Monitoring methods and systems for Deep Argo buoys

Abstract

This invention discloses a monitoring method and system for Deep Argo buoys, belonging to the field of marine environmental monitoring technology. The method includes: acquiring and preprocessing three types of data: buoy cabin environment, buoy attitude control, and ocean physical observations; constructing a coupled state model of the cabin environment to capture the nonlinear coupling and time-series dependence between environmental parameters to identify anomalies; constructing a buoy motion physics model to calculate theoretical depth and identify buoy attitude control anomalies; combining theoretical depth with ocean physical observation data for cross-domain verification to distinguish whether the source of anomalies in the ocean physical observation data is the buoy itself or actual ocean physical signals; fusing multi-source information to complete anomaly detection and fault diagnosis, and generating a buoy health status report. The Deep Argo buoy monitoring method and system provided by this invention can accurately identify anomalies and trace the source of data anomalies, ensuring the reliability of monitoring data and the stability of buoy operation.

Description

Technical Field

[0001] This invention belongs to the field of marine environmental monitoring technology, and in particular relates to a monitoring method and system for Deep Argo buoys. Background Technology

[0002] To conduct in-depth research on scientific issues such as global ocean circulation, thermohaline transport, and climate change, it is necessary to deploy DeepArgo buoys on a large scale to obtain comprehensive deep-sea environmental parameters. These buoys need to withstand hydrostatic pressures of up to 60 MPa and near-zero temperatures in extreme deep-sea environments. Their long-term, wide-range autonomous operation mode places stringent requirements on buoy reliability, data acquisition accuracy, and system stability.

[0003] The existing Deep Argo buoy monitoring technology has significant drawbacks: First, it relies on a single sensor to monitor the cabin pressure, which can only identify obvious faults such as cabin damage, ignoring the complex coupling relationship between cabin temperature, humidity, and air pressure. This leads to the failure to detect early, hidden faults such as minor leaks and overheating of electronic components in a timely manner. Second, it has not established a fine physical correlation model between buoy operating parameters and depth, making it impossible to identify nonlinear deviations in oil volume and depth, affecting the accuracy of depth measurement and the accuracy of temperature, salinity, and depth data. Third, it is difficult to distinguish the source of anomalies in oceanographic observation data, and it is impossible to determine whether the anomaly is due to a buoy malfunction or a genuine oceanographic signal, reducing the reliability of the observation data.

[0004] Therefore, there is an urgent need for a DeepArgo buoy monitoring solution that integrates multi-source parameters, constructs coupling and physical models, and accurately distinguishes the sources of anomalies, in order to overcome the limitations of existing technologies. Summary of the Invention

[0005] In view of the shortcomings of the related technologies, the purpose of this invention is to provide a monitoring method and system for Deep Argo buoys to solve the problems mentioned in the background.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A monitoring method for Deep Argo buoys includes the following steps: S1. Acquire environmental parameters, buoy attitude control parameters, and ocean physical observation data inside the buoy cabin, and perform preprocessing. The buoy's internal environmental parameters include internal air pressure, internal temperature, and internal humidity data; the buoy attitude control parameters include buoyancy pump fuel volume, pump motor current, buoy's own depth, tilt angle, and heading angle data; and the oceanographic observation data include external environmental temperature, external environmental salinity, and external depth data. S2. Based on the preprocessed buoy cabin environmental parameters, construct a coupled state model of the cabin environment to capture the nonlinear coupling relationship and temporal dependence between cabin air pressure data, cabin temperature data and cabin humidity data, so as to identify abnormal information of the buoy cabin environment. S3. Based on the preprocessed buoy attitude control parameters, construct a physical model of buoy motion and calculate the theoretical depth of the Deep Argo buoy to identify abnormal information in buoy attitude control. S4. Based on theoretical depth and ocean physical observation data, cross-domain consistency verification is carried out. Combining the abnormal information identified by the in-cabin environment coupled state model and the buoy motion physical model, the source of the abnormality in the ocean physical observation data is distinguished as either abnormality in the buoy's own state or real ocean physical signals. S5. Based on the abnormal information of the buoy cabin environment, the abnormal information of buoy attitude control, the cross-domain consistency verification results, and the differentiated sources of abnormal ocean physical observation data, multi-source anomaly detection and fault diagnosis are performed on the Deep Argo buoy. S6. Based on the results of multi-source anomaly detection and fault diagnosis, generate a buoy health status early warning and diagnosis report that includes a comprehensive score of buoy health status and fault early warning information.

[0007] In some embodiments, step S2 specifically includes: S21. Integrate the preprocessed cabin air pressure data, cabin temperature data and cabin humidity data into a multivariable time series input vector sequence; S22. The Transformer network model is used to learn the multivariate time-series input vector sequence to capture the nonlinear coupling relationship and time-series dependence between cabin air pressure data, cabin temperature data and cabin humidity data; S23. Construct a coupled state model of the cabin environment based on the learning results; S24. Calculate the differences between the predicted values ​​of cabin air pressure, cabin temperature, and cabin humidity output by the cabin environment coupled state model and the corresponding actual observed values, so as to quantify the degree of deviation of the buoy cabin environment parameters and identify abnormal information of the buoy cabin environment.

[0008] In some embodiments, step S3 specifically includes: S31. Input the pre-processed buoyancy pump oil volume data, pump motor current data, buoy depth data, tilt angle data, and heading angle data; S32. Based on the principles of fluid mechanics, and combining the geometric parameters, material density parameters, and real-time seawater density data of the Deep Argo buoy, establish a mathematical model of the buoyancy and resistance of the Deep Argo buoy at different depths. S33. Adaptive Kalman filtering algorithm or unscented Kalman filtering algorithm is used to integrate buoyancy and drag mathematical model and real-time acquired buoy attitude control parameters to construct buoy motion physical model in order to calculate the theoretical depth of Deep Argo buoy; S34. Compare the theoretical depth with the actual depth data of the buoy itself to confirm the deviation of the buoy's kinematic characteristics and identify abnormal information in the buoy's attitude control.

[0009] In some embodiments, in step S33, the calculation formula for the buoy motion physical model is as follows:

[0010] in, The net force acting on the Deep Argo buoy. For the total mass of the Deep Argo buoy, It is the acceleration due to gravity. For the fixed displacement volume of the Deep Argo buoy, The change in displacement volume corresponding to the buoyancy adjustment of the Deep Argo buoy. The coefficient of viscous drag. This is the pressure drag coefficient. This refers to the descent speed of the Deep Argo buoy.

[0011] In some embodiments, in step S33, the theoretical depth is:

[0012] in, The theoretical depth of the Deep Argo buoy predicted by the model; This is a nonlinear function used to map input parameters to theoretical depth; This is the current buoyancy pump oil level data for the Deep Argo buoy; Pump motor current data for Deep Argo buoys; The real-time seawater density data at the depth where the Deep Argo buoy is located is obtained by calculating the external environmental temperature and salinity data from the ocean physical observation data through the equation of state. This is the set of internal parameters of the model.

