Marine ecological heavy metal pollution grade early warning method

Abstract

The present application relates to the technical field of marine environment monitoring, and particularly relates to a heavy metal pollution grade early warning method for marine ecology, which comprises the following steps: obtaining target data, fusing the target data, and generating a pollution situation field; identifying a high uncertainty area and a high risk evolution area based on the pollution situation field, and generating virtual monitoring points in the high uncertainty area and the high risk evolution area; scheduling mobile monitoring units to perform verification monitoring according to the virtual monitoring points, and waking up fixed monitoring points; obtaining verification monitoring real-time data collected by the mobile monitoring units and the fixed monitoring points, and correcting the pollution situation field by using the verification monitoring real-time data to generate an updated pollution situation field; and generating a heavy metal pollution grade early warning signal based on the updated pollution situation field.

Description

Technical Field

[0001] This invention relates to the field of marine environmental monitoring technology, and in particular to a method for early warning of heavy metal pollution levels in marine ecosystems. Background Technology

[0002] Marine heavy metal pollution is a significant environmental problem threatening the health of marine ecosystems and human food security. To address this type of pollution, early warning systems are typically deployed for early detection and rapid response. These systems aim to monitor changes in pollutant concentrations and issue alerts when pollution levels exceed safe thresholds, thereby activating emergency response plans and mitigating the harm caused by pollution incidents. Therefore, developing efficient and accurate early warning methods and systems for marine heavy metal pollution levels is crucial for marine environmental protection.

[0003] Existing marine heavy metal pollution early warning methods typically rely on a network of fixed monitoring stations deployed in key sea areas. These stations periodically collect water samples or measure heavy metal concentrations in real time using sensors. When the collected data exceeds a preset alarm threshold, the system triggers an alarm. Some advanced methods incorporate simple hydrodynamic models to make preliminary inferences about the possible diffusion direction of pollutants, thus aiding in early warning decision-making.

[0004] However, existing technologies have the following drawbacks: Fixed-site monitoring networks have limited coverage, resulting in numerous monitoring blind spots across vast sea areas, and they cannot effectively track pollution sources that dynamically change with ocean currents and meteorological conditions. Their early warning mechanisms are mostly passively triggered, meaning an alarm is only issued when pollutants reach the monitoring point and exceed standards, leading to a significant lag in the early warning response. Furthermore, they do not adequately consider the combined impact of multiple complex factors such as marine hydrology, meteorology, and pollution sources, and the accuracy and foresight of the assessment results need improvement, making it difficult to meet the demand for rapid and accurate early warnings for sudden and dynamic pollution events. Summary of the Invention

[0005] This invention employs techniques such as constructing a pollution situation field, generating virtual monitoring points to schedule monitoring resources, and using confirmatory data to close the loop and correct the situation field. This approach can generate highly reliable early warning signals and improve the accuracy of early warnings.

[0006] The technical solution proposed in this invention is: a method for early warning of heavy metal pollution levels in marine ecosystems, the method comprising:

[0007] Acquire target data and perform fusion processing on the target data to generate a pollution situation field;

[0008] Based on the pollution situation field, high uncertainty areas and high-risk evolution areas are identified, and virtual monitoring points are generated in the high uncertainty areas and high-risk evolution areas;

[0009] Based on the virtual monitoring points, the mobile monitoring unit is dispatched to conduct verification monitoring and the fixed monitoring points are activated.

[0010] Acquire real-time verification monitoring data collected by mobile monitoring units and fixed monitoring points, and use the real-time verification monitoring data to correct the pollution situation field and generate an updated pollution situation field.

[0011] Based on the updated pollution situation field, a heavy metal pollution level early warning signal is generated.

[0012] Preferably, the step of acquiring target data and fusing the target data to generate a pollution situation field includes:

[0013] Acquire real-time monitoring data including heavy metal concentration data and sensor location information; acquire ocean hydrological data including tidal current velocity and water temperature parameters; acquire meteorological data including wind speed and wind direction parameters;

[0014] Acquire pollution source data including industrial emission locations and river input flow parameters; perform spatiotemporal alignment and normalization processing on real-time monitoring data, marine hydrological data, meteorological data, and pollution source data to generate a fused dataset;

[0015] By utilizing fused datasets and performing dynamic simulations through predictive models, a pollution state field is generated. ;

[0016] The prediction model is built on a long short-term memory neural network, with the input being a fused dataset and the output being a pollution state field; the spatiotemporal alignment uses time interpolation, and the normalization process uses the Min-Max scaling algorithm.

