Method, device, equipment and medium for predicting microplastic transport paths based on complex hydrodynamic environments

Abstract

This application discloses a method, device, equipment, and medium for predicting the transport path of microplastics based on complex hydrodynamic environments, involving the fields of environmental monitoring and water pollution control technology. The method integrates a watershed digital elevation model, a river vector diagram, microplastic properties, and real-time hydrological data. The method then uses a dynamic adjacency matrix to capture changes in water flow direction, combines multimodal feature fusion to model the physical properties of microplastics and terrain resistance, and uses a cross-basin transfer learning strategy to address the scarcity of monitoring data. The method accurately predicts the transport path of microplastics, achieves high-precision simulation of microplastic migration in complex hydrodynamic environments, and improves the efficiency of pollution prevention and control.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring and water pollution control, and in particular to a method, device, equipment and medium for predicting the migration path of microplastics based on a complex hydrodynamic environment. Background Art

[0002] Microplastics (plastic particles smaller than 5 mm in size) are an emerging environmental pollutant and have been widely detected in rivers, lakes, and marine ecosystems. Due to their persistence, bioaccumulation, and ecotoxicity, microplastics pose a potential threat to aquatic ecosystems and human health. They originate from diverse sources, including cosmetics, textile fibers, and the degradation of agricultural plastic films. Their transport pathways are influenced by multiple factors, including hydrodynamics, topography, and particle characteristics.

[0003] Current research on predicting microplastic transport pathways in complex hydrodynamic environments focuses on three main technologies: fluid dynamics models based on physical mechanisms, data-driven statistical learning models, and hybrid models that fuse physics and data. Fluid dynamics models use the Navier-Stokes equations and particle settling formulas to simulate water flow and construct river channel topology networks in conjunction with GIS data. Data-driven models, such as LSTM and random forests, use machine learning to extract features from historical data to improve computational efficiency. Hybrid models attempt to combine the advantages of both approaches, for example by using fluid dynamics equations as neural network constraints to improve prediction accuracy and adaptability.

[0004] Despite recent progress, existing technologies still face limitations in several areas. First, while fluid dynamics models can accurately simulate water flow, they rely on the complex Navier-Stokes equations, requiring high-performance computing resources in three-dimensional simulation scenarios and struggling to meet the demands of real-time response to sudden pollution incidents. Furthermore, these models assume microplastics are homogeneous spherical particles and fail to fully account for the influence of varying materials and morphologies. Second, data-driven models excel at processing time series or local spatial features, but are limited in their ability to express complex topological relationships between upstream and downstream watersheds and between tributaries and the mainstream. Their "black-box" nature limits their application in pollution control decision-making. Furthermore, while hybrid models attempt to combine physical modeling with data-driven approaches, their coupling between physics and data is relatively crude, resulting in unstable training processes and a lack of dynamic weight adjustment mechanisms, making it difficult to balance the computational weights between real-time hydrological data and long-term physical laws. Therefore, an efficient and accurate path prediction method is urgently needed to predict microplastic transport pathways in complex hydrodynamic environments. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device, equipment and medium for predicting the microplastic transport path based on a complex hydrodynamic environment, aiming to solve the technical problem of how to improve the accuracy of microplastic path prediction in a complex hydrodynamic environment.

[0006] To achieve the above objectives, this application proposes a method for predicting microplastic transport pathways based on a complex hydrodynamic environment, including:

[0007] Obtain watershed digital elevation model data, river channel vector maps, microplastic properties, and hydrological data;

[0008] Dividing the watershed digital elevation model data and the river channel vector map to obtain a node set, wherein the node includes elevation, slope, river channel bifurcation point and watershed outlet;

[0009] Performing calculations based on the hydrological data and the watershed digital elevation model data to generate a dynamic adjacency matrix;

[0010] The microplastic properties are integrated with the watershed topographic features to obtain integrated topographic features, wherein the watershed topographic features are obtained based on the watershed digital elevation model data and the river channel vector map, and the watershed topographic features include slope data, roughness coefficient, and river channel curvature;

[0011] Calculating based on the dynamic adjacency matrix and the fused terrain features to obtain a target microplastic concentration;

[0012] Determining a concentration change direction and a concentration change intensity based on the target microplastic concentration;

[0013] Determining a transport node and a transport distance based on the concentration change direction and the concentration change intensity;

[0014] When the transfer node is a watershed outlet in the node set, the transfer paths are combined to obtain a target transfer path prediction result.

[0015] In one embodiment, the step of performing calculations based on the hydrological data and the watershed digital elevation model data to generate a dynamic adjacency matrix includes:

[0016] Calculating by using the hydrological data to obtain flow direction similarity;

[0017] The flow direction consistency between two adjacent nodes is calculated based on the flow direction similarity. The specific formula is:

[0018]

[0019] in, and Node and nodes The flow rate, is the flow direction similarity;

[0020] The terrain resistance is obtained according to the digital elevation model data of the watershed. The specific formula is:

[0021]

[0022] in, For nodes and The elevation difference between For nodes and The horizontal space distance between is the resistance factor;

[0023] According to the flow direction consistency and the terrain resistance, a dynamic adjacency matrix is obtained, and the specific formula is:

[0024]

[0025] in is the flow direction consistency, is the terrain resistance, Represents an indicative function that only allows adjacent nodes in the downstream direction to participate in the calculation.

[0026] In one embodiment, the step of fusing the microplastic properties with the watershed topographic features to obtain fused topographic features, wherein the watershed topographic features are obtained based on the watershed digital elevation model data and the river channel vector map, and the watershed topographic features include slope data, roughness coefficient, and river channel curvature, comprises:

[0027] The basin terrain features are encoded to obtain the initial feature vector, which is specifically expressed as:

[0028]

[0029] in, is the slope data, is the roughness coefficient, is the river channel curvature;

[0030] The initial feature vector is processed by a double-layer GCN to obtain the first feature vector and the second feature vector. The specific formula is:

[0031]

[0032]

[0033] in, and Represent the first layer weight and the second layer weight of GCN respectively, For nodes The set of first-order neighbors of ;

[0034] The microplastic properties are encoded to obtain the property coding characteristics. The specific formula is:

[0035]

[0036] represents the density of microplastics, is the shape factor, Represents a function that converts a material category into a numerical value;

[0037] Based on the attribute coding feature and the second feature vector, a fused terrain vector is obtained. The specific formula is:

[0038]

[0039]

[0040]

[0041] in, are the mapping matrices of terrain features, bond vectors and microplastic properties, is the key vector dimension, is the attention weight, represents the normalization function, is the fusion feature vector;

[0042] A fused terrain feature is obtained based on the fused terrain vector.

[0043] In one embodiment, the step of calculating based on the dynamic adjacency matrix and the fused terrain features to obtain a target microplastic concentration includes:

[0044] Based on the dynamic adjacency matrix and the fused terrain features, the microplastic concentration is updated and calculated through the graph propagation algorithm. The specific formula is:

[0045]

[0046] in, For nodes In time Prediction of microplastic concentrations, is the learnable weight matrix;

[0047] Calculating based on the watershed digital elevation model data and the river vector diagram to obtain the water volume corresponding to the node;

[0048] Obtaining a flow rate from a node to a corresponding adjacent node according to the hydrological data;

[0049] Initialize update weights;

[0050] The concentration loss value is obtained by combining the water volume and the flow rate with the microplastic concentration through the mass conservation equation, and the formula is:

[0051]

[0052] in, For nodes The corresponding water volume, For nodes arrive The flow rate, Representation node arrive Flow rate;

[0053] The concentration loss value is updated until the update weight reaches a preset value to obtain a target loss value;

[0054] The target loss value and the microplastic concentration are combined for calculation to obtain the target microplastic concentration.

