Search

CN-121999918-A - Microplastic distribution prediction and tracing method

CN121999918ACN 121999918 ACN121999918 ACN 121999918ACN-121999918-A

Abstract

The application discloses a method for predicting and tracing the distribution of microplastic, and relates to the field of model detection. The method comprises the steps of collecting multisource monitoring data of a target water area, carrying out space-time alignment processing on the multisource monitoring data to generate a structured space-time data set, constructing a three-dimensional ocean current dynamic model of the target water area, simulating ocean current field distribution under different space-time conditions by ocean environment auxiliary data, extracting ocean current dynamic characteristic data, processing the space-time data set and the ocean current dynamic characteristic data through a micro-plastic prediction sub-model, predicting and outputting micro-plastic diffusion paths and micro-plastic concentration predicted values of all monitoring stations, processing current micro-plastic distribution data and ocean current dynamic characteristic data through a micro-plastic tracing sub-model, and predicting and outputting current micro-plastic distribution retrospective sea inlet pollution sources. The scheme realizes the prediction of the micro-plastic diffusion path and the tracing of the pollution source, can accurately predict the micro-plastic diffusion path and trace the pollution source, and improves the efficiency and the accuracy of marine micro-plastic pollution control.

Inventors

  • ZOU BOFU

Assignees

  • 邹博夫

Dates

Publication Date
20260508
Application Date
20260121

Claims (10)

  1. 1. A method for predicting and tracing the distribution of microplastic, which is characterized by comprising the following steps: Collecting multisource monitoring data of a target water area, and carrying out space-time alignment processing on the multisource monitoring data to generate a structured space-time data set, wherein the multisource monitoring data comprise marine monitoring site data of microplastic, river entrance sea mouth monitoring data and marine environment auxiliary data; Constructing a three-dimensional ocean current dynamics model of a target water area, simulating ocean current field distribution under different air-space conditions by using the ocean environment auxiliary data, and extracting ocean current dynamic characteristic data; The method comprises the steps of processing the space-time data set and the ocean current dynamic characteristic data through a micro-plastic prediction sub-model, predicting and outputting a micro-plastic diffusion path and micro-plastic concentration predicted values of each monitoring station in future target time, and processing current micro-plastic distribution data and the ocean current dynamic characteristic data through a micro-plastic tracing sub-model, and predicting and outputting current micro-plastic distribution traced sea inlet pollution sources.
  2. 2. The method of claim 1, wherein the marine monitoring site data comprises site location, monitoring time, microplastic concentration, microplastic particle size and morphology, the river estuary monitoring data comprises estuary location, discharge flux, microplastic type, the marine environmental assistance data comprises ocean current velocity, tidal cycle, water temperature, salinity and wind speed; The current microplastic distribution data is obtained based on a marine monitoring station sensor array, a river entrance sea mouth monitoring terminal and a data receiving unit, wherein the marine monitoring station sensor array is used for collecting microplastic concentration, particle size, morphology, water temperature and salinity data in real time, the river entrance sea mouth monitoring terminal is used for collecting entrance sea mouth microplastic discharge flux and type data, and the data receiving unit is used for receiving data of all monitoring devices through a wireless communication module and storing the data in a local database.
  3. 3. The method of claim 1, wherein the microplastic predictor model and the microplastic traceback model are constructed based on a trend perceptive graph neural network; After the multisource monitoring data are acquired, constructing graph structure data of the ocean area based on space coordinates, taking a monitoring site and a river sea entrance as graph nodes, taking standardized monitoring data at positions corresponding to the graph nodes as node characteristics, and constructing the graph structure data; When the micro plastic predictor model and the micro plastic traceability sub model are trained, based on constructed graph structure data as input parameters, the spatial dependence relationship of micro plastic concentration along the node migration diffusion of the graph structure and the influence of micro plastic pollution sources on a ocean current path is constructed.
  4. 4. A method according to claim 3, wherein constructing graph structure data comprises: acquiring the space coordinates of all nodes, and calculating the space distance between every two nodes; Determining K adjacent nodes of each node based on the space distance of each node and the ocean current connectivity, and constructing node edges according to the flow direction; Node edge weights The method is calculated according to the following formula: Wherein, the Representing the spatial distance between adjacent nodes, Representing the current speed of the previous node pointing to the next node, Indicating the maximum current velocity within the region, Representing the weight coefficient.
  5. 5. A method according to claim 3, wherein said performing a spatiotemporal alignment process on said multisource monitoring data generates a structured spatiotemporal dataset comprising: abnormal data is removed through a 3 sigma criterion, and a linear interpolation method is adopted to supplement a missing value; For the data sequence of the selected monitoring index, the arithmetic average value is calculated firstly And standard deviation Then determine each data point Whether or not to meet If it meets the requirement, then look at Removing abnormal values; for the missing values in the time sequence or the space sequence, two adjacent points And Estimating a point in the middle of the data of (a) (Satisfy the following ) Values of (2) The formula is as follows: Wherein, the As the missing value to be interpolated, For its corresponding time or position coordinates; Mapping all data to a [0,1] interval through min-max standardization; And carrying out space-time matching on the marine monitoring site data, the river entrance data and the environment auxiliary data based on the time stamp and the longitude and latitude coordinates to generate the structured space-time data set.
  6. 6. The method of claim 5, wherein the training process of the predictor model comprises: Taking the space-time data set, the ocean current dynamic characteristic data and the graph structure data as input, taking the historical monitored microplastic concentration distribution data as a label, optimizing model parameters through back propagation, and learning the mapping relation between the microplastic concentration and the multisource data and ocean current characteristics through iterative training of an Adam optimizer.
  7. 7. The method of claim 5, wherein the training process of the traceback sub model comprises: The microplastic distribution data and the corresponding ocean current dynamic characteristic data at the historical moment are used as input, the river entrance microplastic discharge flux and the discharge time before the moment are used as labels, and the inverse mapping relation between the current distribution of the microplastic and the source discharge is learned through iterative training of an Adam optimizer.
  8. 8. The method according to claim 1, wherein after the micro plastic prediction sub-model predicts and outputs the micro plastic diffusion path and the micro plastic concentration predicted value of each monitoring station in the future target time, the area with the concentration exceeding the set threshold is marked as a deposition hot area, and a diffusion path and deposition hot area distribution thermodynamic diagram is generated; And after the microplastic tracing sub-model predicts and outputs the current microplastic distribution and tracing sea entrance pollution sources, generating a migration path tracing diagram of the microplastic from the candidate sources to the current distribution area based on each candidate pollution source and the corresponding pollution contribution value.
  9. 9. The method of claim 5, wherein a three-dimensional ocean current dynamics model of the target ocean area is built based on a finite volume method, and a hexahedral mesh is divided; And inputting the preprocessed water temperature, salinity, wind speed and tide cycle data, solving a Navier-Stokes equation, simulating to obtain ocean current velocity vectors and flow direction data at different moments, and outputting the ocean current dynamic characteristic data.
  10. 10. The method of claim 6 or 7, wherein the microplastic predictor model and the microplastic traceback model construct a composite loss function based on cross entropy loss and mean square error The following is indicated: Wherein, the For cross entropy loss, the method is used for classifying tasks (such as judging contribution types of candidate sources when tracing the sources); for mean square error loss, for regression tasks (e.g., predicting concentration values of microplastic); l1 regularization-based construction and model parameters A penalty term proportional to the sum of absolute values of (a) to form an optimized objective function : Wherein, the Is a super-parameter that controls the strength of the regularization, Is the sum of the absolute values of the model ownership parameters.