[0017] Preferably, the step of identifying high-uncertainty areas and high-risk evolution areas based on the pollution situation field, and generating virtual monitoring points in the high-uncertainty areas and high-risk evolution areas, includes:

[0018] The pollution situation field is analyzed to extract pollutant concentration distribution and diffusion trend parameters; the diffusion trend parameters are the spatiotemporal gradient vectors of the pollution situation field. ;in Indicates the pollution situation;

[0019] Based on pollutant concentration distribution and diffusion trend parameters, regional uncertainty indicators and risk evolution indicators are calculated. Among them, diffusion rate Directional weighting factor ; Indicates the angle between the direction of pollutant diffusion and the direction of the sensitive ecological zone; Indicates the weighting coefficient;

[0020] Based on regional uncertainty indicators and risk evolution indicators, identify regions with high uncertainty and regions with high risk evolution;

[0021] Within high-uncertainty and high-risk evolution zones, virtual monitoring points are generated, containing location information and monitoring priorities. The location information includes the latitude and longitude of the uncertain and high-risk evolution zones; the monitoring priorities... Among them, regional uncertainty indicators ; The value of the spatiotemporal gradient vector;

[0022] The high uncertainty region is the region with a regional uncertainty index greater than 0.7, and the high risk evolution region is the region with a risk evolution index greater than 0.8.

[0023] Preferably, the step of scheduling the mobile monitoring unit to perform verification monitoring based on the virtual monitoring point and waking up the fixed monitoring point includes:

[0024] Obtain the location information and monitoring priority of virtual monitoring points;

[0025] Based on location information and monitoring priority, a mobile monitoring unit scheduling instruction is generated;

[0026] Based on the scheduling instructions of the mobile monitoring unit, control the mobile monitoring unit to go to the virtual monitoring point to perform verification monitoring;

[0027] Simultaneously, based on the spatial distribution of the virtual monitoring points, the nearby fixed monitoring points with dormancy functions are awakened, and their acquisition frequency is adjusted.

[0028] Preferably, the step of using real-time data from confirmatory monitoring to correct the pollution situation field and generate an updated pollution situation field includes:

[0029] Receive real-time verification monitoring data and perform quality verification and anomaly filtering on it using the Z-score anomaly detection algorithm to generate a calibration dataset;

[0030] The calibration dataset is input into a data assimilation algorithm that can correct model parameters to generate corrected simulation parameters; the data assimilation algorithm is an ensemble Kalman filter algorithm.

[0031] The corrected simulation parameters are input into the pre-trained prediction model to generate an updated pollution situation field.

[0032] Preferably, the step of generating a heavy metal pollution level early warning signal based on the updated pollution situation field includes:

[0033] Analyze the updated pollution situation field and extract parameters such as peak concentration and diffusion range of heavy metal pollutants;

[0034] Based on the peak concentration and diffusion range parameters of heavy metal pollutants, and combined with a rule matrix containing pollution level thresholds, the warning level is calculated.

[0035] Based on the warning level, a heavy metal pollution level warning signal is generated and sent to the emergency platform through the corresponding alarm equipment.

[0036] The graded early warning signal includes the early warning level, peak concentration, diffusion range, and timestamp; the rule matrix is... The row vector elements within the rule matrix represent the warning levels corresponding to different extended ranges within the same peak concentration range.

[0037] Preferred options also include:

[0038] The internal weights of the adaptive learning model are trained using the corrected simulation parameters to optimize the prediction accuracy of the adaptive learning model for the pollutant diffusion pattern.

[0039] Based on the optimized adaptive learning model, the generation rules of the virtual monitoring points are dynamically adjusted.

[0040] The adaptive learning model is built on a convolutional neural network, taking a pollution situation field as input and outputting a prediction error; the training process is carried out through backpropagation with a learning rate of 0.001.

[0041] Preferably, the step of generating a mobile monitoring unit scheduling instruction based on location information and monitoring priority includes:

[0042] Obtain the current status and location parameters of the mobile monitoring unit;

[0043] By combining the monitoring priority of virtual monitoring points with the current status and location parameters of mobile monitoring units, the scheduling path and resource allocation scheme are calculated.

[0044] Based on the scheduling path and resource allocation scheme, a mobile monitoring unit scheduling instruction is generated.

[0045] Preferably, the method further includes:

[0046] After the heavy metal pollution level warning signal is issued, collect real-time data from confirmatory monitoring and feedback data including on-site investigation results;

[0047] By utilizing feedback data, the accuracy of early warnings and the efficiency of response are evaluated to obtain evaluation results;

[0048] Based on the assessment results, the construction parameters of the pollution situation field and the generation rules of virtual monitoring points were adjusted.