[0055] In one embodiment, the step of determining the concentration change direction and the concentration change intensity according to the target microplastic concentration includes:

[0056] determining a predicted concentration distribution based on the target microplastic concentration;

[0057] The concentration gradient of the node is calculated based on the predicted concentration distribution to determine the concentration change direction and concentration change intensity. The formula is:

[0058]

[0059]

[0060] in, represents the predicted concentration distribution, Representation node The position vector of and are the horizontal and vertical directions of the spatial coordinate axes.

[0061] In one embodiment, before the step of determining the transport node and the transport distance based on the concentration change direction and the concentration change intensity, the step includes:

[0062] Obtaining target watershed data, wherein the target watershed data includes microplastic concentration and terrain attribute data of the target watershed;

[0063] The target watershed data of a preset proportion is selected and the Kroginli spatial interpolation algorithm is used to generate pseudo labels for unlabeled watersheds. The specific formula is:

[0064]

[0065] in, represents the Kriging spatial interpolation algorithm, represents the concentration of microplastics in the target watershed, Represents the terrain attribute data of the target watershed;

[0066] Evaluating the pseudo labels of the unlabeled watershed to obtain corresponding confidence scores;

[0067] When the confidence level exceeds a preset confidence level, obtaining unlabeled watershed data corresponding to the confidence level;

[0068] An optimization strategy is used for the target watershed data and the unlabeled watershed data to obtain a loss value. The specific formula is:

[0069]

[0070] in, represents the maximum mean difference loss value, represents the mean square error loss value, represents the trade-off coefficient, Indicates unlabeled watershed data;

[0071] The loss value is updated until the trade-off coefficient reaches a preset value, and the step of determining the concentration change direction and concentration change intensity based on the target microplastic concentration is performed.

[0072] In one embodiment, the step of determining the transport node and the transport distance based on the concentration change direction and the concentration change intensity includes:

[0073] Determine, according to the concentration change direction, a set of downstream adjacent nodes that is consistent with the change direction;

[0074] Selecting a node corresponding to a preset concentration change intensity in the concentration change intensity from the downstream adjacent node set as a transfer node;

[0075] The transport distance is obtained based on the watershed digital elevation model data and the river vector diagram in combination with the current node and the transport node.

[0076] In addition, to achieve the above objectives, the present application also proposes a device for predicting the transport path of microplastics based on a complex hydrodynamic environment, the device comprising:

[0077] The acquisition module is used to obtain watershed digital elevation model data, river vector diagrams, microplastic properties and hydrological data;

[0078] A partitioning module is used to partition the watershed digital elevation model data and the river channel vector map to obtain a node set, wherein the node includes elevation, slope, river channel bifurcation point and watershed outlet;

[0079] A calculation module, configured to perform calculations based on the hydrological data and the watershed digital elevation model data to generate a dynamic adjacency matrix;

[0080] A fusion module is used to fuse the microplastic properties with the watershed topographic characteristics to obtain fused topographic characteristics, wherein the watershed topographic characteristics are obtained based on the watershed digital elevation model data and the river channel vector map, and the watershed topographic characteristics include slope data, roughness coefficient, and river channel tortuosity;

[0081] A calculation module is further used to calculate based on the dynamic adjacency matrix and the fused terrain features to obtain a target microplastic concentration;

[0082] The calculation module is further used to determine the concentration change direction and concentration change intensity according to the target microplastic concentration;

[0083] The calculation module is further used to determine the transport node and the transport distance based on the concentration change direction and the concentration change intensity;

[0084] The result module is used to combine the transport paths when the transport node is a watershed outlet in the node set to obtain a target transport path prediction result.

[0085] In addition, to achieve the above-mentioned purpose, the present application also proposes a medium, which is a computer-readable medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the microplastic transport path prediction method based on a complex hydrodynamic environment as described above are implemented.

[0086] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the microplastic transport path prediction method based on a complex hydrodynamic environment as described above.

[0087] This application obtains watershed digital elevation model data, river vector maps, microplastic properties, and hydrological data, divides the watershed digital elevation model data and river vector maps to obtain a node set, performs calculations based on the hydrological data and watershed digital elevation model data, generates a dynamic adjacency matrix, fuses the microplastic properties with the watershed terrain characteristics, obtains fused terrain characteristics, and then calculates based on the dynamic adjacency matrix and fused terrain characteristics to obtain the target microplastic concentration. The concentration change direction and concentration change intensity are determined based on the target microplastic concentration, and the transport node and transport distance are determined based on the concentration change direction and concentration change intensity. Finally, when the transport node is the watershed outlet in the node set, the transport distance is combined to obtain the target transport path prediction result. By integrating the watershed digital elevation model, river vector maps, microplastic properties, and real-time hydrological data, and using the dynamic adjacency matrix and fused terrain characteristics, the microplastic transport path is accurately predicted, achieving high-precision simulation of microplastic migration in complex hydrodynamic environments and improving pollution prevention and control efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0089] Figure 1 This is a flow chart of the first embodiment of the method for predicting microplastic transport pathways based on complex hydrodynamic environments of this application;

[0090] Figure 2 This is a flow chart of the second embodiment of the method for predicting microplastic transport pathways based on complex hydrodynamic environments of this application;

[0091] Figure 3 This is a flow chart of the third embodiment of the method for predicting microplastic transport pathways based on complex hydrodynamic environments of this application;

[0092] Figure 4 This is a schematic diagram of the module structure of the device for predicting microplastic transport paths based on complex hydrodynamic environments according to the first embodiment of the method for predicting microplastic transport paths based on complex hydrodynamic environments of the present application;

[0093] Figure 5 Schematic diagram of the device structure of the hardware operating environment involved in the method for predicting microplastic transport paths based on complex hydrodynamic environments in an embodiment of the present application.

[0094] The purpose, features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0095] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0096] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0097] Globally, microplastics (plastic particles smaller than 5 mm in size) are a new environmental pollutant and have become widespread in rivers, lakes, and marine ecosystems. Due to their persistence, bioaccumulation, and potential ecotoxicity, microplastics not only absorb harmful substances such as persistent organic pollutants and heavy metals but can also be transmitted to humans through the food chain, posing a serious threat to aquatic ecosystems and human health. However, current research on the transport patterns of microplastics faces numerous challenges, particularly in tracing their source, predicting accuracy, and addressing response issues. Traditional methods based on fixed-point sampling struggle to dynamically track the sources and migration pathways of microplastics across the entire river basin. Existing prediction models primarily rely on empirical fluid dynamics formulas and employ static hydrological parameters for simulation. These models struggle to accurately reflect the dynamic impacts of sudden rainfall and runoff changes, hydraulic infrastructure regulation, and other factors on water flow direction and velocity, limiting prediction accuracy. Furthermore, in the absence of precise transport path prediction technology, pollution interception measures often lag behind. This is particularly true during sudden pollution events such as rainstorms, where high-precision pollutant dispersion predictions cannot be provided in a timely manner, hindering the effectiveness of pollution prevention and control.