[0013] In some embodiments, step S4 specifically includes: S41. Based on historical oceanographic observation data under normal operating conditions of Deep Argo buoys, establish an observation data baseline model to characterize the typical features and spatiotemporal variation patterns of temperature profiles, salinity profiles, and depth profiles in oceanographic observation data. S42. Compare the actual collected external environmental temperature data, external environmental salinity data, and external depth data with the observation data baseline model to detect abnormal points or abnormal profiles in the marine physical observation data and complete the cross-domain consistency verification. S43. When anomalies are detected in ocean physical observation data, the abnormal information identified by the coupled state model of the cabin environment and the buoy motion physical model is combined, and a fusion decision-making mechanism based on rule reasoning and probabilistic graphical model is adopted to distinguish whether the source of the anomaly in the ocean physical observation data is the buoy's own abnormal state or a real ocean physical signal.

[0014] In some embodiments, in step S43, the fusion decision-making mechanism based on rule-based reasoning and probabilistic graphical models uses the following calculation formula:

[0015] in, The probability that an anomaly in the observation data is caused by a real ocean physical signal when anomalies in the observation data, anomalies in the cabin environment, and anomalies in the buoy attitude control occur simultaneously. These are real ocean physical signals; Anomalies were detected in the marine physical observation data; This refers to abnormal information regarding the environment inside the buoy compartment; This refers to abnormal information related to buoy attitude control. During calculation, if or and If they are highly synchronized in time and satisfy the correlation of physical mechanisms, then it is determined that... The source is an abnormal state of the buoy itself; if and None of them occurred, and If it consistently does not conform to historical climatic characteristics, then it is judged The source is a real ocean physical signal.

[0016] In some embodiments, step S5 specifically includes: S51. The abnormal information of the buoy cabin environment, the abnormal information of buoy attitude control, the cross-domain consistency verification results, and the differentiated sources of abnormal ocean physical observation data are weighted and fused to form a unified multi-dimensional feature vector. S52. Input the multi-dimensional feature vector into the pre-trained classification model, which is a multi-class support vector machine model or a decision tree model. S53. Combining the historical failure mode library, the potential failure types of the buoy are classified and diagnosed through a classification model. Potential failure types include slow leakage of the hull, overheating of internal electronic components, performance degradation of the buoyancy pump, minor leakage of oil circuits, and pseudo-anomalies in observation data caused by biofouling.

[0017] In some embodiments, in step S1, the preprocessing specifically includes timestamp synchronization, data alignment, missing value filling, outlier filtering, and feature normalization of the buoy cabin environmental parameters, buoy attitude control parameters, and ocean physical observation data.

[0018] A monitoring system for Deep Argo buoys, applied to the aforementioned monitoring method for Deep Argo buoys, the monitoring system comprising: The data acquisition module is used to acquire environmental parameters inside the buoy compartment, buoy attitude control parameters, and ocean physical observation data. The environmental parameters inside the buoy compartment include internal air pressure data, internal temperature data, and internal humidity data. The buoy attitude control parameters include buoyancy pump fuel volume data, pump motor current data, buoy depth data, tilt angle data, and heading angle data. The ocean physical observation data includes external environmental temperature data, external environmental salinity data, and external depth data. The data preprocessing module is connected to the data acquisition module. The data preprocessing module is used to perform time stamp synchronization, data alignment, missing value filling, outlier filtering and feature normalization preprocessing on the environmental parameters, attitude control parameters and ocean physical observation data in the buoy cabin. The in-cabin environment coupled state modeling module is connected to the data preprocessing module. The in-cabin environment coupled state modeling module is used to construct an in-cabin environment coupled state model based on the preprocessed buoy in-cabin environmental parameters, and capture the nonlinear coupling relationship and time-series dependence between in-cabin air pressure data, in-cabin temperature data and in-cabin humidity data, so as to identify abnormal information of the buoy in-cabin environment. The buoy motion physics modeling module is connected to the data preprocessing module. The buoy motion physics modeling module is used to build a buoy motion physics model and calculate the theoretical depth of the Deep Argo buoy in order to identify abnormal information in the buoy attitude control. The cross-domain data consistency verification and anomaly differentiation module is connected to the data preprocessing module, the in-cabin environment coupled state modeling module, and the buoy motion physics modeling module. The cross-domain data consistency verification and anomaly differentiation module is used to perform cross-domain consistency verification based on theoretical depth and ocean physical observation data. Combining the anomaly information identified by the in-cabin environment coupled state model and the buoy motion physics model, it distinguishes whether the source of the anomaly in the ocean physical observation data is the buoy's own state anomaly or a real ocean physical signal. The multi-source anomaly detection and fault diagnosis module is connected to the in-cabin environment coupled state modeling module, the buoy motion physics modeling module, and the cross-domain data consistency verification and anomaly differentiation module, respectively. The multi-source anomaly detection and fault diagnosis module is used to perform multi-source anomaly detection and fault diagnosis on the Deep Argo buoy based on anomaly information of the buoy's in-cabin environment, anomaly information of the buoy's attitude control, cross-domain consistency verification results, and the differentiated sources of anomalies in the ocean physical observation data. The status warning and reporting module is connected to the multi-source anomaly detection and fault diagnosis module. The status warning and reporting module is used to generate a buoy health status warning and diagnosis report that includes a comprehensive score of buoy health status and fault warning information based on the results of multi-source anomaly detection and fault diagnosis.

[0019] In some embodiments, the data acquisition module includes a MEMS barometric pressure sensor, a platinum resistance temperature sensor, a capacitive humidity sensor, an oil level sensor, a pump motor current sensor, and a CTD sensor.

[0020] Compared with the prior art, the beneficial effects of the present invention are: 1. The monitoring method and system for Deep Argo buoys provided by this invention integrate multi-source heterogeneous information such as buoy cabin environmental parameters, buoy attitude control parameters, and ocean physical observation data to achieve three-dimensional perception of the complex operating state of the buoy; at the same time, by establishing a dynamic nonlinear coupling relationship model between cabin air pressure, temperature, and humidity, it can accurately identify early hidden fault precursors that are difficult to detect due to complex parameter correlations in related technologies, and significantly improve the sensitivity and accuracy of fault early warning.

[0021] 2. The Deep Argo buoy monitoring method and system provided by this invention constructs a buoy motion physical model based on fluid dynamics principles, and combines an adaptive Kalman filter algorithm to fuse multi-source buoy operating parameters to achieve accurate estimation of the buoy's theoretical depth. It can effectively identify kinematic deviations caused by buoyancy pump performance degradation, minor oil circuit leaks, and abnormal buoy attitude, thereby improving the accuracy of attitude control anomaly detection.