[0049] A computer-readable storage medium storing a computer program that is executed by a processor to implement the aforementioned method for early warning of heavy metal pollution levels for marine ecosystems.

[0050] The beneficial effects of this invention are:

[0051] 1. This invention constructs a pollution situation field and generates virtual monitoring points, intelligently guiding mobile monitoring units to conduct precise surveys while simultaneously activating fixed monitoring points, thus building a dynamically reconfigurable monitoring network. This enhances the ability to capture the spatial evolution of pollution events.

[0052] 2. This invention uses real-time verification monitoring data collected from mobile and fixed monitoring points to instantly correct and update the pollution situation field, thereby improving the accuracy and reliability of early warning assessment and effectively reducing the risk of false alarms and missed alarms.

[0053] 3. By simulating pollution patterns and identifying high-risk evolution areas, this invention achieves a shift from passive monitoring to proactive early warning. It can predict the diffusion path and potential impact of pollutants, deploy monitoring resources to key locations in advance, and issue early warning signals earlier, greatly improving response speed. Attached Figure Description

[0054] Figure 1 This is a flowchart of a heavy metal pollution level early warning method for marine ecology according to the present invention. Detailed Implementation

[0055] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious modifications will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0056] It is understood that the term "a" should be understood as "at least one" or "one or more," that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0057] Example 1: Reference Figure 1 The technical solution provided by this invention is: a method for early warning of heavy metal pollution levels in marine ecosystems, the method comprising:

[0058] Step 1: Acquire target data and perform fusion processing on the target data to generate a pollution situation field, specifically including the following steps:

[0059] Acquire real-time monitoring data including heavy metal concentration data and sensor location information; acquire marine hydrological data including tidal current velocity and water temperature parameters; acquire meteorological data including wind speed and wind direction parameters; acquire pollution source data including industrial emission location and river input flow parameters;

[0060] For real-time monitoring data (e.g., Cu concentration from the Bohai Bay sensor = The data, including marine hydrological data (e.g., tidal current velocity of 0.5 m / s), meteorological data, and pollution source data, are spatiotemporally aligned and normalized to generate a fused dataset. The spatiotemporal alignment uses time interpolation, and the normalization process employs the Min-Max scaling algorithm.

[0061] By utilizing fused datasets and performing dynamic simulations through predictive models, a pollution state field is generated.

[0062] in, Represents spatial coordinates and time The pollution situation field, each grid node stores the predicted heavy metal pollution concentration value (unit: The spatial grid resolution is set to 100m×100m×10m (depth) or risk level (1-5, with 1 being the lowest risk). The default settings are 100m×100m×10m (depth) and 1-hour time step. This represents a pre-trained prediction model built on a Long Short-Term Memory (LSTM) neural network. The network architecture includes: an input layer (dimension equal to the number of features in the fused dataset), two LSTM hidden layers (128 neurons each), and an output layer (dimension matching the contamination situation grid). Training process: A historical dataset (≥10,000 samples) was used, with 100 training epochs. The loss function was mean squared error (MSE), and the optimizer was Adam.

[0063] The fused dataset is input into the prediction model, and the output is the pollution state field. These represent the sets of real-time monitoring data, marine hydrological data, meteorological data, and pollution source data extracted and preprocessed from the fused dataset, respectively.

[0064] The rule matrix is: The row vector elements within the rule matrix represent the warning levels corresponding to different extended ranges within the same peak concentration range.

[0065] when heavy metal At that time, if the diffusion range ,but =1; if ,but ;if ,but ;

[0066] when If the diffusion range ,but =2; if ,but ;if ,but ;

[0067] when If the diffusion range ,but =3; if ,but ;if ,but .in, and Indicates the concentration threshold. and This represents the diffusion range threshold. In this embodiment, , , , .

[0068] ;

[0069] Output .

[0070] The purpose of this step is to transform multi-source, heterogeneous data into a unified and predictive pollution situation field. In this embodiment, the system acquires four types of key data through preset interfaces. Real-time monitoring data comes from fixed or mobile sensor networks deployed in the ocean, and its core content is heavy metal concentration data at a specific timestamp and the corresponding sensor location information. Ocean hydrological data usually comes from ocean numerical models or measured buoys, mainly including tidal current velocity, tidal cycle, and water temperature parameters that affect the rate of material diffusion, describing the state of seawater movement. Meteorological data is usually obtained from meteorological forecasting systems, including wind speed, wind direction, and precipitation intensity parameters that affect the transport and deposition of pollutants on the sea surface. Pollution source data comes from monitoring known land-based or marine sewage outlets, specifically industrial discharge locations and river input flow parameters. After acquiring the above data, spatiotemporal alignment and normalization processing are performed.