[0098] Therefore, in order to overcome the above problems, the present application proposes an efficient and accurate method for predicting the transport path of microplastics based on a complex hydrodynamic environment. The main solution of the embodiment of the present application is: obtain the digital elevation model data of the watershed, the river vector map, the microplastic properties and the hydrological data, divide the digital elevation model data of the watershed and the river vector map to obtain a node set, calculate based on the hydrological data and the digital elevation model data of the watershed, generate a dynamic adjacency matrix, fuse the microplastic properties and the watershed terrain characteristics to obtain a fused terrain feature, and then calculate based on the dynamic adjacency matrix and the fused terrain feature to obtain the target microplastic concentration, determine the concentration change direction and concentration change intensity according to the target microplastic concentration, determine the transport node and transport distance based on the concentration change direction and concentration change intensity, and finally, when the transport node is the watershed outlet in the node set, combine the transport distance to obtain the target transport path prediction result.

[0099] Based on the above, the present application embodiment also provides a method for predicting the transport path of microplastics based on a complex hydrodynamic environment, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the method for predicting microplastic transport paths based on complex hydrodynamic environments of this application.

[0100] In this embodiment, the method for predicting the microplastic transport path based on a complex hydrodynamic environment includes steps S10 to S80:

[0101] Step S10: Obtain watershed digital elevation model data, river vector map, microplastic properties and hydrological data.

[0102] It is important to note that watershed digital elevation model (DEM) data provides detailed topographic information, including elevation and slope, which is crucial for understanding water flow direction and velocity. By analyzing high-resolution DEM data (resolution ≤ 10 m), it is possible to precisely delineate watershed boundaries and identify key nodes such as elevation points, slope aspects, channel bifurcations, and outlets, laying the foundation for the subsequent construction of a dynamic adjacency matrix. River channel vector maps provide network structure information for river systems, helping to understand flow paths and channel connectivity. Incorporating natural channel connectivity, an initial edge set can be established to further refine flow patterns within the watershed. Furthermore, microplastic properties, including physical parameters such as particle density, shape factor, and material classification (e.g., PE / PET), directly influence their migration behavior in water, such as settling rate and suspension time. Accurately encoding these properties and integrating them with topographic features allows for more realistic simulation of microplastic migration under complex hydrodynamic conditions. Hydrological data, such as flow rate, rainfall, and tidal changes, are used to update the flow state within the basin in real time, reflecting the impact of hydrological dynamics on the transport pathways of microplastics. Using this data to generate a dynamic adjacency matrix can capture changes in flow direction over time, enabling accurate simulation of the microplastic migration process.

[0103] Step S20 : dividing the watershed digital elevation model data and the river channel vector map to obtain a node set.

[0104] It should be noted that by loading high-resolution DEM data (resolution ≤ 10 meters), we can extract the topographic information of the basin, including altitude, slope, and aspect. These geographical attributes are important for understanding the direction of water flow, flow rate, and potential sedimentation areas. In order to further refine the study area, the Voronoi algorithm can be used to process the DEM data to generate a node set. , each node This process helps accurately capture the changes in hydrological conditions within the basin, providing a foundation for subsequent modeling. Furthermore, the river vector diagram provides information on the network structure of the river system, including its starting points, end points, branching points, and their connections.

[0105] Furthermore, based on this information, an initial edge set can be established to represent the natural river connection between each node. The specific formula is:

[0106]

[0107] in, is a binary attribute of the static adjacency matrix. Specifically, if there is a direct waterway connection between two adjacent nodes, the corresponding edge is marked as 1; conversely, if there is no direct connection, it is marked as 0. This method constructs a static adjacency matrix that provides a preliminary description of the flow pattern within the watershed. The initial edge set is obtained by combining multiple static adjacency matrices.

[0108] Step S30 , performing calculations based on the hydrological data and the watershed digital elevation model data to generate a dynamic adjacency matrix.

[0109] It should be noted that the dynamic adjacency matrix accurately reflects the dynamic changes in water flow direction and velocity by fusing real-time hydrological data with DEM data. There are two main factors to consider when constructing a dynamic adjacency matrix: flow direction similarity calculation and terrain resistance. The flow direction similarity calculation relies on the velocity data obtained in real time to evaluate the consistency of flow direction between nodes. Specifically, cosine similarity is used to measure the consistency of flow velocity direction between two nodes, and a flow direction consistency is set. If the angle is less than the set value, it is considered that there is unidirectional transmission between the two nodes; otherwise, the weight is forced to be adjusted to zero. In addition, considering the impact of terrain on water flow, it is necessary to introduce terrain resistance. The terrain resistance is generated by combining the elevation difference and horizontal distance with the resistance factor to ensure that water transmission is allowed only in the downstream direction.

[0110] Specifically, the flow direction similarity is calculated through hydrological data, and the flow direction similarity between adjacent nodes is calculated through vector analysis method. The specific formula is:

[0111]

[0112] in and Node and nodes The flow rate, It represents the angle between the flow velocity directions of two nodes. The closer the flow direction similarity is to 1, the more consistent the flow direction is. The flow direction consistency between two adjacent nodes is calculated based on the flow direction similarity. The specific formula is:

[0113]

[0114] in, and Node and nodes The flow rate, is the flow direction similarity;

[0115] The terrain resistance is obtained based on the watershed digital elevation model data. The specific formula is:

[0116]

[0117] in, For nodes and The elevation difference between For nodes and The horizontal space distance between is the resistance factor.

[0118] According to the flow consistency and terrain resistance, the dynamic adjacency matrix is obtained. The sigmoid function is used to map the adjacency weight to the range of (0, 1) to facilitate the weight normalization in subsequent GNN training. The specific formula is:

[0119]

[0120] in is the flow direction consistency, is the terrain resistance, Represents an indicative function, which is used to filter node pairs that meet the conditions and only allow adjacent nodes in the downstream direction to participate in the calculation.

[0121] Step S40: Fusing the microplastic properties with the watershed topographic features to obtain fused topographic features.

[0122] It should be noted that the above-mentioned basin topographic characteristics are obtained based on the basin digital elevation model data and river channel vector diagrams. The basin topographic characteristics include slope data, roughness coefficient and river channel curvature. Slope data reflects the ability of water flow to accelerate, the roughness coefficient quantifies the frictional resistance of the surface to water flow, and the river channel curvature describes the degree of river channel tortuosity, affecting the complexity of water flow and the possibility of turbulence formation. Microplastic properties include key parameters such as particle density, shape factor and material classification (such as PE / PET). These factors directly affect the sedimentation rate and suspension time of microplastics in water bodies. For example, microplastic particles of different densities will sink or float at different speeds; and the shape factor determines their resistance and stability in water flow. Fibrous particles are more easily carried by water flow and remain suspended than spherical particles.