[0022] 3. The Deep Argo buoy monitoring method and system provided by this invention can efficiently distinguish between false anomalies in observation data caused by changes in the buoy's own state and real ocean physical signals by comparing the theoretical depth calculated by the model with the actual observation depth and combining the correlation analysis of temperature and salinity profile data quality, thus greatly improving the reliability and efficiency of Deep Argo buoy observation data. Attached Figure Description

[0023] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating an embodiment of the monitoring method and system for the Deep Argo buoy of the present invention. Figure 2 This is a structural principle block diagram of an embodiment of the monitoring method and system for the Deep Argo buoy of the present invention. Detailed Implementation

[0024] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0025] In the description of this invention, it should be understood that the terms "center", "lateral", "longitudinal", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0026] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0027] Example 1: See appendix Figures 1 to 2 This paper presents an illustrative embodiment of the monitoring method for the Deep Argo buoy proposed in this invention. The monitoring method for the Deep Argo buoy includes the following steps: S1. Acquire environmental parameters, buoy attitude control parameters, and ocean physical observation data inside the buoy cabin, and perform preprocessing. The buoy's internal environmental parameters include internal air pressure, internal temperature, and internal humidity data; the buoy attitude control parameters include buoyancy pump fuel volume, pump motor current, buoy's own depth, tilt angle, and heading angle data; and the oceanographic observation data include external environmental temperature, external environmental salinity, and external depth data. S2. Based on the preprocessed buoy cabin environmental parameters, construct a coupled state model of the cabin environment to capture the nonlinear coupling relationship and temporal dependence between cabin air pressure data, cabin temperature data and cabin humidity data, so as to identify abnormal information of the buoy cabin environment. S3. Based on the preprocessed buoy attitude control parameters, construct a physical model of buoy motion and calculate the theoretical depth of the Deep Argo buoy to identify abnormal information in buoy attitude control. S4. Based on theoretical depth and ocean physical observation data, cross-domain consistency verification is carried out. Combining the abnormal information identified by the in-cabin environment coupled state model and the buoy motion physical model, the source of the abnormality in the ocean physical observation data is distinguished as either abnormality in the buoy's own state or real ocean physical signals. S5. Based on the abnormal information of the buoy cabin environment, the abnormal information of buoy attitude control, the cross-domain consistency verification results, and the differentiated sources of abnormal ocean physical observation data, multi-source anomaly detection and fault diagnosis are performed on the Deep Argo buoy to identify the potential fault types of the Deep Argo buoy. S6. Based on the results of multi-source anomaly detection and fault diagnosis, generate a buoy health status early warning and diagnosis report that includes a comprehensive score of buoy health status and fault early warning information.

[0028] By integrating environmental parameters inside the buoy cabin, buoy attitude control parameters, and ocean physical observation data, a coupling correlation model and a physical correlation model among multiple parameters are established, thereby enabling consistency detection, anomaly identification, and fault diagnosis of buoy operation parameters.

[0029] In step S1, environmental parameters inside the buoy cabin, buoy attitude control parameters, and ocean physical observation data are acquired, specifically including the following sub-steps: S101. Acquire environmental parameters inside the buoy chamber, including internal air pressure, temperature, and humidity data. The Deep Argo buoy is equipped with a high-precision MEMS pressure sensor, a platinum resistance temperature sensor, and a capacitive humidity sensor. The acquired raw analog signals are first converted into digital signals by a high-resolution analog-to-digital converter, then undergo preliminary data calibration and encapsulation by the buoy's internal microcontroller, and finally transmitted to the buoy's main control unit via the internal bus.

[0030] S102. Acquire buoy attitude control parameters, including buoy oil level and pump motor current. The Deep Argo buoy achieves depth control through its internal hydraulic pump system. This system precisely controls the transfer of hydraulic oil between the inner and outer oil bladders to adjust the buoy's net buoyancy. The buoy oil level sensor measures the remaining hydraulic oil in the inner oil bladder in real time with microliter-level accuracy. The pump motor current sensor monitors the actual operating current of the DC motor driving the hydraulic pump. This current value reflects the pump's load and operating efficiency. These data are digitized and timestamped before being packaged and sent by the buoy's main control unit. These parameters are used to evaluate the buoy's depth control capability, buoyancy pump health, and predict the buoy's remaining mission life.

[0031] S103. Acquire marine physical observation data, including depth data, temperature profile data, and salinity profile data. The Deep Argo buoy is equipped with a high-precision conductivity-temperature-depth sensor. During the buoy's descent and ascent, the sensor synchronously collects seawater depth, temperature, and conductivity data at a high frequency.

[0032] In step S1, preprocessing specifically includes timestamping, data alignment, missing value imputation, outlier filtering, and feature normalization of the buoy cabin environmental parameters, buoy attitude control parameters, and ocean physical observation data. In this embodiment, all collected data are timestamped to ensure temporal consistency of data from different sensors; all collected data are aligned to unify data from different sampling frequencies to a preset sampling frequency; missing values ​​are imputed using a temporal interpolation algorithm to fill in data gaps; outlier filtering is performed to identify and remove isolated data points that deviate from the normal range; and feature normalization is performed to convert data with different dimensions to a unified numerical range.

[0033] Step S2 performs multi-parameter coupling consistency detection on the environmental parameters inside the buoy chamber to identify abnormalities in the chamber environment. Specifically, by analyzing the inherent physical coupling relationship between the three parameters—indoor air pressure data, indoor temperature data, and indoor humidity data—subtle anomalies that are not easily detected by a single parameter can be identified.

[0034] Step S2 specifically includes: S21, processing the pre-processed cabin pressure data. Cabin temperature data and cabin humidity data Integrate into a multivariable time-series input vector sequence S22. A Transformer network model is used to learn the multivariate time-series input vector sequence to capture the nonlinear coupling relationship and temporal dependence between the cabin air pressure data, cabin temperature data, and cabin humidity data, and to extract features from the vector sequence; S23. A cabin environment coupling state model is constructed based on the learning results; S24. The differences between the predicted cabin air pressure, predicted cabin temperature, and predicted cabin humidity values ​​output by the cabin environment coupling state model and the corresponding actual observed values ​​are calculated to quantify the degree of deviation of the buoy cabin environment parameters and identify abnormal information in the buoy cabin environment.

[0035] The Transformer network model includes a multi-head self-attention mechanism, the calculation formula of which is:

[0036] Where Q, K, and V are the query vector, key vector, and value vector, respectively. It is a vector dimension. Through this mechanism, the model can capture the nonlinear coupling characteristics of cabin pressure, temperature, and humidity over long time scales, and the model output is the predicted value of the environmental parameters for the next time step. By calculating the residual vector Furthermore, the Mahalanobis distance is introduced to quantify the deviation between the actual observation point and the predicted distribution. If the deviation exceeds the preset confidence interval threshold, it is determined that there is abnormal coupling in the cabin environment.