[0071] Step 2: Based on the pollution situation field, identify high-uncertainty areas and high-risk evolution areas, and generate virtual monitoring points in these areas. This includes the following steps:

[0072] The pollution situation field is analyzed to extract pollutant concentration distribution and diffusion trend parameters; the diffusion trend parameters are the spatiotemporal gradient vectors of the pollution situation field. .

[0073] Based on pollutant concentration distribution and diffusion trend parameters, regional uncertainty indicators and risk evolution indicators are calculated. diffusion rate Directional weighting factor ; Indicates the angle between the direction of pollutant diffusion and the direction of the sensitive ecological zone; In this embodiment, the weighting coefficient is represented. =0.6, =0.4.

[0074] Based on regional uncertainty indicators and risk evolution indicators, high uncertainty regions and high-risk evolution regions are identified. High uncertainty regions are defined as those with a regional uncertainty index greater than 0.7, and high-risk evolution regions are defined as those with a risk evolution index greater than 0.8. The thresholds of 0.7 and 0.8 are calibrated based on a historical pollution event dataset (sample size = 500) to ensure a 95% confidence level.

[0075] Within high-uncertainty and high-risk evolution zones, virtual monitoring points are generated, containing location information and monitoring priorities. The location information includes the latitude and longitude of the uncertain and high-risk evolution zones; the monitoring priorities... Among them, regional uncertainty indicators ; The value of the spatiotemporal gradient vector; the generation density of virtual monitoring points is After virtual monitoring points are generated, a real-time update mechanism is implemented: the data is recalculated every 30 minutes. and In order to adapt to changes in the situation.

[0076] The purpose of this step is to transform the macroscopic pollution situation field generated in the previous step into specific, executable monitoring task instructions. Specifically, in this embodiment, the pollution situation field is analyzed and treated as a digital grid containing multi-dimensional information. At each grid node, two core parameters are extracted: pollutant concentration distribution and diffusion trend parameter. The pollutant concentration distribution describes the spatial distribution of heavy metals at the current moment or predicts the future moment, while the diffusion trend parameter is obtained by calculating the gradient of the pollution situation field in space and the rate of change in time, thus quantifying the speed and direction of pollutant diffusion.

[0077] After obtaining parameters for pollutant concentration distribution and diffusion trends, regional uncertainty indicators and risk evolution indicators are calculated. Regional uncertainty indicators measure the confidence level of the simulation model's predictions for a specific region, while risk evolution indicators comprehensively assess the potential hazards of pollution events.

[0078] When pollutants are moving towards the sensitive area The value increases. When the direction of pollutant diffusion is directly opposite to the sensitive ecological area... The risk evolution index indicates that the risk in a region depends not only on the rate of pollutant diffusion, but also on the direction of the threat it poses.

[0079] For example, when the regional uncertainty index is 0.8, Generate a virtual monitoring point (121.0, 38.5, 0.85), where 0.85 represents the monitoring priority.

[0080] Step 3: Based on the virtual monitoring points, schedule the mobile monitoring unit to conduct verification monitoring and wake up the fixed monitoring points. This includes the following steps:

[0081] Step 3.1: Obtain the location information and monitoring priority of the virtual monitoring points;

[0082] Based on location information and monitoring priority, a mobile monitoring unit scheduling instruction is generated, which specifically includes: obtaining the current status and location parameters of the mobile monitoring unit; calculating the scheduling path and resource allocation scheme by combining the monitoring priority of the virtual monitoring point and the current status and location parameters of the mobile monitoring unit; and generating the mobile monitoring unit scheduling instruction based on the scheduling path and resource allocation scheme.

[0083] In this embodiment, real-time telemetry data of each available mobile monitoring unit is acquired. This data constitutes its current state and location parameters, specifically including high-precision geographic coordinates obtained through the Global Positioning System (GPS) and key performance indicators such as energy reserves and navigation speed obtained through onboard sensors. A genetic optimization algorithm is used to deeply integrate this data with the virtual monitoring points and their monitoring priorities obtained from the previous step to calculate the optimal scheduling path and resource allocation scheme. The objective function is: ;in, Indicates the assignment of mobile monitoring units Head to virtual monitoring point The required sailing time is calculated based on the unit's current position parameters and sailing capabilities, as well as the geographical location information of the target point. Indicates the number of virtual monitoring points; Represents virtual monitoring points The monitoring priority is a dimensionless weighting coefficient; the higher the value, the more urgent and important the monitoring task at that point. In this embodiment, the process of solving the objective function of the optimization algorithm is as follows: initialization is performed to obtain a list of mobile monitoring units and a list of virtual detection points (arranged in descending order of monitoring priority); then the cost matrix is ​​calculated. Set the population size to 50, the crossover rate to 0.8, and the number of iterations to 100; output the optimal scheduling path.

