marine environmental conditions. This technical solution aligns with the principles of marine meteorology, accurately reflects the coordinated changes in various meteorological elements within the marine environment, and effectively improves the accuracy and reliability of anomaly data detection.

[0056] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0057] The specific implementation of step S01 is to construct a cross-parameter correlation network model using an iterative regression algorithm. The correlation rules between air temperature and air pressure are expressed by a piecewise correlation function as follows:

[0058] ;

[0059] In the formula, The correlation coefficient between air temperature and air pressure is dimensionless. This is the temperature value, in °C. The linear coefficients are shown in °C. and ℃ ; The coefficient of the constant term is dimensionless; The coefficient of the quadratic term is given in °C. ; The error correction term is dimensionless, ranging from -0.02 to 0.02 and from -0.03 to 0.03, respectively. The linear correlation rule between humidity and temperature is expressed as: In the formula, The humidity-temperature correlation coefficient is dimensionless. The linear slope coefficient is expressed in °C. ; The intercept constant is dimensionless. The random error term is dimensionless and ranges from -0.05 to 0.05. The square root correlation rule for wind speed and air pressure gradients is expressed as: In the formula, is the correlation coefficient between wind speed and air pressure gradient, which is dimensionless; This represents the pressure gradient, measured in Pa / km. This is the square root coefficient, with units of (Pa / km). ; The offset constant is dimensionless. This is a nonlinear error term, dimensionless, ranging from -0.08 to 0.08.

[0060] The parameter acquisition method is as follows: The method is obtained through least squares iterative regression, including step 1: collecting at least 5 years of historical temperature and pressure observation data; step 2: performing more than 1000 iterations of optimization using the gradient descent algorithm; and step 3: determining the optimal parameter combination through cross-validation. The Pearson correlation coefficient was obtained using the following steps: Step 1: Statistical analysis of the linear relationship between humidity and temperature observation data; Step 2: Fitting the straight line parameters using a linear regression algorithm. The nonlinear fitting algorithm is used to obtain the data, including step 1: calculating the scatter distribution of wind speed data and the square root of the pressure gradient; step 2: using the Levenberg-Marquardt algorithm to estimate the nonlinear parameters. The method uses statistical environmental data, including step 1: collecting historical environmental change data of the marine area; and step 2: calculating the environmental stability index as a correction factor.

[0061] The specific implementation of step S02 is to use a sliding window algorithm to calculate the real-time correlation coefficient. The formula for calculating the real-time correlation coefficient is as follows:

[0062] ;

[0063] In the formula, for The real-time correlation coefficient at any given time, dimensionless; For the first The parameter observation value at each moment, with the unit depending on the specific parameter; This is the average value of the parameters within the sliding window, with the same unit as the corresponding parameter. The sliding window length is set to 6 data points. The index is for the current time. The trigger condition for the correlation anomaly flag is: In the formula, is the corresponding standard correlation coefficient, which is dimensionless.

[0064] The specific implementation of step S03 is to establish a multi-dimensional abnormal signal detection mechanism, and the numerical calculation formula for the parameter threshold over-limit intensity is as follows:

[0065] ;

[0066] In the formula, The value of the intensity exceeding the parameter threshold is dimensionless. The units for the observed parameter values ​​depend on the parameter type. This represents the parameter threshold boundary, with the same units as the observed parameters. This represents the normal range of parameter variation, and the units are the same as those for the observed parameters. The environmental correction factor is dimensionless and ranges from 0.8 to 1.2. The numerical formula for calculating the spatiotemporal rate of change anomaly intensity is: In the formula, The numerical value of the spatiotemporal rate of change anomaly intensity is dimensionless. This is the rate of change of the parameter over time, expressed in units of the parameter per hour. This is the rate of change of the parameter space, expressed in parameter units per kilometer. Spatial coordinates; The spatial weighting coefficient is dimensionless and ranges from 0.5 to 1.5. The numerical formula for calculating the probability density distribution migration intensity is: In the formula, The value of the probability density distribution shift intensity is dimensionless. The probability density function of the current data; The probability density function for historical normal data; The parameter is a variable that takes values. The numerical formula for calculating the failure strength of the correlation check is: In the formula, This is a dimensionless numerical value representing the failure strength of the correlation check. Number of parameter pairs; For the first Real-time correlation coefficients of parameter pairs; For the first The expected correlation coefficient of each parameter pair; For the first The importance weights of each parameter pair are dimensionless.