[0037] The in-cabin environment coupled state model is used to record the dynamic correlation patterns of in-cabin air pressure, temperature, and humidity under normal operating conditions. The monitoring system first extracts a large number of in-cabin environmental parameter sequences from historical buoy data confirmed to be in normal operating conditions, constructing a multivariate time series dataset. Based on this dataset, deep learning models such as vector autoregressive moving average models, long short-term memory neural networks, or variational autoencoders are used to learn the linear and nonlinear dependencies, hysteresis effects, and periodic patterns among these three parameters. The model's input is the air pressure, temperature, and humidity sequences within a past time window, and the output is the predicted values ​​of these parameters and their uncertainty range for the next time step. The model is dynamically optimized on normal datasets to ensure it accurately captures the normal behavioral patterns of in-cabin environmental parameters.

[0038] After determining the coupled state model of the cabin environment, a coupling consistency index is calculated between parameters. This model is used to predict the actual observed cabin environmental parameters at the current moment, and the deviation between the predicted and actual observed values ​​is calculated to quantify the coupling consistency between parameters. Specifically, for each monitoring moment, the monitoring system inputs the actual observation data from a time window prior to the current moment into the coupled state model of the cabin environment to obtain predicted values ​​for cabin air pressure, cabin temperature, and cabin humidity at the next moment. Subsequently, these predicted values ​​are compared with the actual values ​​collected at the same moment. The consistency index is calculated using the Mahalanobis distance to comprehensively reflect the degree to which multiple variables deviate from the normal coupling pattern. When the Mahalanobis distance... The larger the value, the greater the deviation of the current observation data from the normal coupling pattern, and the worse the consistency between parameters. This indicator can effectively identify situations where a single parameter change is not significant, but the overall coupling relationship has become abnormal, such as a slight decrease in air pressure due to a minor leak but a significant increase in humidity. The formula for calculating the distance to Maharanobis is:

[0039] in, This represents a vector of currently observed cabin environmental parameters, including air pressure, temperature, and humidity. The mean vector of parameters predicted by the coupled state model of the cabin environment; The covariance matrix representing the prediction error of the parameters is calculated using historical normal data. This covariance matrix reflects the correlation between the prediction errors of each parameter under normal operating conditions.

[0040] Finally, the system determines whether there are any anomalies in the cabin environment based on the consistency index. After calculating the coupling consistency index for each monitoring moment, the monitoring system compares it with a preset anomaly threshold, which is determined based on the statistical distribution of consistency indices from historical normal operation data. If the coupling consistency index at the current moment exceeds this threshold, an anomaly is determined to exist in the cabin environment. For example, when the Mahalanobis distance is significantly greater than the normal fluctuation range, even if the change in a single parameter is not significant, it indicates that the internal coupling relationship has been disturbed, which may indicate an early failure. In addition, to improve the robustness of the judgment, a strategy can be adopted to trigger an anomaly alarm only when the consistency index exceeds the threshold for multiple consecutive moments (such as three consecutive moments), in order to reduce false alarms caused by sporadic measurement noise.

[0041] Step S3 involves performing a physical correlation consistency test between the buoy attitude control parameters and the depth data in the ocean physical observation data to identify buoy attitude control anomalies. This involves establishing a refined physical correlation model to detect problems such as decreased buoyancy pump performance, minor oil leaks, or buoy attitude anomalies, thereby ensuring the accuracy of depth measurements and the reliability of the data.

[0042] Step S3 specifically includes: S31, inputting preprocessed buoyancy pump oil volume data, pump motor current data, buoy depth data, tilt angle data, and heading angle data; S32, based on fluid dynamics principles, combined with the geometric parameters, material density parameters, and real-time seawater density data of the Deep Argo buoy, establishing a mathematical model of buoyancy and drag at different depths for the Deep Argo buoy; S33, using an adaptive Kalman filter algorithm or an unscented Kalman filter algorithm, fusing the buoyancy and drag mathematical model with the real-time acquired buoy attitude control parameters, constructing a physical model of buoy motion to calculate the theoretical depth of the Deep Argo buoy; S34, comparing the theoretical depth with the actual collected buoy depth data to confirm deviations in the buoy's kinematic characteristics and identify abnormal information in the buoy attitude control.

[0043] A physical correlation model between buoy attitude control parameters and depth data is constructed, namely, the buoy motion physics model, to quantify the physical relationship between buoy fuel level, pump motor current, and the depth reached by the buoy under a specific seawater density profile. Utilizing the principles of buoy kinematics, combined with fluid dynamics and buoyancy, a nonlinear regression model is constructed. The model's input includes time-series data of the current buoy fuel level and pump motor current, as well as the current seawater density profile information (which can be indirectly calculated from temperature and salinity profiles). The model's output is the theoretical depth the buoy should reach. The buoy motion physics model is fitted and optimized using real observation data from historical buoy dives and ascents, such as employing a Kalman filter and particle filter fusion algorithm, to achieve adaptive learning and accurate estimation of the buoy's dynamic parameters.

[0044] In step S33, the calculation formula for the buoy motion physical model is as follows:

[0045] in, The net force acting on the Deep Argo buoy. For the total mass of the Deep Argo buoy, It is the acceleration due to gravity. For the fixed displacement volume of the Deep Argo buoy, The change in displacement volume corresponding to the buoyancy adjustment of the Deep Argo buoy. The coefficient of viscous drag. This is the pressure drag coefficient. This refers to the descent speed of the Deep Argo buoy.

[0046] In step S33, the theoretical depth is:

[0047] in, The theoretical depth of the Deep Argo buoy predicted by the model; This is a nonlinear function used to map input parameters to theoretical depth; This is the current buoyancy pump oil level data for the Deep Argo buoy; Pump motor current data for Deep Argo buoys; The real-time seawater density data at the depth where the Deep Argo buoy is located is obtained by calculating the external environmental temperature and salinity data from the ocean physical observation data through the equation of state. This is a set of internal parameters for the model, including the buoy's own mass, volume, compressibility, rate of change of inner and outer bladder volume, pump efficiency, and drag coefficient, which are optimized during model training.

[0048] The consistency deviation between the model's predicted depth and the actual observed depth is calculated. Specifically, after constructing the buoy's motion physical model, at each monitoring moment, the current buoy oil level, pump motor current, and real-time seawater density profile are input into the buoy's motion physical model to calculate the theoretical depth the buoy should reach in the current state. This theoretical depth is then compared with the depth measured by the Deep Argo buoy's actual sensors to calculate the consistency deviation. The deviation can be calculated using indicators such as absolute error, relative error, or root mean square error. If there is a persistent and significant difference between the actual depth and the theoretical depth, and this difference exceeds the measurement error range of the sensor and the prediction error range of the model, it indicates that there may be an anomaly in the buoy's attitude control or depth measurement. This deviation may be caused by various factors such as buoyancy pump failure, oil leakage, external biological adhesion altering the buoy's hydrodynamic characteristics, or even depth sensor drift.