[0084] The above method, through the coordinated operation of synchronously waking up fixed dormant monitoring points, pre-constructs a local, high-density observation background field in the target area before the arrival of the mobile monitoring unit. This ensures that subsequent verification monitoring is not an isolated point measurement, greatly improving the efficiency of data acquisition and the effectiveness of verification.

[0085] Step 3.2: Based on the mobile monitoring unit scheduling instructions, control the mobile monitoring unit to proceed to the virtual monitoring point to perform verification monitoring. Specifically, based on the calculated optimal scheduling path, generate one or more machine-executable mobile monitoring unit scheduling instructions. These instructions contain a clear sequence of target coordinates and speed parameters.

[0086] Step 3.3: Based on the spatial distribution of the virtual monitoring points, wake up the nearby fixed monitoring points with sleep function and adjust their acquisition frequency.

[0087] The purpose of steps 3.2 and 3.3 is to transform the abstract monitoring task generated in the previous steps into actual command and control of physical monitoring resources. This process begins with acquiring a set of virtual monitoring points, each containing explicit geographical location information and quantified monitoring priorities. Specifically, in this embodiment, for each available mobile monitoring unit, such as an unmanned surface vessel or underwater robot, its own state parameters are then acquired, including its current location, energy reserves, and navigation capabilities.

[0088] Based on these inputs, an optimized scheduling algorithm is executed to generate scheduling instructions for mobile monitoring units. The goal of this algorithm is to optimize resource utilization while meeting task requirements. Its core is solving a problem of minimizing a cost function, which can be expressed as: ,in, Represented as a mobile monitoring unit Assign virtual monitoring points The resulting overall costs. This indicates the number of monitoring units; the cost is based on the number of units. Current location to target virtual monitoring point The weighted calculation results of the required travel time and energy consumption for location information. Indicates the target virtual monitoring point The monitoring priority is a dimensionless weight; the higher the value, the greater the monitoring value of that point.

[0089] The goal of this optimization solution is to find a task allocation scheme. This approach minimizes the overall unit priority cost, ensuring that the highest priority tasks receive a response at the fastest speed and lowest cost. The algorithm's output is a structured mobile monitoring unit scheduling instruction, which specifies the target virtual monitoring point and recommended flight path for each mobile unit.

[0090] For example, while generating instructions, based on the target path of the mobile monitoring unit, the preset fixed dormant monitoring points within a 5-kilometer radius along the route or near the target point are identified, and wake-up instructions are sent to them to enable them to enter a high-frequency working mode in coordination, that is, to adjust their collection frequency from once per hour to once per 10 minutes.

[0091] Mobile monitoring unit The location is (120.0, 38.0). The genetic optimization algorithm outputs the shortest path value of the virtual monitoring point. The travel time is 10 minutes. The nearby fixed dormant monitoring point is woken up. The monitoring point ID is 005.

[0092] Step 3 enables intelligent scheduling of monitoring resources through "virtual-real linkage" and "dynamic-static combination." It not only dispatches mobile monitoring units to precisely survey key predicted points but also simultaneously activates surrounding fixed monitoring points to form a high-density local observation network. Through the synergy of these two approaches, the system can acquire the most critical and timely verification data on high-risk areas in both spatial and temporal dimensions with extremely high efficiency and low cost. This transforms the entire monitoring network from a passive, static system into an active, reconfigurable intelligent system, providing high-quality data input for subsequent closed-loop feedback and model correction, fundamentally improving the response speed and verification efficiency of early warnings.

[0093] Step 4: Acquire real-time confirmatory monitoring data collected by the mobile monitoring unit and fixed monitoring points, and use the real-time confirmatory monitoring data to correct the pollution situation field and generate an updated pollution situation field. This specifically includes the following steps:

[0094] Receive real-time verification monitoring data and perform quality verification and anomaly filtering using the Z-score anomaly detection algorithm to generate a calibration dataset; when If this occurs, it is considered a data anomaly;

[0095] The calibration dataset is input into a data assimilation algorithm that can correct model parameters to generate corrected simulation parameters; the data assimilation algorithm is an ensemble Kalman filter algorithm (ensemble size = 50, covariance matrix initialization based on historical error).