[0067] The parameter acquisition method is as follows: The method involves obtaining geographic information through analysis, including step 1: analyzing the topographic and geomorphological characteristics of the sea area; and step 2: determining the spatial weighting coefficients based on the complexity of the sea area. The method of obtaining the parameters by means of expert evaluation includes step 1: inviting marine meteorological experts to score the importance of the parameters; and step 2: statistically processing the expert scores to obtain the weight coefficients.

[0068] The specific implementation of step S04 is to execute an adaptive weight optimization algorithm. The dynamic weight value calculation adopts the output of the ocean parameter adaptive weight optimization model, and the hierarchical weight adjustment function is expressed as follows:

[0069] ;

[0070] In the formula, It is a dynamic weight vector, dimensionless; This is the current sea depth data, in meters (m). This is geographic latitude data, in degrees; This data represents seasonal temperature variations, expressed in °C. The data quality assessment score is dimensionless and ranges from 0 to 1. This is the neural network mapping function. The formula for calculating the comprehensive adjustment factor is: In the formula, The value of the comprehensive adjustment factor is dimensionless; The maximum reference water depth is set at 6000m; The maximum seasonal temperature difference is set to 30℃. The formula for adjusting the hierarchical fusion weight parameters is as follows: In the formula, The adjusted hierarchical fusion weight parameters are dimensionless. These are the basic weight parameters, which are dimensionless. The adjustment coefficient is dimensionless and ranges from -0.5 to 0.5.

[0071] The specific implementation of step S05 involves applying fuzzy comprehensive evaluation calculation processing. The formula for calculating the comprehensive anomaly index is as follows:

[0072] ;

[0073] In the formula, This is a comprehensive anomaly index, dimensionless. This is the transpose of the dynamic weight vector; This is the anomalous intensity vector; For the first Dynamic weight values ​​for each dimension, dimensionless; For the first The anomaly intensity values ​​in each dimension are dimensionless, where... , , , The fuzzy relation matrix is ​​represented as:

[0074] ;

[0075] In the formula, For the first The evaluation factor for the first The membership degree of each evaluation level is dimensionless. The triangular membership function is expressed as: In the formula, The membership function value of a triangle is dimensionless. For input variables; These are the left boundary, peak point, and right boundary parameters of the trigonometric function.

[0076] The specific implementation of step S06 is to perform multi-sensor cross-validation processing, and the sensor health assessment function is expressed as:

[0077] ;

[0078] In the formula, The sensor reliability score is dimensionless. The variance of the sensor output data is expressed in units that vary depending on the sensor type. This is a dimensionless coefficient representing the data drift trend. This is a cross-validation consistency metric, dimensionless. The sensor's operating time, in hours; The sensor's design lifespan is expressed in hours. This is a dimensionless value used to assess the degree of environmental corrosion. The weighting coefficients are dimensionless and satisfy the following conditions: The specific values ​​are 0.2, 0.25, 0.3, 0.15, and 0.1, respectively. The formula for calculating the ANOVA statistic is: In the formula, This is a dimensionless statistic for analysis of variance. Mean square between groups; Within-group mean square; Number of sensor groups; For the first Number of sensors in a group; For the first Average value observed by the group of sensors; This represents the overall observed mean. For the first Group 1 Observations from individual sensors; This represents the total number of sensors.

[0079] The parameter acquisition method is as follows: The data was obtained using statistical methods, including step 1: collecting the sensor's output data for 24 consecutive hours under stable conditions; and step 2: calculating the sample variance of the data sequence. The trend analysis method is used to obtain the data, including step 1: performing linear regression analysis on the historical output data of the sensor; step 2: calculating the slope of the regression line as the drift trend coefficient. The correlation analysis method is used to obtain the correlation coefficient, which includes step 1: calculating the Pearson correlation coefficient between the observation data of the target sensor and the redundant sensor; and step 2: averaging the correlation coefficients of multiple sensors. The data was obtained through experimental measurement, including step 1: periodically checking the corrosion level of the sensor housing; and step 2: calculating the corrosion level assessment value based on the corrosion area ratio, using the following formula: ,in For the corrosion area, This represents the total surface area of ​​the sensor. This represents the corrosion sensitivity coefficient of the material.