[0123] Specifically, basin topographic features (such as slope, roughness coefficient, and river channel curvature) are encoded as initial feature vectors and processed through a two-layer GCN to capture the spatial dependencies between adjacent nodes. Microplastic properties are also mapped into numerical form, and the settling velocity calculated according to the Stokes formula is incorporated into the model as a physical prior. Next, a multi-head cross-attention mechanism is used to achieve dynamic interaction and adaptive fusion between the two, resulting in a fused terrain vector, which is then converted into fused terrain features.

[0124] Step S50: Calculate based on the dynamic adjacency matrix and fused terrain features to obtain the target microplastic concentration.

[0125] It is important to note that the dynamic adjacency matrix provides a fundamental framework for modeling microplastic migration by capturing real-time changes in flow direction and velocity within a watershed, as well as the influence of topographic resistance. This matrix not only reflects natural river connectivity but also responds to changing hydrological conditions such as rainfall fluctuations and dam regulation. Fusion topographic features are combined with the dynamic adjacency matrix to accurately predict microplastic concentration distributions. The fused topographic features encompass both the watershed's topographic characteristics and microplastic properties, which together influence the migration of microplastics within the water column. A graph neural network (GNN) model, combining the dynamic adjacency matrix and fused topographic features, is used to calculate the target microplastic concentration. Graph propagation calculations use a message passing mechanism to update the predicted microplastic concentration at each node based on the dynamic adjacency matrix and node features. Each message passing step considers information from neighboring nodes and their own physical properties, enabling the model to dynamically adapt to changing hydrological conditions. In addition, to ensure that the prediction results comply with the law of conservation of mass, a physical loss calculation step is introduced to force the model prediction to satisfy the discrete conservation equation to prevent unreasonable physical contradictions. The mass conservation constraint is added as a regularization term in the loss function to supervise and optimize the physical law of concentration change.

[0126] Specifically, the microplastic concentration is obtained by updating the calculation based on the dynamic adjacency matrix and fused terrain features through the graph propagation algorithm. The specific formula is:

[0127]

[0128] in, For nodes In time Prediction of microplastic concentrations, is a learnable weight matrix. Calculations are performed based on the watershed digital elevation model data and river vector diagrams to obtain the water volume corresponding to the node. The flow velocity from the node to the corresponding adjacent node is obtained based on the hydrological data. The weights are initialized and updated. The concentration loss value is obtained by applying the mass conservation equation based on the water volume and flow velocity combined with the microplastic concentration. The formula is:

[0129]

[0130] in, For nodes The corresponding water volume, For nodes arrive The flow rate, Representation node arrive The mass conservation equation accounts for the total amount of microplastics flowing from one node to another. The concentration loss value is updated until the update weight reaches a preset value, resulting in a target loss value. This process ensures the model can dynamically adapt to changing hydrological conditions and provide accurate microplastic concentration predictions. The target loss value and the microplastic concentration are combined for calculation to obtain the target microplastic concentration.

[0131] Step S60: Determine the concentration change direction and concentration change intensity based on the target microplastic concentration.

[0132] It should be noted that after obtaining the target microplastic concentration for each node, the direction of concentration change can be determined by analyzing the concentration differences between adjacent nodes. Specifically, if the microplastic concentration of a node is significantly higher than that of its neighboring nodes, it can be inferred that microplastics are diffusing from high-concentration areas to low-concentration areas, that is, the direction of concentration change is from high-concentration areas to low-concentration areas.

[0133] Specifically, the predicted concentration distribution is determined based on the target microplastic concentration, and then the concentration gradient of the node is calculated based on the predicted concentration distribution to determine the direction and intensity of concentration change. In order to quantify the intensity of this concentration change, the concentration gradient between adjacent nodes can be calculated. The concentration gradient not only reflects the diffusion trend of microplastics in the water body, but also indirectly reveals the influence of factors such as water flow velocity and terrain resistance on the migration of microplastics. The specific formula is:

[0134]

[0135]

[0136] in, represents the predicted concentration distribution, Representation node The position vector of and The larger the absolute value of the concentration gradient, the more obvious the migration trend of microplastics in that direction and the faster the migration speed.

[0137] Step S70: determining the transport node and the transport distance based on the concentration change direction and the concentration change intensity.

[0138] It should be noted that after determining the direction and intensity of concentration change, the transport node can be further selected. In the dynamic adjacency matrix of the watershed, each node has multiple possible downstream nodes. By comparing the concentration gradient direction and intensity of these downstream nodes, the node with the most consistent concentration gradient direction and the greatest intensity with the current node is selected as the target transport node. This process not only takes into account the natural trend of microplastic migration, but also combines the topological structure of the watershed to ensure that the migration path meets the hydrological conditions. At the same time, the calculation of the transport distance is also crucial. Based on DEM data and river vector diagrams, the actual transport distance from the current node to the target node can be accurately calculated. This distance not only reflects the spatial range of microplastic migration, but can also be combined with the intensity of concentration changes to further infer the migration rate and time of microplastics.

[0139] Specifically, the downstream adjacent node set consistent with the concentration change direction is determined according to the concentration change direction, and then the node corresponding to the preset concentration change intensity in the concentration change intensity is selected from the downstream adjacent node set as the transfer node. Specifically, the node with the largest absolute value of the concentration gradient is selected as the transfer node of the current node. Based on the digital elevation model data of the watershed and the river vector diagram, the current node and the transfer node are combined to obtain the transfer distance. Specifically, the transfer distance is determined based on the initial edge set, that is, the natural river connection between each node. When the node and nodes When there is a direct river connection between nodes, the transport distance is the length of the river centerline. and nodes When there is no direct river connection between them, the nodes are calculated using the watershed digital elevation model data. and nodes The shortest path length between nodes can be used to determine the microplastics from To Node transport distance.

[0140] In this way, not only can the migration path of microplastics be tracked, but also their diffusion speed and range in the watershed can be predicted, providing important spatiotemporal information for pollution control.

[0141] Step S80: When the transfer node is a watershed outlet in the node set, the transfer distances are combined to obtain a target transfer path prediction result.

[0142] It is important to note that starting from the initial release point of microplastics, their migration path is tracked by analyzing the direction and intensity of concentration changes at each node until they reach the watershed outlet. During this process, the transport distance between each node is calculated and recorded. The accumulation of these transport distances forms a complete transport path for microplastics from the source to the watershed outlet. By integrating these paths, the migration trajectory of microplastics within the entire watershed can be obtained, including important information such as the key nodes they pass through, the length of the path, and the migration time.

[0143] This embodiment obtains watershed digital elevation model data, river vector maps, microplastic properties, and hydrological data, divides the watershed digital elevation model data and river vector maps to obtain a node set, performs calculations based on the hydrological data and watershed digital elevation model data, generates a dynamic adjacency matrix, fuses the microplastic properties with the watershed terrain characteristics to obtain fused terrain features, and then calculates based on the dynamic adjacency matrix and fused terrain features to obtain a target microplastic concentration. The direction and intensity of concentration change are determined based on the target microplastic concentration, and the transport node and transport path are determined based on the direction and intensity of concentration change. Finally, when the transport node is the watershed outlet in the node set, the transport path is combined to obtain the target transport path prediction result. By integrating the watershed digital elevation model, river vector maps, microplastic properties, and real-time hydrological data, and using the dynamic adjacency matrix and fused terrain features, the microplastic transport path is accurately predicted, achieving high-precision simulation of microplastic migration in complex hydrodynamic environments and improving the efficiency of pollution prevention and control.