Description

Microplastic distribution prediction and tracing method Technical Field The application relates to the field of model detection, in particular to a micro-plastic distribution prediction method. Background In recent years, microplastic is a typical tiny particle pollutant, widely exists in water environments such as oceans, rivers and the like, and forms a serious threat to ecological system balance and human health. The diffusion path and the deposition hot zone distribution of the microplastic in the ocean are precisely mastered, and pollution sources of the microplastic are traced, so that the microplastic is a precondition for formulating an effective prevention and control strategy. In the related art, the monitoring technology for water environment focuses on detection and evaluation of water quality parameters (such as pH value, dissolved oxygen, heavy metal content and the like), and related patent achievements mainly surround optimization of water quality monitoring equipment and simple prediction development of water quality change trend. On the one hand, the related technology lacks deep analysis of a diffusion mechanism of the micro-plastic micro-particle pollutants in a complex marine environment, and is difficult to accurately simulate the influence of dynamic factors such as ocean currents, tides and the like on the migration and diffusion of the micro-plastic. On the other hand, most of the existing prediction models are one-way diffusion trend prediction, and reverse deduction based on the current distribution state cannot be realized, namely, the pollution source and migration path of the microplastic cannot be traced back from the current distribution of the microplastic in the ocean, so that the pollution prevention and control work lacks pertinence, and the diffusion and the spread of the microplastic pollution are difficult to be restrained from the source. In addition, the conventional micro-plastic migration simulation technology depends on a traditional fluid dynamic model, only can consider environmental factors of a single dimension, and space-time correlation characteristics of multi-source monitoring data are not fully fused, so that the prediction accuracy is low, the generalization capability is insufficient, and the actual requirements of micro-plastic distribution prediction and tracing in a complex marine environment cannot be met. Therefore, development of a microplastic distribution prediction and tracing system capable of integrating multi-source data, accurately simulating a diffusion mechanism and realizing reverse tracing becomes a technical problem to be solved in the current marine environment treatment field. Disclosure of Invention The application provides a microplastic distribution prediction and tracing method, which can accurately predict the microplastic diffusion path and trace the pollution source, and improve the efficiency and accuracy of marine microplastic pollution control. Collecting multisource monitoring data of a target water area, and carrying out space-time alignment processing on the multisource monitoring data to generate a structured space-time data set, wherein the multisource monitoring data comprise marine monitoring site data of microplastic, river entrance sea mouth monitoring data and marine environment auxiliary data; Constructing a three-dimensional ocean current dynamics model of a target water area, simulating ocean current field distribution under different air-space conditions by using the ocean environment auxiliary data, and extracting ocean current dynamic characteristic data; The method comprises the steps of processing the space-time data set and the ocean current dynamic characteristic data through a micro-plastic prediction sub-model, predicting and outputting a micro-plastic diffusion path and micro-plastic concentration predicted values of each monitoring station in future target time, and processing current micro-plastic distribution data and the ocean current dynamic characteristic data through a micro-plastic tracing sub-model, and predicting and outputting current micro-plastic distribution traced sea inlet pollution sources. Specifically, the marine monitoring site number comprises site position, monitoring time, microplastic concentration, microplastic particle size and form; the river estuary monitoring data comprise estuary positions, discharge flux and micro-plastic types, and the marine environment auxiliary data comprise ocean current speed, tidal cycle, water temperature, salinity and wind speed; The current microplastic distribution data is obtained based on a marine monitoring station sensor array, a river entrance sea mouth monitoring terminal and a data receiving unit, wherein the marine monitoring station sensor array is used for collecting microplastic concentration, particle size, morphology, water temperature and salinity data in real time, the river entrance sea mouth monitoring terminal is used for col