[0080] The specific implementation of step S07 is to perform abnormal data repair and reconstruction processing. The inverse calculation equations of the correlation network are expressed as follows:

[0081] ;

[0082] In the formula, The parameters to be solved are temperature, humidity, and wind speed, respectively, in ℃, %, and m / s. The target air pressure value is expressed in Pa. The values ​​are known normal humidity and wind speed observations, in % and m / s, respectively. Let be the correlation coefficient between each parameter pair, which is dimensionless. The estimated values ​​of outlier parameters are obtained using Cramer's rule. In the formula, These are estimated values ​​for abnormal temperature parameters, in °C. It is a coefficient matrix; This is the matrix obtained by replacing the first column of the coefficient matrix with a constant vector; This refers to matrix determinant operations. The formula for calculating the matrix determinant is: In the formula, For matrix The Middle Line number The elements of the column.

[0083] The principle and effect of each formula are explained below. Piecewise correlation function of air temperature and air pressure. Based on the physical principle of the nonlinear relationship between air temperature and air pressure in marine meteorology, this paper accurately describes the correlation variation law under different temperature ranges through piecewise function form, where a linear relationship is used in the low temperature range. To reflect a stable thermodynamic state, a quadratic function is used in the high-temperature range. This model describes complex nonlinear thermal convection processes and, compared to traditional single linear correlation models, more accurately reflects the complex thermodynamic processes in the marine environment, significantly improving the accuracy and applicability of correlation modeling. Real-time correlation coefficient calculation formula. The sliding window Pearson correlation coefficient algorithm is used, through the numerator part Calculate the covariance between parameters, denominator part Standardized processing is performed, and real-time changes in the relationship between parameters are captured through dynamic time windows. Compared with static correlation analysis methods over fixed time periods, this approach can promptly detect abnormal fluctuations in parameter relationships, providing a sensitive and reliable basis for real-time quality control. Parameter threshold exceedance intensity calculation formula. Combined with the degree of parameter deviation and environmental correction factors By quantifying the degree of exceeding limits and considering environmental impacts, an accurate assessment of anomaly intensity can be achieved. Compared to simple limit judgment methods, this approach provides a more refined quantification of anomaly severity, offering a reliable numerical basis for subsequent comprehensive evaluation. The formula for the anomaly intensity of spatiotemporal change rate is as follows: Based on partial differential equation theory, considering the variation characteristics of parameters in time and space dimensions, gradient calculation is used. By capturing the dynamic change patterns of parameters, this method can comprehensively reflect the spatiotemporal evolution of the marine environment compared to traditional single-dimensional rate of change detection methods, and effectively identify various types of dynamic anomalies. (Probability density distribution migration intensity formula) Using integral distance This method measures the degree of difference between the current data distribution and the historical normal distribution. Based on statistical principles, it achieves a quantitative comparison of distribution characteristics. Compared with simple mean or variance comparison methods, it can comprehensively reflect the overall changes in data distribution, providing a rigorous mathematical foundation for statistical anomaly detection. The comprehensive anomaly index calculation formula is as follows: Through weighted linear combination This method achieves the fusion of multi-dimensional anomaly information and, combined with a dynamic weight optimization strategy, adjusts the importance of each dimension based on environmental characteristics. Compared to fixed-weight comprehensive evaluation methods, it can adapt to the detection needs under different sea areas and seasonal conditions, significantly improving the accuracy and adaptability of comprehensive judgment. Sensor health assessment function. Based on the multi-factor comprehensive evaluation theory, through the exponential function and The exponential decay law describing sensor performance, the reciprocal function The linear term reflects the negative impact of variance on reliability. This approach directly reflects the contribution to consistency and, compared to single-indicator health assessment methods, comprehensively reflects the sensor's operational status, providing a reliable basis for weight allocation in cross-validation. The inverse calculation equations of the correlation network are based on linear algebra theory and the parameter correlation network structure. Through matrix operations, it reconstructs and calculates abnormal parameters. The coefficient matrix reflects the correlation relationships between parameters, the constant vector contains known parameter information, and the Cramer's rule is used for the solution process. By using the ratio of determinants to ensure the uniqueness of the solution, compared with simple interpolation or mean replacement methods, physical consistency between parameters can be maintained, ensuring that the reconstructed data conforms to the laws of marine meteorology, and providing a scientific and reliable means of repairing data integrity.