[0144] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction and will not be described in detail later. Figure 2 The microplastic transport path prediction method based on a complex hydrodynamic environment, step S40, further includes steps S201 to S205:

[0145] Step S201: Encode the watershed terrain features to obtain an initial feature vector.

[0146] It should be noted that the slope information from the high-resolution DEM data is extracted, that is, the degree of inclination of the surface relative to the horizontal plane. The slope directly affects the speed and direction of water flow. Steep slopes usually accelerate water flow, while gentle areas may cause sedimentation. Next, the roughness coefficient is calculated. This is an indicator that quantifies the roughness of the surface and reflects the amount of friction between the water flow and the riverbed or riverbank. A higher roughness coefficient means greater resistance, which slows down the water flow and affects the migration speed of microplastics. Finally, the curvature of the river channel is evaluated to determine the degree of river tortuosity by measuring the curvature of the centerline of the river channel. Highly curved river channels will lead to complex flow changes and increase the residence time of microplastics in local areas.

[0147] Furthermore, after obtaining the above features, we integrate them to obtain an initial feature vector, which is specifically expressed as:

[0148]

[0149] in, is the slope data, is the roughness coefficient, The curvature of the river channel.

[0150] Step S202: Process the initial feature vector through a double-layer GCN to obtain a first feature vector and a second feature vector.

[0151] It’s important to note that processing these initial feature vectors with a two-layer graph convolutional network (GCN) can capture the complex relationships between adjacent nodes, thereby better simulating actual water flow conditions. GCN is a deep learning method specifically designed for graph-structured data, effectively capturing the spatial correlations between nodes.

[0152] Specifically, in the first layer of GCN processing, the input is the initial feature vector of each node and is linearly transformed through the first layer weights, followed by a nonlinear activation function (such as ReLU), as shown in the following formula:

[0153]

[0154] Next, when processing the second layer of GCN, the second layer weights are used to further refine the features. The formula is:

[0155]

[0156] in, and Represent the first layer weight and the second layer weight of GCN respectively, For nodes This two-layer GCN processing approach can improve understanding of hydrological conditions within a watershed, particularly in areas with complex topography and variable flow patterns. For example, steep slopes may accelerate water flow and carry more microplastic particles, while areas with high roughness coefficients may slow water flow and increase the retention time of microplastics.

[0157] Step S203: Encode the properties of the microplastics to obtain property coding features.

[0158] It should be noted that the properties of microplastics, including density, shape factor, and material classification (such as PE / PET), will affect their migration behavior in water bodies. The density of microplastics determines its floating or sinking state in water: particles with a density less than that of water tend to float, while particles with a density greater than that of water will sink. Secondly, the shape factor reflects the morphology of the particles. Fibrous or flaky microplastics are more easily carried by water flow and remain suspended than spherical particles. In order to convert these physical properties into a form that can be used for model processing, we use a specific function to convert the material category into a numerical form. The specific formula is:

[0159]

[0160] represents the density of microplastics, is the shape factor, Represents a function that converts material categories into numerical values. In addition, the theoretical sedimentation velocity of each microplastic particle is calculated based on the Stokes sedimentation formula and added to the model as a physical prior. The specific formula is:

[0161]

[0162] in, Indicates the sedimentation velocity of microplastic particles, represents the density of microplastic particles, represents the density of water, represents the acceleration due to gravity, Indicates the movement of microplastic particles, represents the dynamic viscosity of water. The above formula shows that the greater the density difference between microplastic particles and water, the faster they settle. The larger the radius of the microplastic particle, the faster it settles, and the greater the viscosity of the water, the slower it settles. This encoding method not only quantifies the basic physical properties of microplastics but also allows the model to adjust its predictions based on these properties, thereby more accurately simulating the migration trajectories of microplastics under complex hydrodynamic conditions.

[0163] Step S204: obtaining a fused terrain vector based on the attribute coding feature and the second feature vector.

[0164] It should be noted that the attribute encoding features and the second feature vector are combined through the multi-head cross attention mechanism. In this process, the geographical features are used as the query and the microplastic attributes are used as the key value. , projecting terrain features, key vectors, and microplastic properties into the same space and calculating attention weights. The purpose of this is to dynamically adjust the weight distribution between different features to capture the most relevant parts. Finally, through a normalization function such as Layer Normalization, a fused feature vector is obtained. The specific formula is:

[0165]

[0166]

[0167]

[0168] in, are the mapping matrices of terrain features, bond vectors and microplastic properties, is the key vector dimension, is the attention weight, represents the normalization function, This fusion method not only integrates information about terrain and microplastic properties, but also enhances the model's expressiveness and adaptability.

[0169] Step S205: obtaining fused terrain features based on the fused terrain vectors.

[0170] It should be noted that the fused vector not only contains information on natural geographical elements such as terrain slope, roughness coefficient, and river channel curvature, but also takes into account physical properties such as microplastic particle density, shape factor, and material classification.

[0171] Next, these fused terrain vectors are used to construct the final fused terrain features. This process usually involves further processing of the fused vectors, such as enhancing the model's expressive power through nonlinear transformations (such as the ReLU activation function) or using normalization techniques to ensure scale consistency between different features. In addition, additional physical constraints or prior knowledge (such as the law of conservation of mass) can be introduced to improve the reliability and interpretability of model predictions. Fusing terrain features can more accurately simulate the migration behavior of microplastics under complex hydrodynamic conditions. For example, for areas with special terrain features (such as steep slopes or highly winding rivers), changes in water flow speed and direction can be better captured, thereby providing more accurate predictions of microplastic concentration distribution.

[0172] This example encodes watershed topographic features and uses GCN processing to generate enhanced spatial correlation feature vectors. Combined with the microplastic physical property encoding, a multi-head cross-attention mechanism is used to generate fused terrain vectors. Finally, fused terrain features are extracted from these fused terrain vectors, improving the accuracy of simulations of microplastic migration behavior under complex hydrological conditions and providing strong support for precise pollution prevention and control.

[0173] Based on the first embodiment of the present application, in the third embodiment of the present application, the same or similar contents as those in the first embodiment can be referred to the above introduction and will not be described in detail later. Figure 3 The microplastic transport path prediction based on a complex hydrodynamic environment further includes steps S301 to S306 before step S70:

[0174] Step S301: Acquire target watershed data.

[0175] It should be noted that the target basin data mentioned above includes microplastic concentration and terrain attribute data for the target basin. Acquisition of microplastic concentration data typically requires a combination of field sampling and laboratory analysis. Within the selected target basin, several monitoring points are set up based on factors such as water flow characteristics, population density, and industrial distribution, and water or sediment samples are collected regularly. After processing, these samples are then measured for microplastic concentration using advanced analytical techniques (such as micro-infrared spectroscopy and pyrolysis gas chromatography-mass spectrometry). Furthermore, when monitoring data is sparse, cross-basin transfer learning strategies can be combined with the Kriging spatial interpolation algorithm to generate pseudo-labels based on existing data to supplement microplastic concentration information in unlabeled areas.