[0096] The corrected simulation parameters are input into the pre-trained prediction model to generate an updated pollution situation field.

[0097] The purpose of this step is to perform real-time calibration of macroscopic predictions using newly acquired real data. Specifically, in this embodiment, real-time verification monitoring data from scheduled mobile monitoring units and awakened fixed dormant monitoring points is continuously monitored and received. This data includes heavy metal concentration values ​​at high spatiotemporal resolution, along with their precise acquisition coordinates and timestamps. After acquiring the data, the data processing flow is initiated. Quality verification and anomaly filtering are performed on the verification monitoring data. Quality verification aims to assess the reliability of the data, for example, by comparing it with historical data ranges or data from neighboring sensors to eliminate obvious sensor malfunctions or transmission errors. Anomaly filtering uses statistical methods to identify and process extreme outliers in the data, which may be transient disturbances caused by non-polluting factors. After processing, a high-quality calibration dataset is formed. This calibration dataset is used as a key input and fed back to the prediction model, whose objective function can be expressed as: ;in, It is the cost function that needs to be minimized. This represents the state vector, i.e., the simulation parameters that need to be corrected. This represents the observation vector. It is an observation operator (linear interpolation function) that will measure the model state Mapping to the observation space means simulating the values ​​that should exist at the observation point. It is the observation error covariance matrix, which characterizes the uncertainty of the observation data. It is the background field state vector before correction, i.e., the initial prediction parameters of the prediction model. It is the background field error covariance matrix ( Initialized as an identity matrix, this represents the uncertainty in the model's predictions. By solving this objective function, the optimal simulation parameters are obtained. .Will The model takes a predictive input and outputs an updated pollution situation field. This update process is real-time, ensuring that the pollution situation field can quickly reflect changes in perception brought about by new data.

[0098] The above steps, through dynamic and continuous calibration and refinement using measured data, endow the pollution situation field with powerful adaptive and self-correcting capabilities. The pollution diffusion range, intensity, and path it depicts become increasingly accurate with the infusion of real data. This fundamentally solves the problem of traditional open-loop prediction models being out of sync with real-world dynamics, ensuring the reliability and accuracy of subsequent early warning decisions.

[0099] Step 5: Based on the updated pollution situation field, generate a heavy metal pollution level early warning signal, which specifically includes the following steps:

[0100] Analyze the updated pollution situation field and extract parameters such as peak concentration and diffusion range of heavy metal pollutants;

[0101] Based on the peak concentration and diffusion range parameters of heavy metal pollutants, and combined with a rule matrix containing pollution level thresholds, the warning level is calculated.

[0102] Based on the warning level, a heavy metal pollution level warning signal is generated and sent to the emergency platform through the corresponding alarm equipment.

[0103] The graded early warning signal includes the early warning level, peak concentration, diffusion range, and timestamp; the rule matrix is... The row vector elements within the rule matrix represent the warning levels corresponding to different extended ranges within the same peak concentration range.

[0104] when heavy metal At that time, if the diffusion range ,but =1; if ,but ;if ,but ;

[0105] when If the diffusion range ,but =2; if ,but ;if ,but ;

[0106] when If the diffusion range ,but =3; if ,but ;if ,but .

[0107] , Indicates the pollutant concentration threshold. , Indicates the extended range threshold;

[0108] The purpose of this step is to transform the dynamic pollution situation field, which has been calibrated with real-time data, into early warning information that can be used for decision-making.

[0109] For example, if for If the diffusion range is 6 square kilometers and the rule matrix mapping is a level 4 warning, then:

[0110] The heavy metal pollution level warning signal is {Level: 4, peak concentration 25}. / L, diffusion range: 6 square kilometers, peak location (121.0, 38.5), time: 2023-10-01 10:30}.

[0111] Example 2: To improve the long-term learning and evolution capabilities of the entire early warning method, the following technical solution is proposed based on Example 1. In addition to real-time correction, a deeper optimization of the model itself is added. The specific solution is as follows:

[0112] The internal weights of the adaptive learning model are trained using the corrected simulation parameters to optimize the prediction accuracy of the adaptive learning model for the pollutant diffusion pattern.

[0113] Based on the optimized adaptive learning model, the generation rules of the virtual monitoring points are dynamically adjusted.

[0114] The adaptive learning model is built on a convolutional neural network, taking a pollution situation field as input and outputting a prediction error.

[0115] 100 historical data points were retrieved from the database as training data for the adaptive learning model. The adaptive learning model was trained using mean squared error as the loss function and backpropagation was used during the training process. The learning rate was 0.001.