[0084] To better understand and implement this invention, the following is a specific application scenario example 2: A technical team deployed multiple sets of marine meteorological observation equipment in a certain sea area and needed to perform real-time quality control processing on the collected marine meteorological data. The technical team adopted a marine meteorological data quality control method based on cross-parameter correlation networks to perform quality control processing on 30 consecutive days of observation data in July 2024.

[0085] The technical team first constructed a correlation network model for marine meteorological parameters and used an iterative regression algorithm to analyze and process historical marine meteorological data of the sea area over the past three years. Through 1200 iterations of optimization calculations, the correlation rule between air temperature and air pressure was determined: when the air temperature is above 25℃, the correlation coefficient is -0.62, applicable to the temperature range of 25℃ to 35℃. Simultaneously, a linear correlation rule between humidity and air temperature was established, and the correlation coefficient was determined to be 0.78 using the Pearson correlation coefficient calculation method. The square root correlation rule for wind speed and air pressure gradients was established based on fluid dynamics principles, and the function parameters were determined through nonlinear fitting analysis. A cross-parameter correlation network structure is formed, which includes six core parameters: air temperature, air pressure, humidity, wind speed, wind direction, and sea surface temperature. The parameter nodes are connected by correlation coefficients.

[0086] During the real-time data processing phase, the technical team implemented a sliding window correlation coefficient calculation, employing a 30-minute sliding window algorithm to calculate the correlation coefficients of each parameter in the cross-parameter correlation network structure in real time. At 10:30 AM on July 15th, the calculated real-time correlation coefficients for temperature and air pressure were -0.48, deviating from the standard correlation coefficient of -0.62 by 22.6%, but not exceeding the 25% threshold. However, at 11:00 AM, the real-time correlation coefficient changed to -0.42, deviating by 32.3%, exceeding the 25% threshold, triggering a correlation anomaly flag.

[0087] The technical team established a multi-dimensional anomaly signal detection mechanism, identifying and quantifying four dimensions separately. The detection results at 11:00 showed: the observed temperature was 33.8℃, exceeding the upper limit of the normal range of 32.5℃, generating a parameter threshold exceedance intensity value of 0.85. The wind speed abruptly changed from 8.2 m / s to 15.7 m / s, a change rate of 91.5%, exceeding the reasonable range by 40%, generating a spatiotemporal change rate anomaly intensity value of 0.92. The probability density distribution of humidity data deviated by 0.31 from the historical normal distribution pattern, generating a probability density distribution deviation intensity value of 0.76. In the correlation verification, the real-time correlation coefficient between temperature and humidity was 0.52, deviating from the standard value of 0.78 by 33.3%, generating a correlation verification failure intensity value of 0.88.

[0088] The technical team implemented an adaptive weight optimization algorithm, calculating dynamic weight values ​​through an adaptive weight optimization model for ocean parameters. Inputs included current sea area depth data of 2850m, geographic latitude data of 18.5°, and seasonal temperature range data of 12.3℃. The model processed these values ​​using a hierarchical attention network architecture. Figure 3 As shown, the attention mechanism layer assigns weights to different features: water depth data has a weight of 0.35 for affecting marine environmental stability, latitude data has a weight of 0.28 for affecting climate features, and seasonal temperature difference has a weight of 0.37 for affecting temporal variation. The calculated dynamic weight values ​​are as follows: parameter threshold exceeding limit 0.24, spatiotemporal change rate anomaly 0.31, probability density distribution offset 0.22, and correlation verification failure 0.23.

[0089] In the fuzzy comprehensive evaluation calculation and processing stage, the technical team performed a weighted fusion calculation on the anomaly intensity values ​​and dynamic weight values. A fuzzy relation matrix was established using the fuzzy comprehensive evaluation calculation rules, and the calculated comprehensive anomaly index was 0.85, exceeding the preset threshold of 0.7, thus indicating data anomaly. The hierarchical weight adjustment function, based on current sea area depth data, geographic latitude data, seasonal temperature difference data, and data quality assessment scores, calculated a comprehensive adjustment factor value of 0.68, falling within the range of 0.5 to 0.75. A weight allocation strategy prioritizing shallow features was adopted to adjust the hierarchical fusion weight parameters of the model.