[0049] Finally, the deviation is used to determine whether there is an anomaly in the buoy attitude control. The calculated consistency deviation is compared with a preset anomaly threshold, which is determined based on the statistical distribution of model prediction errors in historical normal operation data. To improve the accuracy of the judgment, the duration and trend of the deviation (such as whether the deviation increases gradually or appears suddenly) and the working stage of the buoy (diving, surfacing, or stationary) can be combined for a comprehensive judgment. In this way, nonlinear deviations between oil volume and depth can be identified, providing a basis for early warning of potential buoyancy pump failures or inaccurate depth measurements.

[0050] Step S4 performs adaptive anomaly identification on ocean physical observation data and distinguishes between buoy's own state anomalies and real ocean physical signals. Specifically, it achieves accurate identification and source differentiation of observation data anomalies by establishing a baseline model and combining the anomaly detection results of the in-cabin environment coupled state model and the buoy motion physical model.

[0051] Step S4 specifically includes: S41, establishing an observation data baseline model based on historical oceanographic observation data under normal operating conditions of the Deep Argo buoy, to characterize the typical features and spatiotemporal variation patterns of temperature, salinity, and depth profiles in the oceanographic observation data; S42, comparing the actually collected external environmental temperature, salinity, and depth data with the observation data baseline model to detect anomalies or anomalies in the oceanographic observation data and complete cross-domain consistency verification; S43, when anomalies are detected in the oceanographic observation data, combining the anomaly information identified by the in-cabin environment coupled state model and the buoy motion physics model, adopting a fusion decision-making mechanism based on rule-based reasoning and probabilistic graphical models to distinguish whether the source of the anomalies in the oceanographic observation data is an anomaly in the buoy's own state or a real oceanographic signal.

[0052] The observational data baseline model is used to collect typical characteristics and spatiotemporal variation patterns of ocean temperature and salinity profiles collected by Deep Argo buoys under normal operating conditions. Utilizing a global ocean observation database and historical Deep Argo buoy data, a regional ocean temperature, salinity, and climatology model is constructed as the observational data baseline model. This model can predict the temperature and salinity profiles that should be observed within a specific geographical location, season, and depth range, and provides the range of uncertainty in the prediction. Furthermore, this baseline model includes an adaptive evaluation mechanism for data quality, dynamically adjusting the sensitivity of anomaly detection based on historical data distribution to adapt to the complex and variable marine environment.

[0053] After constructing the observation data baseline model, the actual temperature and salinity profile data collected by the Deep Argo buoy are compared with the established observation data baseline model to identify outliers or abnormal profiles in the data, thereby detecting anomalies in the observation data and completing cross-domain consistency verification.

[0054] By combining the abnormalities in the buoy cabin environment and the abnormalities in buoy attitude control, the source of the abnormalities in the observation data is distinguished. After initially identifying the abnormalities in the observation data, the abnormal information in the buoy cabin environment (such as abnormal increase in cabin humidity) and the abnormal information in buoy attitude control (such as the deviation of oil quantity from the consistency of depth) are used as auxiliary judgment criteria to distinguish the source of the abnormalities in the observation data. That is, a fusion decision-making mechanism based on rule reasoning and probabilistic graphical model is adopted.

[0055] In step S43, the fusion decision-making mechanism based on rule-based reasoning and probabilistic graphical models adopts the following calculation formula:

[0056] in, The probability that an anomaly in the observation data is caused by a real ocean physical signal when anomalies in the observation data, anomalies in the cabin environment, and anomalies in the buoy attitude control occur simultaneously. These are real ocean physical signals; Anomalies were detected in the marine physical observation data; This refers to abnormal information regarding the environment inside the buoy compartment; This refers to abnormal information related to buoy attitude control. In actual calculations, if both internal environmental anomalies and buoy attitude control anomalies are detected simultaneously, and these internal anomalies are highly synchronized in time with the observed data anomalies and correlated in physical mechanisms, the observed data anomalies are tended to be attributed to anomalies in the buoy's own state and judged as false anomaly signals. Conversely, if all internal state parameters are normal, but the observed data consistently exhibits characteristics significantly inconsistent with historical climatic states, it is tended to be judged as a genuine ocean physical signal, that is, if... or and If they are highly synchronized in time and satisfy the correlation of physical mechanisms, then it is determined that... The source is an abnormal state of the buoy itself; if and None of them occurred, and If it consistently does not conform to historical climatic characteristics, then it is judged The source is a real ocean physical signal.

[0057] Step S5 specifically includes: S51, weighted fusion of abnormal information from the buoy's internal environment, abnormal information from the buoy's attitude control, cross-domain consistency verification results, and differentiated sources of abnormal oceanographic observation data, integrating them into a unified multidimensional feature vector; S52, inputting the multidimensional feature vector into a pre-trained classification model, which is either a multi-class support vector machine model or a decision tree model; S53, combining a historical fault mode library, classifying and diagnosing potential fault types of the buoy through the classification model, including slow leakage of the hull, overheating of internal electronic components, performance degradation of the buoy pump, minor leaks in the oil circuit, and pseudo-anomalies in observation data caused by biofouling.

[0058] In step S51, when fusing the anomaly detection results from various sources, the fusion weights are dynamically adjusted based on the confidence level of each anomaly detection module, the prior probability of historical failure modes, and the severity of their impact on buoy operation. The fusion process employs evidence theory, Bayesian networks, or multilayer perceptron models to integrate discrete or continuous anomaly information from different sources into a unified multidimensional feature vector, comprehensively reflecting the current anomaly state of the buoy.

[0059] In step S52, the classification model is based on machine learning algorithms such as support vector machines, decision trees, random forests, or deep neural networks, and is trained on a labeled dataset containing various known fault modes (such as sensor failure, seal leakage, buoyancy pump performance degradation, power system anomalies, communication failures, etc.). The classification model learns the abnormal feature vector patterns corresponding to different fault modes to classify and diagnose the current abnormal state of the buoy. The output includes not only the most likely fault type but also the confidence level for each fault type.

[0060] In this embodiment, step S6 generates and executes corresponding early warning or control commands based on the fault diagnosis results. After obtaining accurate fault diagnosis results, the monitoring system automatically generates and executes corresponding early warning or control commands based on preset emergency response strategies and fault severity. Early warning commands are sent to operators and maintenance personnel via email, SMS, and other methods, informing them of the fault type, severity, and current buoy status.