[0116] Example 3: To optimize the long-term performance of the entire early warning system, this example proposes the following technical solution based on Example 2:

[0117] After the heavy metal pollution level warning signal is issued, collect real-time data from confirmatory monitoring and feedback data including on-site investigation results;

[0118] By utilizing feedback data, the accuracy of early warnings and the efficiency of response are evaluated to obtain evaluation results;

[0119] Based on the assessment results, the construction parameters of the pollution situation field and the generation rules of virtual monitoring points were adjusted.

[0120] Specifically, in this embodiment, after the event concludes, two key types of feedback data are archived and collected. The first type is the acquired confirmatory real-time monitoring data, i.e., the ground truth values ​​collected by mobile and fixed monitoring units during the event. The second type is the feedback data after the issuance of the warning signal, which is a type of a posteriori verification information. Its sources include the analysis results of samples collected on-site by the emergency response team, assessment reports on the ecological impact of the pollution event, or the actual pollution range interpreted through high-resolution remote sensing imagery.

[0121] After collecting complete feedback data, an evaluation process is initiated to quantify the overall performance of this early warning event. This process first uses the feedback data to assess the accuracy of the early warning. The predicted pollution situation at the last moment before the warning was issued is spatially and numerically compared with the actual pollution situation constructed from the feedback data to calculate an accuracy index. ;in, This indicates the accuracy of the early warning system; the closer the value is to 1, the higher the accuracy. It is the predicted distribution of pollutant concentrations during the early warning period. This is a reconstruction of the true concentration distribution based on feedback data. The integral is applied across the entire affected area. It will be carried out internally.

[0122] If the assessment results show an accuracy index for early warning If the accuracy falls below the preset threshold, the prediction error for this event will be used as a strong feedback signal to adjust the construction parameters of the pollution situation field, including retraining the prediction model or modifying the weights of different data sources in the fusion process.

[0123] For example, feedback data includes field samples. , ,area It is 5 square kilometers. Calculation. . If the value is less than the preset threshold of 0.95, the prediction model needs to be adjusted. middle The weight (e.g., increase by 10%).

[0124] The present invention also provides a computer-readable storage medium storing a computer program, which is executed by a processor to implement the aforementioned method for early warning of heavy metal pollution levels for marine ecosystems.

[0125] The processes described above with reference to the flowcharts in the embodiments disclosed in this invention can be implemented as computer software programs. The embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), it performs the functions defined in the methods of this application. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wire segments, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless segments, wire segments, optical fibers, RF, etc., or any suitable combination thereof.

[0126] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation that may be implemented in systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0127] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been shown and explained in the embodiments. Without departing from the principles described, the implementation of the present invention may have any changes or modifications.

Claims

1. A method for early warning of heavy metal pollution levels in marine ecosystems, characterized in that, The method includes: Acquire target data and perform fusion processing on the target data to generate a pollution situation field; Based on the pollution situation field, high uncertainty areas and high-risk evolution areas are identified, and virtual monitoring points are generated in the high uncertainty areas and high-risk evolution areas; Based on the virtual monitoring points, the mobile monitoring unit is dispatched to conduct verification monitoring and the fixed monitoring points are activated. Acquire real-time verification monitoring data collected by mobile monitoring units and fixed monitoring points, and use the real-time verification monitoring data to correct the pollution situation field and generate an updated pollution situation field. Based on the updated pollution situation field, a heavy metal pollution level early warning signal is generated. The process of identifying high-uncertainty areas and high-risk evolution areas based on the pollution situation field, and generating virtual monitoring points in these areas, includes: The pollution situation field is analyzed to extract pollutant concentration distribution and diffusion trend parameters; the diffusion trend parameters are the spatiotemporal gradient vectors of the pollution situation field. ;in Indicates the pollution situation; Based on pollutant concentration distribution and diffusion trend parameters, regional uncertainty indicators and risk evolution indicators are calculated. Among them, diffusion rate Directional weighting factor ; Indicates the angle between the direction of pollutant diffusion and the direction of the sensitive ecological zone; Indicates the weighting coefficient; based on regional uncertainty indicators and risk evolution indicators, identifies high uncertainty regions and high-risk evolution regions; Within high-uncertainty and high-risk evolution zones, virtual monitoring points are generated, containing location information and monitoring priorities. The location information includes the latitude and longitude of the uncertain and high-risk evolution zones; the monitoring priorities... Among them, regional uncertainty indicators ; The value of the spatiotemporal gradient vector; The high uncertainty region is the region with a regional uncertainty index greater than 0.7, and the high risk evolution region is the region with a risk evolution index greater than 0.