[0090] The technical team implemented multi-sensor cross-validation processing, initiating the multi-sensor cross-validation mechanism for identified data anomalies. Three redundant sensors, numbered A01, A02, and A03, were deployed in this sea area. The results of the cross-comparison analysis are shown in Table 1.

[0091] Table 1. Sensor Cross-Validation Comparison Data Table

[0092]

[0093] Comparative analysis of redundant sensor data revealed anomalies in the temperature and wind speed data from sensor A01. The technical team then used a sensor health assessment function to evaluate each sensor, as shown in Table 2.

[0094] Table 2 Sensor Health Assessment Results

[0095]

[0096] Sensor A01 has a reliability score of 0.58, which is below the threshold of 0.6, indicating that maintenance and calibration are required. Sensors A02 and A03 have reliability scores of 0.89 and 0.87 respectively, indicating that they are in good working order.

[0097] The technical team performed abnormal data repair and reconstruction, employing a virtual sensor data reconstruction algorithm for the confirmed abnormal data from sensor A01. The input included the normal parameter values ​​of sensors A02 and A03 in the cross-parameter correlation network structure, as well as the correlation rules for air temperature and air pressure, the linear correlation rule for humidity and air temperature, and the square root correlation rule for wind speed and air pressure gradient. The estimated values ​​of the abnormal parameters were obtained by solving the inverse calculation equations of the correlation network: the repaired air temperature was 29.4℃, and the wind speed was 9.0m / s.

[0098] During 30 consecutive days of quality control processing, the system detected 246 data anomalies, including 89 instances of parameter threshold exceeding limits, 73 instances of spatiotemporal change rate anomalies, 52 instances of probability density distribution deviation anomalies, and 32 instances of correlation verification failure anomalies. Multi-sensor cross-validation confirmed 198 genuine anomalies, resulting in a false alarm rate of 19.5%. The virtual sensor data reconstruction algorithm successfully repaired 192 anomalies, achieving a repair success rate of 97.0%. Relevant charts and graphs from the implementation process are shown below. Figures 2-4 As shown.

[0099] This invention represents a significant advancement over traditional quality control methods. Traditional methods primarily rely on threshold judgments for single parameters and simple statistical tests, lacking analysis of inter-parameter correlations and prone to false alarms and missed alarms. This invention, by constructing a cross-parameter correlation network model, fully leverages the physical relationships between marine meteorological parameters, enabling the identification of complex anomalies that are difficult to detect using traditional methods. The multi-dimensional anomaly signal detection mechanism comprehensively evaluates data quality from four perspectives: parameter thresholds, spatiotemporal variations, probability distribution, and correlation verification, providing more comprehensive anomaly identification capabilities. The adaptive weight optimization algorithm dynamically adjusts the detection strategy based on sea area and seasonal characteristics, avoiding the insufficient adaptability issues associated with fixed weights. The multi-sensor cross-validation mechanism effectively reduces the false alarm rate and improves the accuracy of anomaly identification. The virtual sensor data reconstruction algorithm infers anomaly data based on inter-parameter correlations, ensuring data continuity and integrity, and providing a reliable data foundation for subsequent marine environmental analysis.

[0100] It should be noted that the variables involved in this invention are explained in detail in Tables 3 and 4.

[0101] Table 3. Variable Explanation Table (Part 1)

[0102]

[0103] Table 4. Variable Explanation Table (Part Two)

[0104]