[0061] First, the data preprocessing module employs advanced data cleaning methods such as the three-standard-deviation criterion and Kalman filtering to perform quality control and normalization of the data, providing a high-accuracy foundation for subsequent analysis. Second, based on the preprocessed buoy internal environmental parameters, a Transformer network model is used to learn the cabin pressure, temperature, and humidity data to construct a coupled state model of the cabin environment. This model captures the nonlinear coupling relationships and time-series dependencies between parameters. Simultaneously, the consistency of coupling between parameters is quantified using Mahalanobis distance, effectively identifying subtle anomalies. Furthermore, a relationship between buoy attitude control parameters and depth data is established. The invention employs a physical correlation model, combining fluid dynamics and buoyancy principles for depth prediction, thus solving the problem of existing technologies failing to identify buoy attitude control anomalies due to the absence of this model. Furthermore, the invention utilizes multimodal data fusion technology and a complex deep learning architecture, which, compared to traditional statistical methods or simple machine learning models commonly used in existing technologies, can identify cross-modal anomaly patterns. Moreover, by integrating multiple anomaly information through Bayesian networks or multilayer perceptron models to form a unified multidimensional feature vector, the invention achieves dynamic fusion of multi-source anomaly detection and fault diagnosis, overcoming the limitations of existing technologies that only focus on a single anomaly detection module, resulting in higher diagnostic accuracy.

[0062] In the above illustrative embodiments, the monitoring method of Deep Argo buoys preprocesses multi-dimensional data, constructs a coupled state model of the cabin environment and a physical model of buoy motion, and combines cross-domain consistency verification and probabilistic graph fusion decision-making mechanisms to accurately identify abnormal information in the buoy cabin environment and buoy attitude control, trace the source of abnormalities in ocean physical observation data, and achieves accurate determination of fault types through multi-source information weighted fusion and classification diagnosis, ensuring the reliability of monitoring data and the stability of buoy operation, and improving the accuracy and continuity of deep-sea observation.

[0063] Example 2: See appendix Figures 1 to 2 This paper presents an illustrative embodiment of the Deep Argo buoy monitoring system proposed in this invention, applied to the Deep Argo buoy monitoring method of Embodiment 1. The Deep Argo buoy monitoring system includes a data acquisition module, a data preprocessing module, an in-cabin environment coupling state modeling module, a buoy motion physical modeling module, a cross-domain data consistency verification and anomaly differentiation module, a multi-source anomaly detection and fault diagnosis module, and a status early warning and reporting module.

[0064] The data acquisition module is used to acquire environmental parameters inside the buoy chamber, buoy attitude control parameters, and ocean physical observation data. The environmental parameters inside the buoy chamber include air pressure data, temperature data, and humidity data. The buoy attitude control parameters include buoyancy pump fuel volume data, pump motor current data, buoy depth data, tilt angle data, and heading angle data. The ocean physical observation data includes external environmental temperature data, external environmental salinity data, and external depth data.

[0065] The data preprocessing module is connected to the data acquisition module. The data preprocessing module is used to perform time stamp synchronization, data alignment, missing value filling, outlier filtering, and feature normalization preprocessing on the environmental parameters, attitude control parameters, and marine physical observation data inside the buoy cabin.

[0066] The in-cabin environment coupled state modeling module and the data preprocessing module are connected. The in-cabin environment coupled state modeling module is used to construct an in-cabin environment coupled state model based on the preprocessed buoy in-cabin environmental parameters, and to capture the nonlinear coupling relationship and time-series dependence between in-cabin air pressure data, in-cabin temperature data and in-cabin humidity data, so as to identify abnormal information in the buoy in-cabin environment.

[0067] Specifically, the cabin environment coupled state modeling module processes pre-processed cabin pressure, temperature, and humidity data. Internally, this module incorporates multivariate time series models, such as vector autoregression (VAR) or long short-term memory (LSTM) networks, to capture the dynamic coupling relationships among these three parameters. Furthermore, the module includes a feature extraction unit capable of extracting advanced features from the time-series data, such as rate of change, fluctuation amplitude, and correlation coefficients, to characterize the stability and potential anomalies of the cabin environment.

[0068] The buoy motion physics modeling module and the data preprocessing module are connected. The buoy motion physics modeling module is used to build a buoy motion physics model and calculate the theoretical depth of the Deep Argo buoy in order to identify abnormal information in buoy attitude control.

[0069] Specifically, the buoy motion physics modeling module is used to process buoy oil volume data, pump motor current data, and corresponding depth data. By establishing physical models such as the buoy-depth model based on Archimedes' principle and fluid dynamics for the buoy buoyancy control system, it predicts the theoretical depth that the buoy should reach under specific oil volume and pump motor current. At the same time, it analyzes the temporal characteristics of the pump motor current to evaluate the working efficiency and potential mechanical failures of the buoyancy pump.

[0070] The cross-domain data consistency verification and anomaly differentiation module is connected to the data preprocessing module, the cabin environment coupled state modeling module, and the buoy motion physics modeling module, respectively. The cross-domain data consistency verification and anomaly differentiation module is used to perform cross-domain consistency verification based on theoretical depth and ocean physical observation data. Combining the anomaly information identified by the cabin environment coupled state model and the buoy motion physics model, it distinguishes whether the source of the anomaly in the ocean physical observation data is the buoy's own state anomaly or a real ocean physical signal.

[0071] Specifically, the cross-domain data consistency verification and anomaly differentiation module receives and processes marine physical observation data such as temperature and salinity profiles. Internally, this module includes an oceanographic background knowledge base and a historical observation database, enabling preliminary quality assessment of the input temperature and salinity profile data. Furthermore, it integrates multi-scale feature extraction algorithms to capture various scale features of the thermocline, halocline depth, thickness, and seasonal variations in marine physical signals.

[0072] The multi-source anomaly detection and fault diagnosis module is connected to the cabin environment coupled state modeling module, the buoy motion physics modeling module, and the cross-domain data consistency verification and anomaly differentiation module, respectively. The multi-source anomaly detection and fault diagnosis module is used to perform multi-source anomaly detection and fault diagnosis on the Deep Argo buoy based on the anomaly information of the buoy cabin environment, the anomaly information of the buoy attitude control, the cross-domain consistency verification results, and the differentiated sources of anomalies in the ocean physical observation data, so as to identify the potential fault types of the Deep Argo buoy.

[0073] In this embodiment, the multi-source anomaly detection and fault diagnosis module is the core intelligent unit of the Deep Argo buoy monitoring system. It uses multimodal data fusion technology and deep learning models, such as neural networks with fused attention mechanisms, to comprehensively analyze features from different modules and identify cross-modal anomaly patterns.

[0074] The status warning and reporting module is connected to the multi-source anomaly detection and fault diagnosis module. The status warning and reporting module is used to generate a buoy health status warning and diagnosis report that includes a comprehensive score of buoy health status and fault warning information based on the results of multi-source anomaly detection and fault diagnosis.

[0075] The Status Warning and Reporting module receives anomaly detection results from the Multi-Source Anomaly Detection and Fault Diagnosis module. Internally, this module stores a knowledge base or trained classification models containing various fault modes, such as decision tree-based, support vector machine-based, or neural network-based classifiers. Based on the characteristics of the anomaly patterns, the module maps them to known Deep Argo buoy fault types, such as hull microleakage, sensor drift, buoyancy pump performance degradation, or communication failure, and provides the corresponding fault confidence level.