[0176] Secondly, terrain attribute data includes key geographic information such as elevation, slope, roughness coefficient, and river channel sinuosity. This data is primarily derived from digital elevation models (DEMs). To ensure accuracy, it is recommended to use DEM data with a resolution of no more than 10 meters and extract relevant terrain parameters using GIS software. River channel vector maps are also necessary to identify river bifurcations and the start and end points of watersheds.

[0177] Step S302 : Select a preset proportion of target watershed data and use the Kirchnli spatial interpolation algorithm to generate pseudo labels for unlabeled watersheds.

[0178] It should be noted that Kriging interpolation is a geostatistical method that not only considers the distance between sample points, but also uses semivariogram analysis to quantify spatial autocorrelation, thereby providing the best estimate and error variance for each predicted location.

[0179] Specifically, 10% of representative monitoring points in the target watershed are selected as the initial data source. The selection of these monitoring points should be based on the principle of uniform distribution of watershed characteristics to ensure coverage of different terrain and hydrological conditions to improve the accuracy of the interpolation results. First, the semi-variogram model between the microplastic concentration of known monitoring points and geographical attributes (such as slope, elevation, etc.) is calculated. Then, based on this model, the unknown area is interpolated to generate pseudo labels. In this way, the data gaps can be effectively filled and the microplastic concentration of the entire watershed can be obtained. The specific formula is:

[0180]

[0181] in, represents the Kriging spatial interpolation algorithm, represents the concentration of microplastics in the target watershed, Represents the terrain attribute data of the target watershed.

[0182] Step S303: Evaluate the pseudo labels of the unlabeled watersheds to obtain corresponding confidence levels.

[0183] It should be noted that the Monte Carlo simulation method is used. By randomly sampling the data of known monitoring points multiple times and repeating the interpolation process, we can observe the distribution of prediction results under different sampling conditions and then calculate the confidence interval and probability distribution of the pseudo-label for each unlabeled watershed.

[0184] Step S304: When the confidence level exceeds a preset confidence level, unlabeled watershed data corresponding to the confidence level is obtained.

[0185] It should be noted that after generating pseudo-labels, a confidence threshold needs to be set. In this example, it is set to 0.85 based on existing research. Only areas with predicted probabilities exceeding this threshold are retained to ensure the reliability of the interpolation results. For each predicted area, if the predicted value of its microplastic concentration falls within the set high confidence interval, the data in that area is marked as a reliable pseudo-label.

[0186] Furthermore, these high-confidence unlabeled watershed data include not only estimated microplastic concentrations but also associated confidence scores, geographic attributes (such as slope and elevation), and other factors that may influence microplastic migration. This detailed information allows for more accurate simulation of the dynamic trends of microplastics within the target watershed. Next, incorporating this data into the domain adaptation fine-tuning phase will help adjust model parameters to better suit the specific conditions of the target watershed. A stepwise refinement approach is employed to integrate these high-confidence pseudo-labels.

[0187] Step S305: An optimization strategy is applied to the target watershed data and the unlabeled watershed data to obtain a loss value.

[0188] It should be noted that the domain difference alignment loss is calculated by methods such as maximum mean difference (MMD), aiming to narrow the feature distribution gap between the target domain (i.e., the data-rich training set) and the unlabeled domain. The loss value is obtained, and then the parameters are adjusted to reduce the loss value to enhance the cross-domain generalization ability of the model. The specific formula is:

[0189]

[0190] in, represents the maximum mean difference loss value, represents the mean square error loss value, represents the trade-off coefficient, It is set to 0.1 at the beginning of training and linearly increases to 0.9 in each round to achieve a smooth transition from MSE to MMD. Indicates unlabeled watershed data.

[0191] Step S306: Update the loss value until the trade-off coefficient reaches a preset value, and execute the step of determining the concentration change direction and concentration change intensity according to the target microplastic concentration.

[0192] It should be noted that a smaller trade-off coefficient (e.g. = 0.1), and gradually increase the coefficient as the number of training rounds increases until it reaches the preset maximum value (such as =1.0). This allows the model to initially rely more on data-driven methods for rapid convergence, while gradually increasing its focus on physical constraints in later stages, ensuring that the final results are consistent with both observed data and the laws of nature. Once the trade-off coefficient reaches the preset value, the step of determining the direction and intensity of concentration change based on the target microplastic concentration is executed. This step uses the optimized model to predict the microplastic concentration distribution at each node in the watershed at future times and calculates the direction (i.e., flow direction) and intensity (i.e., flow velocity and diffusion rate) of concentration change.

[0193] This embodiment obtains target watershed data, generates pseudo labels using kriging spatial interpolation, and screens high-quality pseudo labels using confidence levels. Finally, an optimization strategy is used to calculate the loss value and dynamically adjust the trade-off coefficient until the model is stable. This improves data coverage and prediction accuracy in areas with few samples, and also ensures the reliability of the results by combining physical constraints.

[0194] Based on the first embodiment of the present application, the present application also provides a device for predicting the transport path of microplastics based on a complex hydrodynamic environment. Figure 4 , the device comprises:

[0195] The acquisition module 10 is used to obtain watershed digital elevation model data, river vector map, microplastic properties and hydrological data.

[0196] The division module 20 is used to divide the watershed digital elevation model data and the river vector map to obtain a node set, where the nodes include elevation, slope, river bifurcation point and watershed outlet.

[0197] The calculation module 30 is used to perform calculations based on the hydrological data and the watershed digital elevation model data to generate a dynamic adjacency matrix.

[0198] The fusion module 40 is used to fuse the microplastic properties with the basin terrain characteristics to obtain fused terrain characteristics. The basin terrain characteristics are obtained based on the basin digital elevation model data and river channel vector map. The basin terrain characteristics include slope data, roughness coefficient and river channel curvature.

[0199] The calculation module 30 is also used to perform calculations based on the dynamic adjacency matrix and fused terrain features to obtain the target microplastic concentration.

[0200] The calculation module 30 is also used to determine the direction and intensity of concentration change based on the target microplastic concentration.

[0201] The calculation module 30 is further used to determine the transport node and the transport distance based on the concentration change direction and the concentration change intensity.

[0202] The result module 50 is used to combine the transport paths when the transport node is a watershed outlet in the node set to obtain a target transport path prediction result.

[0203] The device for predicting the transport path of microplastics based on a complex hydrodynamic environment provided by this application adopts the method for predicting the transport path of microplastics based on a complex hydrodynamic environment in the above-mentioned embodiment, and can solve the technical problem of how to improve the accuracy of microplastic path prediction in a complex hydrodynamic environment. Compared with the existing technology, the beneficial effects of the device for predicting the transport path of microplastics based on a complex hydrodynamic environment provided by this application are the same as the beneficial effects of the method for predicting the transport path of microplastics based on a complex hydrodynamic environment provided by the above-mentioned embodiment, and the other technical features of the device for predicting the transport path of microplastics based on a complex hydrodynamic environment are the same as the features disclosed in the above-mentioned embodiment method, and are not further described here.

[0204] In one embodiment, the calculation module 30 is further used to calculate using hydrological data to obtain flow direction similarity; calculate based on the flow direction similarity to obtain flow direction consistency between two adjacent nodes; obtain terrain resistance based on the watershed digital elevation model data; and obtain a dynamic adjacency matrix based on the flow direction consistency and terrain resistance.