8.

2. The method for early warning of heavy metal pollution levels for marine ecology according to claim 1, characterized in that, The process of acquiring target data and fusing the target data to generate a pollution situation field includes: Acquire real-time monitoring data containing heavy metal concentration data and sensor location information; Acquire ocean hydrological data including tidal current velocity and water temperature parameters; Obtain meteorological data including wind speed and wind direction parameters; Acquire pollution source data including industrial emission locations and river input flow parameters; Spatiotemporal alignment and normalization are performed on real-time monitoring data, marine hydrological data, meteorological data, and pollution source data to generate a fused dataset; By utilizing fused datasets and performing dynamic simulations through predictive models, a pollution state field is generated. ; The prediction model is built on a long short-term memory neural network, with the input being a fused dataset and the output being a pollution state field; the spatiotemporal alignment uses time interpolation, and the normalization process uses the Min-Max scaling algorithm.

3. The method for early warning of heavy metal pollution levels for marine ecology according to claim 1, characterized in that, The step of scheduling mobile monitoring units to perform verification monitoring based on virtual monitoring points and waking up fixed monitoring points includes: Obtain the location information and monitoring priority of virtual monitoring points; Based on location information and monitoring priority, a mobile monitoring unit scheduling instruction is generated; Based on the scheduling instructions of the mobile monitoring unit, control the mobile monitoring unit to go to the virtual monitoring point to perform verification monitoring; Simultaneously, based on the spatial distribution of the virtual monitoring points, nearby fixed monitoring points with hibernation functions are awakened, and their acquisition frequency is adjusted.

4. The method for early warning of heavy metal pollution levels for marine ecology according to claim 3, characterized in that, The step of using real-time data from confirmatory monitoring to correct the pollution situation field and generate an updated pollution situation field includes: Receive real-time verification monitoring data and perform quality verification and anomaly filtering on it using the Z-score anomaly detection algorithm to generate a calibration dataset; The calibration dataset is input into a data assimilation algorithm that can correct model parameters to generate corrected simulation parameters; the data assimilation algorithm is an ensemble Kalman filter algorithm. The corrected simulation parameters are input into the pre-trained prediction model to generate an updated pollution situation field.

5. The method for early warning of heavy metal pollution levels for marine ecology according to claim 4, characterized in that, The generation of heavy metal pollution level early warning signals based on the updated pollution situation field includes: Analyze the updated pollution situation field and extract parameters such as peak concentration and diffusion range of heavy metal pollutants; Based on the peak concentration and diffusion range parameters of heavy metal pollutants, and combined with a rule matrix containing pollution level thresholds, the warning level is calculated. Based on the warning level, a heavy metal pollution level warning signal is generated and sent to the emergency platform through the corresponding alarm equipment. The graded early warning signal includes the early warning level, peak concentration, diffusion range, and timestamp; the rule matrix is... The row vector elements in the rule matrix represent the warning levels corresponding to different extended ranges when the same peak concentration range is reached.

6. The method for early warning of heavy metal pollution levels for marine ecology according to claim 5, characterized in that, Also includes: The internal weights of the adaptive learning model are trained using the corrected simulation parameters to optimize the prediction accuracy of the adaptive learning model for the pollutant diffusion pattern. Based on the optimized adaptive learning model, the generation rules of the virtual monitoring points are dynamically adjusted. The adaptive learning model is built on a convolutional neural network, taking a pollution situation field as input and outputting a prediction error; the training process is carried out through backpropagation with a learning rate of 0.

001.

7. The method for early warning of heavy metal pollution levels for marine ecology according to claim 6, characterized in that, The step of generating mobile monitoring unit scheduling instructions based on location information and monitoring priority includes: Obtain the current status and location parameters of the mobile monitoring unit; By combining the monitoring priority of virtual monitoring points with the current status and location parameters of mobile monitoring units, the scheduling path and resource allocation scheme are calculated. Based on the scheduling path and resource allocation scheme, a mobile monitoring unit scheduling instruction is generated.

8. A method for early warning of heavy metal pollution levels for marine ecology according to claim 7, characterized in that, The method further includes: After the heavy metal pollution level warning signal is issued, collect real-time data from confirmatory monitoring and feedback data including on-site investigation results; By utilizing feedback data, the accuracy of early warnings and the efficiency of response are evaluated to obtain evaluation results; Based on the assessment results, the construction parameters of the pollution situation field and the generation rules of virtual monitoring points were adjusted.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which is executed by a processor to implement the heavy metal pollution level early warning method for marine ecology as described in any one of claims 1-8.

Images