[0105] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for quality control of marine meteorological data based on cross-parameter correlation networks, characterized in that, This includes constructing a marine meteorological parameter correlation network model to obtain correlation rules for air temperature and pressure, linear correlation rules for humidity and air temperature, and square root correlation rules for wind speed and pressure gradients, forming a cross-parameter correlation network structure; implementing sliding window correlation coefficient calculation to calculate the correlation coefficient of each parameter in the cross-parameter correlation network structure in real time, triggering a correlation anomaly flag when the real-time correlation coefficient deviates from the standard correlation coefficient by more than a preset threshold; establishing a multi-dimensional anomaly signal detection mechanism to identify and quantify signals of parameter threshold exceeding limits, spatiotemporal change rate anomalies, probability density distribution shifts, and correlation verification failures, generating corresponding intensity values; and executing an adaptive weight optimization algorithm to calculate dynamic weight values ​​through a marine parameter adaptive weight optimization model. Fuzzy comprehensive evaluation calculation is applied to perform weighted fusion calculation on intensity values ​​and dynamic weight values ​​to calculate a comprehensive anomaly index. When the comprehensive anomaly index exceeds a preset threshold, it is judged as data anomaly. Multi-sensor cross-validation is implemented to determine the authenticity of data anomalies and calculate sensor reliability scores. Anomaly data repair and reconstruction is performed by using a virtual sensor data reconstruction algorithm to solve the estimated values ​​of anomaly parameters through the inverse calculation equation of the correlation network, and outputting repaired marine meteorological parameter data. The steps for constructing a marine meteorological parameter correlation network model are as follows: First, an iterative regression algorithm is used to analyze and process historical marine meteorological data to obtain the correlation rules between air temperature and air pressure. Second, linear correlation rules between humidity and air temperature and square root correlation rules between wind speed and air pressure gradient are established to form a cross-parameter correlation network structure containing 6 sets of core parameters. In the cross-parameter correlation network structure, each parameter node is connected through a correlation coefficient. Specifically, the temperature-pressure correlation rule is derived by using least squares regression analysis and polynomial fitting algorithms to iteratively calculate a piecewise negative correlation between temperature and pressure from a large amount of historical marine meteorological observation data. The final correlation coefficient and applicable temperature range are determined through more than 1000 iterations of optimization. The humidity-temperature linear correlation rule is derived by using a linear regression algorithm to statistically analyze the observation data of relative humidity and temperature in the marine environment to obtain a linear correlation. The correlation coefficient is determined using the Pearson correlation coefficient calculation method. The wind speed-pressure gradient square root correlation rule is derived by establishing a square root function relationship between wind speed and pressure gradient based on fluid dynamics principles. The function parameters are determined through nonlinear fitting analysis of measured wind speed data and concurrent pressure gradient data. Specifically, generating the corresponding intensity values ​​involves generating the corresponding parameter threshold exceeding intensity values, spatiotemporal change rate anomaly intensity values, probability density distribution offset intensity values, and correlation verification failure intensity values. Specifically, the step of executing the adaptive weight optimization algorithm involves calculating dynamic weight values ​​for the intensity of parameter threshold exceeding limits, the intensity of spatiotemporal change rate anomalies, the intensity of probability density distribution shifts, and the intensity of correlation verification failures through the ocean parameter adaptive weight optimization model. The ocean parameter adaptive weight optimization model adjusts the hierarchical fusion weight parameters of the model based on the current sea area depth data, geographic latitude data, and seasonal temperature difference data.

2. The marine meteorological data quality control method based on cross-parameter correlation networks according to claim 1, characterized in that, The aforementioned cross-parameter correlation network structure is specifically a network topology model based on the principles of marine meteorology, which establishes the interconnection between multiple meteorological parameters. It describes the coordinated change patterns of various meteorological elements in the marine environment by quantifying the mathematical relationships between different parameters.

3. The marine meteorological data quality control method based on cross-parameter correlation networks according to claim 2, characterized in that, The steps for implementing the sliding window correlation coefficient calculation are as follows: a sliding window algorithm with a time window length of 30 minutes is used to calculate the correlation coefficient of each parameter in the cross-parameter correlation network structure in real time. When the calculated real-time correlation coefficient deviates by more than 25% from the standard correlation coefficient corresponding to the temperature and pressure correlation rule, the humidity and temperature linear correlation rule, and the wind speed and pressure gradient square root correlation rule, a correlation anomaly indicator is triggered.

4. The marine meteorological data quality control method based on cross-parameter correlation networks according to claim 3, characterized in that, The steps of applying fuzzy comprehensive evaluation calculation are as follows: the numerical values ​​of the parameter threshold exceeding the limit, the numerical values ​​of the spatiotemporal change rate anomaly, the numerical values ​​of the probability density distribution deviation, and the numerical values ​​of the correlation verification failure are weighted and fused with the dynamic weight value. The comprehensive anomaly index is calculated according to the fuzzy comprehensive evaluation calculation rules. When the comprehensive anomaly index exceeds the preset threshold of 0.7, it is determined to be a data anomaly.

Images