[0205] In one embodiment, the fusion module 40 is also used to encode the basin terrain features to obtain an initial feature vector; obtain a first feature vector and a second feature vector based on the initial feature vector through double-layer GCN processing; encode the microplastic properties to obtain attribute coding features; obtain a fused terrain vector based on the attribute coding features and the second feature vector; and obtain a fused terrain feature based on the fused terrain vector.

[0206] In one embodiment, the calculation module 30 is also used to update the calculation based on the dynamic adjacency matrix and the fused terrain features through the graph propagation algorithm to obtain the microplastic concentration; calculate according to the watershed digital elevation model data and the river vector map to obtain the water volume corresponding to the node; obtain the flow rate from the node to the corresponding adjacent node according to the hydrological data; initialize the updated weight; obtain the concentration loss value through the mass conservation equation based on the water volume and flow rate combined with the microplastic concentration; update the concentration loss value until the updated weight reaches the preset value to obtain the target loss value; calculate in combination with the target loss value and the microplastic concentration to obtain the target microplastic concentration.

[0207] In one embodiment, the calculation module 30 is further used to determine a predicted concentration distribution based on the target microplastic concentration; and calculate the concentration gradient of the node based on the predicted concentration distribution to determine the direction and intensity of concentration change.

[0208] In one embodiment, the calculation module 30 is also used to obtain target watershed data, which includes the microplastic concentration and terrain attribute data of the target watershed; select a preset proportion of target watershed data and use the Kirchnli spatial interpolation algorithm to generate pseudo labels for unlabeled watersheds; evaluate the pseudo labels of the unlabeled watersheds to obtain corresponding confidence levels; when the confidence level exceeds the preset confidence level, obtain the unlabeled watershed data corresponding to the confidence level; use an optimization strategy for the target watershed data and the unlabeled watershed data to obtain a loss value; update the loss value until the trade-off coefficient reaches a preset value, and execute the steps of determining the direction of concentration change and the intensity of concentration change according to the target microplastic concentration.

[0209] In one embodiment, the calculation module 30 is also used to determine a set of downstream adjacent nodes that are consistent with the direction of concentration change based on the direction of concentration change; select a node corresponding to a preset concentration change intensity in the concentration change intensity from the set of downstream adjacent nodes as a transfer node; and obtain a transfer distance based on the digital elevation model data of the watershed and the river vector diagram in combination with the current node and the transfer node.

[0210] The present application provides a device for predicting the transport path of microplastics based on a complex hydrodynamic environment. The device for predicting the transport path of microplastics based on a complex hydrodynamic environment includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for predicting the transport path of microplastics based on a complex hydrodynamic environment in the above-mentioned embodiment one.

[0211] Reference below Figure 5, which shows a schematic structural diagram of a device for predicting microplastic transport paths based on complex hydrodynamic environments suitable for implementing embodiments of the present application. The device for predicting microplastic transport paths based on complex hydrodynamic environments in embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 5 The microplastic transport path prediction device based on a complex hydrodynamic environment shown is merely an example and should not impose any limitations on the functions and scope of use of the embodiments of the present application.

[0212] like Figure 5 As shown, the device for predicting the transport path of microplastics based on a complex hydrodynamic environment may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 1002 or programs loaded from a storage device 1003 into a random access memory (RAM) 1004. RAM 1004 also stores various programs and data required for the operation of the device for predicting the transport path of microplastics based on a complex hydrodynamic environment. The processing device 1001, ROM 1002, and RAM 1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following can be connected to the I / O interface 1006: input devices 1007, such as a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008, such as a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003, such as a magnetic tape, hard disk, etc.; and communication devices 1009. Communication devices 1009 can allow the device for predicting microplastic transport pathways based on complex hydrodynamic environments to communicate wirelessly or wired with other devices to exchange data. While the figure shows various devices for predicting microplastic transport pathways based on complex hydrodynamic environments, it should be understood that not all of the devices shown are required to be implemented or present. More or fewer devices may be implemented or present instead.

[0213] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0214] The device for predicting the transport path of microplastics based on complex hydrodynamic environments provided by this application uses the method for predicting the transport path of microplastics based on complex hydrodynamic environments in the above-mentioned embodiment, which can solve the technical problem of how to improve the accuracy of microplastic path prediction in complex hydrodynamic environments. Compared with the existing technology, the beneficial effects of the device for predicting the transport path of microplastics based on complex hydrodynamic environments provided by this application are the same as the beneficial effects of the method for predicting the transport path of microplastics based on complex hydrodynamic environments provided by the above-mentioned embodiment, and the other technical features of the device for predicting the transport path of microplastics based on complex hydrodynamic environments are the same as those disclosed in the method of the above-mentioned embodiment, and are not further described here.

[0215] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0216] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0217] The present application provides a computer-readable medium having computer-readable program instructions (i.e., a computer program) stored thereon, and the computer-readable program instructions are used to execute the microplastic transport path prediction method based on a complex hydrodynamic environment in the above-mentioned embodiment.

[0218] The computer-readable medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor devices, or any combination thereof. More specific examples of computer-readable media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, 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 device, magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable medium may be any tangible medium that contains or stores a program that can be executed by instructions, used by a device, or used in conjunction with it. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0219] The above-mentioned computer-readable medium can be included in the microplastic transport path prediction device based on a complex hydrodynamic environment; or it can exist independently without being assembled into the microplastic transport path prediction device based on a complex hydrodynamic environment.

[0220] The computer-readable medium carries one or more programs. When executed by the device for predicting the transport path of microplastics in a complex hydrodynamic environment, the device can write computer program code for performing the operations of the present application in one or more programming languages, or a combination thereof. These programming languages include object-oriented programming languages such as Java, Smalltalk, and C++, as well as conventional procedural programming languages such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0221] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based implementation that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0222] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0223] The computer-readable medium provided in this application stores computer-readable program instructions (i.e., a computer program) for executing the aforementioned method for predicting microplastic transport pathways in complex hydrodynamic environments. This method addresses the technical problem of improving the accuracy of microplastic pathway prediction in complex hydrodynamic environments. Compared to the prior art, the beneficial effects of the computer-readable medium provided in this application are similar to those of the method for predicting microplastic transport pathways in complex hydrodynamic environments provided in the aforementioned embodiments, and are not further elaborated here.

[0224] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for predicting microplastic transport paths based on a complex hydrodynamic environment.

[0225] The computer program product provided in this application can solve the technical problem of improving the accuracy of microplastic transport path prediction in complex hydrodynamic environments. Compared with the existing technology, the beneficial effects of the computer program product provided in this application are the same as those of the method for predicting microplastic transport paths in complex hydrodynamic environments provided in the above-mentioned embodiment, and will not be elaborated here.