[0076] In this embodiment, the monitoring system of the Deep Argo buoy also includes a command generation module, which is connected to the status warning and reporting module. Based on fault diagnosis results and preset strategy rules, the command generation module automatically generates and executes corresponding warning or control commands. Warning commands can be sent to the operator via email, SMS, or system interface; control commands are transmitted back to the Deep Argo buoy via satellite communication link, causing the buoy to adjust its operating strategy.

[0077] The data acquisition module includes a MEMS barometric pressure sensor, a platinum resistance temperature sensor, a capacitive humidity sensor, an oil level sensor, a pump motor current sensor, and a CTD sensor.

[0078] The buoy's own depth data is acquired by pressure sensors integrated inside the buoy (such as MEMS pressure sensors). By measuring the correspondence between seawater pressure and depth (seawater pressure increases with depth), the real-time depth of the buoy is calculated. The tilt angle data (including pitch angle and roll angle) is measured by the accelerometer and gyroscope in the buoy's built-in inertial measurement unit (IMU) to sense the buoy's tilt attitude in three-dimensional space. The heading angle data is also acquired by the inertial measurement unit (IMU) in combination with a magnetometer (or Beidou / GPS assisted calibration) to determine the buoy's direction of travel or attitude orientation in the ocean.

[0079] In this embodiment, the monitoring system of the Deep Argo buoy also includes a data receiving module. This module receives various types of data transmitted from the Deep Argo buoy via a satellite communication system. This data includes timestamps, a unique buoy identifier, a sensor type identifier, and corresponding measurement values. The output of the data receiving module is a raw, time-series-arranged buoy data stream, including cabin pressure data, cabin temperature data, cabin humidity data, buoy fuel level data, pump motor current data, depth data, temperature profile data, and salinity profile data.

[0080] The data preprocessing module is connected to the data receiving module and is used to perform quality control and formatting on the received raw data stream. Specifically, the data preprocessing module performs timestamp synchronization, data alignment, missing value imputation, outlier filtering, and feature normalization preprocessing.

[0081] Timestamp calibration and synchronization are performed because there may be slight sampling time deviations between different sensors. The data preprocessing module uses GPS timing information to accurately calibrate the timestamps of all data and align the time series. Time alignment is achieved through interpolation algorithms, such as cubic spline interpolation, to unify data from different sampling frequencies to a preset monitoring frequency.

[0082] For missing values, the data preprocessing module fills in missing data points caused by transmission interruption or sensor failure using extrapolation based on historical data trends or interpolation based on nearby valid data. For example, linear interpolation can be used for short-term missing data points, while Kalman filtering can be used for state estimation and filling for long-term missing data points or specific periodic missing data points.

[0083] Outlier filtering specifically involves the data preprocessing module using statistical methods such as the three-standard-deviation criterion or the median absolute deviation method to identify and remove outlier data points that clearly exceed a reasonable physical range or statistical distribution, in order to ensure data cleanliness.

[0084] Data normalization is used to eliminate the impact of differences in the dimensions and numerical ranges of different parameters on subsequent analysis. The data preprocessing module performs min-max normalization on all data, scaling the data to a uniform range of zero to one.

[0085] In the above illustrative embodiments, the monitoring system of the Deep Argo buoy achieves full-process collaboration of data acquisition, preprocessing, modeling, verification, diagnosis and early warning through modular design. It relies on multiple sensors to accurately capture three types of data, and the method logic of each module is seamlessly connected to efficiently complete model construction, anomaly differentiation and fault diagnosis, realize the automation and intelligence of the monitoring process, provide hardware support for real-time control of buoy status, and ensure the stable and efficient conduct of deep-sea observation missions.

[0086] Finally, it should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0087] The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation of the present invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications and substitutions should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. A monitoring method for Deep Argo buoys, characterized in that, Includes the following steps: S1. Acquire environmental parameters, buoy attitude control parameters, and ocean physical observation data inside the buoy cabin, and perform preprocessing. The buoy cabin environmental parameters include cabin air pressure data, cabin temperature data, and cabin humidity data; the buoy attitude control parameters include buoyancy pump oil volume data, pump motor current data, buoy depth data, tilt angle data, and heading angle data; the ocean physical observation data include external environmental temperature data, external environmental salinity data, and external depth data. S2. Based on the preprocessed environmental parameters inside the buoy chamber, construct a coupled state model of the chamber environment to capture the nonlinear coupling relationship and temporal dependence between the chamber air pressure data, chamber temperature data and chamber humidity data, so as to identify abnormal information of the buoy chamber environment. S3. Based on the preprocessed buoy attitude control parameters, construct a buoy motion physical model and calculate the theoretical depth of the Deep Argo buoy to identify abnormal information in buoy attitude control. S4. Based on the theoretical depth and ocean physical observation data, conduct cross-domain consistency verification. Combine the abnormal information identified by the in-cabin environment coupling state model and the buoy motion physical model to distinguish whether the source of the abnormality in the ocean physical observation data is the buoy's own state abnormality or the real ocean physical signal. S5. Based on the abnormal information of the buoy cabin environment, the abnormal information of buoy attitude control, the cross-domain consistency verification results, and the differentiated sources of abnormal ocean physical observation data, multi-source anomaly detection and fault diagnosis are performed on the Deep Argo buoy. S6. Based on the multi-source anomaly detection and fault diagnosis results, generate a buoy health status early warning and diagnosis report that includes a comprehensive buoy health status score and fault early warning information; Specifically, step S4 includes: S41. Based on historical oceanographic observation data under normal operating conditions of Deep Argo buoys, establish an observation data baseline model to characterize the typical features and spatiotemporal variation patterns of temperature profiles, salinity profiles, and depth profiles in oceanographic observation data. S42. Compare the actual collected external environmental temperature data, external environmental salinity data, and external depth data with the observation data baseline model to detect abnormal points or abnormal profiles in the marine physical observation data and complete the cross-domain consistency verification. S43. When an anomaly is detected in the ocean physical observation data, the anomaly information identified by the coupled state model of the cabin environment and the buoy motion physical model is combined, and a fusion decision-making mechanism based on rule-based reasoning and probabilistic graphical models is adopted to distinguish whether the source of the anomaly in the ocean physical observation data is an anomaly in the buoy's own state or a real ocean physical signal; wherein, the fusion decision-making mechanism based on rule-based reasoning and probabilistic graphical models adopts the following calculation formula: in, The probability that the observation data anomaly is caused by a real ocean physical signal when the observation data anomaly, the cabin environment anomaly, and the buoy attitude control anomaly occur simultaneously. These are real ocean physical signals; Anomalies were detected in the marine physical observation data; This refers to abnormal information regarding the environment inside the buoy compartment; This refers to abnormal information related to buoy attitude control. During calculation, if or and If they are highly synchronized in time and satisfy the correlation of physical mechanisms, then it is determined that... The source is an abnormal state of the buoy itself; if and None of them occurred, and If it consistently does not conform to historical climatic characteristics, then it is judged The source is real ocean physical signals; Step S5 specifically includes: S51. The abnormal information of the buoy cabin environment, the abnormal information of buoy attitude control, the cross-domain consistency verification results, and the differentiated sources of abnormal ocean physical observation data are weighted and fused to form a unified multi-dimensional feature vector. S52. Input the multidimensional feature vector into a pre-trained classification model, wherein the classification model is a multi-class support vector machine model or a decision tree model; S53. Combining the historical failure mode library, the potential failure types of the buoy are classified and diagnosed through the classification model. The potential failure types include slow leakage of the hull, overheating of internal electronic components, performance degradation of the buoyancy pump, minor leakage of oil circuits, and pseudo-anomalies in observation data caused by biofouling.

2. The monitoring method for the Deep Argo buoy according to claim 1, characterized in that, Step S2 specifically includes: S21. Integrate the preprocessed cabin air pressure data, cabin temperature data and cabin humidity data into a multivariable time series input vector sequence; S22. The Transformer network model is used to learn the multivariate time-series input vector sequence to capture the nonlinear coupling relationship and time-series dependence between the cabin air pressure data, cabin temperature data and cabin humidity data; S23. Construct a coupled state model of the cabin environment based on the learning results; S24. Calculate the differences between the predicted values ​​of cabin air pressure, cabin temperature, and cabin humidity output by the cabin environment coupling state model and the corresponding actual observed values, so as to quantify the degree of deviation of the cabin environment parameters of the buoy and identify abnormal information of the cabin environment of the buoy.

3. The monitoring method for the Deep Argo buoy according to claim 1, characterized in that, Step S3 specifically includes: S31. Input the pre-processed buoyancy pump oil volume data, pump motor current data, buoy depth data, tilt angle data, and heading angle data; S32. Based on the principles of fluid mechanics, and combining the geometric parameters, material density parameters, and real-time seawater density data of the Deep Argo buoy, establish a mathematical model of the buoyancy and resistance of the Deep Argo buoy at different depths. S33. Using an adaptive Kalman filter algorithm or an unscented Kalman filter algorithm, the buoyancy and resistance mathematical model and the real-time acquired buoy attitude control parameters are integrated to construct a buoy motion physical model in order to calculate the theoretical depth of the Deep Argo buoy. S34. Compare the theoretical depth with the actual collected buoy depth data to confirm the deviation of the buoy's kinematic characteristics and identify abnormal information in the buoy attitude control.

4. The monitoring method for the Deep Argo buoy according to claim 3, characterized in that, In step S33, the calculation formula for the buoy motion physical model is as follows: in, The net force acting on the Deep Argo buoy. For the total mass of the Deep Argo buoy, It is the acceleration due to gravity. For the fixed displacement volume of the Deep Argo buoy, The change in displacement volume corresponding to the buoyancy adjustment of the Deep Argo buoy. The coefficient of viscous drag. This is the pressure drag coefficient. This refers to the descent speed of the Deep Argo buoy.

5. The monitoring method for the Deep Argo buoy according to claim 3, characterized in that, In step S33, the theoretical depth is: in, The theoretical depth of the Deep Argo buoy predicted by the model; This is a nonlinear function used to map input parameters to theoretical depth; This is the current buoyancy pump oil level data for the Deep Argo buoy; Pump motor current data for Deep Argo buoys; The real-time seawater density data at the depth where the Deep Argo buoy is located is obtained by calculating the external environmental temperature and salinity data from the ocean physical observation data through the equation of state. This is the set of internal parameters of the model.

6. The monitoring method for the Deep Argo buoy according to claim 1, characterized in that, In step S1, the preprocessing specifically includes timestamp synchronization, data alignment, missing value filling, outlier filtering, and feature normalization of the environmental parameters inside the buoy cabin, the buoy attitude control parameters, and the ocean physical observation data.

7. A monitoring system for Deep Argo buoys, characterized in that, The monitoring method applied to the Deep Argo buoy as described in any one of claims 1-6, the monitoring system comprising: The data acquisition module is used to acquire environmental parameters inside the buoy chamber, buoy attitude control parameters, and ocean physical observation data. The environmental parameters inside the buoy chamber include internal air pressure, internal temperature, and internal humidity data. The buoy attitude control parameters include buoyancy pump fuel volume data, pump motor current data, buoy depth data, tilt angle data, and heading angle data. The ocean physical observation data includes external environmental temperature data, external environmental salinity data, and external depth data. A data preprocessing module is connected to the data acquisition module. The data preprocessing module is used to perform time stamp synchronization, data alignment, missing value filling, outlier filtering, and feature normalization preprocessing on the environmental parameters, buoy attitude control parameters, and marine physical observation data inside the buoy cabin. An internal environment coupling state modeling module is connected to the data preprocessing module. The internal environment coupling state modeling module is used to construct an internal environment coupling state model based on the preprocessed internal environment parameters of the buoy, and to capture the nonlinear coupling relationship and time-series dependence between the internal air pressure data, internal temperature data and internal humidity data, so as to identify abnormal information of the internal environment of the buoy. A buoy motion physics modeling module is connected to the data preprocessing module. The buoy motion physics modeling module is used to construct a buoy motion physics model and calculate the theoretical depth of the Deep Argo buoy in order to identify abnormal information in the buoy attitude control. A cross-domain data consistency verification and anomaly differentiation module is connected to the data preprocessing module, the in-cabin environment coupled state modeling module, and the buoy motion physics modeling module. The cross-domain data consistency verification and anomaly differentiation module is used to perform cross-domain consistency verification based on the theoretical depth and ocean physical observation data. Combined with the anomaly information identified by the in-cabin environment coupled state model and the buoy motion physics model, it distinguishes whether the source of the anomaly in the ocean physical observation data is an abnormality in the buoy's own state or a real ocean physical signal. A multi-source anomaly detection and fault diagnosis module is connected to the in-cabin environment coupling state modeling module, the buoy motion physics modeling module, and the cross-domain data consistency verification and anomaly differentiation module, respectively. The multi-source anomaly detection and fault diagnosis module is used to perform multi-source anomaly detection and fault diagnosis on the Deep Argo buoy based on anomaly information of the buoy's in-cabin environment, anomaly information of the buoy's attitude control, cross-domain consistency verification results, and the differentiated sources of anomalies in the ocean physical observation data. The status warning and reporting module is connected to the multi-source anomaly detection and fault diagnosis module. The status warning and reporting module is used to generate a buoy health status warning and diagnosis report containing a comprehensive buoy health status score and fault warning information based on the multi-source anomaly detection and fault diagnosis results.

Images