[0226] The above description is only part of the embodiments of the present application and does not limit the patent scope of the present application. All equivalent structural transformations made by using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for predicting microplastic transport paths based on complex hydrodynamic environments, characterized in that: The method comprises: Obtain watershed digital elevation model data, river channel vector maps, microplastic properties, and hydrological data; Dividing the watershed digital elevation model data and the river channel vector map to obtain a node set, wherein the node includes elevation, slope, river channel bifurcation point and watershed outlet; Performing calculations based on the hydrological data and the watershed digital elevation model data to generate a dynamic adjacency matrix; The microplastic properties are integrated with the watershed topographic features to obtain integrated topographic features, wherein the watershed topographic features are obtained based on the watershed digital elevation model data and the river channel vector map, and the watershed topographic features include slope data, roughness coefficient, and river channel curvature; Calculating based on the dynamic adjacency matrix and the fused terrain features to obtain a target microplastic concentration; Determining a concentration change direction and a concentration change intensity based on the target microplastic concentration; Determining a transport node and a transport distance based on the concentration change direction and the concentration change intensity; When the transfer node is a watershed outlet in the node set, the transfer paths are combined to obtain a target transfer path prediction result.

2. The method according to claim 1, wherein The step of performing calculation based on the hydrological data and the watershed digital elevation model data to generate a dynamic adjacency matrix includes: Calculating by using the hydrological data to obtain flow direction similarity; The flow direction consistency between two adjacent nodes is calculated based on the flow direction similarity. The specific formula is: in, and Node and nodes The flow rate, is the flow direction similarity, For time; The terrain resistance is obtained according to the digital elevation model data of the watershed. The specific formula is: in, For nodes and The elevation difference between For nodes and The horizontal space distance between is the resistance factor; According to the flow direction consistency and the terrain resistance, a dynamic adjacency matrix is obtained, and the specific formula is: in is the flow direction consistency, is the terrain resistance, Represents an indicative function that only allows adjacent nodes in the downstream direction to participate in the calculation.

3. The method according to claim 1, wherein The step of fusing the microplastic properties with the basin topographic features to obtain fused topographic features, wherein the basin topographic features are obtained based on the basin digital elevation model data and the river channel vector map, and the basin topographic features include slope data, roughness coefficient, and river channel curvature, comprises: The basin terrain features are encoded to obtain the initial feature vector, which is specifically expressed as: in, is the slope data, is the roughness coefficient, is the river channel curvature; The initial feature vector is processed by a double-layer GCN to obtain the first feature vector and the second feature vector. The specific formula is: in, and Represent the first layer weight and the second layer weight of GCN respectively, For nodes The first-order neighbor set of represents a dynamic adjacency matrix; The microplastic properties are encoded to obtain the property coding characteristics. The specific formula is: represents the density of microplastics, is the shape factor, Represents a function that converts a material category into a numerical value; Based on the attribute coding feature and the second feature vector, a fused terrain vector is obtained. The specific formula is: in, are the mapping matrices of terrain features, bond vectors and microplastic properties, is the key vector dimension, is the attention weight, represents the normalization function, is the fusion feature vector; A fused terrain feature is obtained based on the fused terrain vector.

4. The method according to claim 1, wherein The step of calculating based on the dynamic adjacency matrix and the fused terrain features to obtain a target microplastic concentration includes: Based on the dynamic adjacency matrix and the fused terrain features, the microplastic concentration is updated and calculated through the graph propagation algorithm. The specific formula is: in, For nodes In time Prediction of microplastic concentrations, is the learnable weight matrix, represents the dynamic adjacency matrix, For nodes The set of first-order neighbors of ; Calculating based on the watershed digital elevation model data and the river vector diagram to obtain the water volume corresponding to the node; Obtaining a flow rate from a node to a corresponding adjacent node according to the hydrological data; Initialize update weights; The concentration loss value is obtained by combining the water volume and the flow rate with the microplastic concentration through the mass conservation equation, and the formula is: in, For nodes The corresponding water volume, For nodes arrive The flow rate, Representation node arrive Flow rate; Updating the concentration loss value until the update weight reaches a preset value to obtain a target loss value; The target loss value and the microplastic concentration are combined for calculation to obtain the target microplastic concentration.

5. The method according to claim 1, wherein The step of determining the concentration change direction and concentration change intensity according to the target microplastic concentration includes: determining a predicted concentration distribution based on the target microplastic concentration; The concentration gradient of the node is calculated based on the predicted concentration distribution to determine the concentration change direction and concentration change intensity. The formula is: in, represents the predicted concentration distribution, Representation node The position vector of and are the horizontal and vertical directions of the spatial coordinate axes.

6. The method according to claim 1, wherein Before the step of determining the transport node and the transport distance based on the concentration change direction and the concentration change intensity, the method includes: Obtaining target watershed data, wherein the target watershed data includes microplastic concentration and terrain attribute data of the target watershed; The target watershed data of a preset proportion is selected and the Kroginli spatial interpolation algorithm is used to generate pseudo labels for unlabeled watersheds. The specific formula is: in, represents the Kriging spatial interpolation algorithm, represents the concentration of microplastics in the target watershed, Represents the terrain attribute data of the target watershed; Evaluating the pseudo labels of the unlabeled watershed to obtain corresponding confidence scores; When the confidence level exceeds a preset confidence level, obtaining unlabeled watershed data corresponding to the confidence level; An optimization strategy is used for the target watershed data and the unlabeled watershed data to obtain a loss value. The specific formula is: in, represents the maximum mean difference loss value, represents the mean square error loss value, represents the trade-off coefficient, Indicates unlabeled watershed data; The loss value is updated until the trade-off coefficient reaches a preset value, and the step of determining the concentration change direction and concentration change intensity based on the target microplastic concentration is performed.

7. The method according to claim 1, wherein The step of determining the transport node and the transport distance based on the concentration change direction and the concentration change intensity includes: Determine, according to the concentration change direction, a set of downstream adjacent nodes that is consistent with the change direction; Selecting a node corresponding to a preset concentration change intensity in the concentration change intensity from the downstream adjacent node set as a transfer node; The transport distance is obtained based on the watershed digital elevation model data and the river vector diagram in combination with the current node and the transport node.

8. A device for predicting microplastic transport paths based on complex hydrodynamic environments, characterized in that: The device comprises: The acquisition module is used to obtain watershed digital elevation model data, river vector diagrams, microplastic properties and hydrological data; A partitioning module is used to partition the watershed digital elevation model data and the river channel vector map to obtain a node set, wherein the node includes elevation, slope, river channel bifurcation point and watershed outlet; A calculation module, configured to perform calculations based on the hydrological data and the watershed digital elevation model data to generate a dynamic adjacency matrix; A fusion module is used to fuse the microplastic properties with the watershed topographic characteristics to obtain fused topographic characteristics, wherein the watershed topographic characteristics are obtained based on the watershed digital elevation model data and the river channel vector map, and the watershed topographic characteristics include slope data, roughness coefficient, and river channel tortuosity; A calculation module is further used to calculate based on the dynamic adjacency matrix and the fused terrain features to obtain a target microplastic concentration; The calculation module is further used to determine the concentration change direction and concentration change intensity according to the target microplastic concentration; The calculation module is further used to determine the transport node and the transport distance based on the concentration change direction and the concentration change intensity; The result module is used to combine the transport paths when the transport node is a watershed outlet in the node set to obtain a target transport path prediction result.

9. A device for predicting microplastic transport paths based on complex hydrodynamic environments, characterized in that: The device includes: a memory, a processor, and a microplastic transport path prediction program based on a complex hydrodynamic environment stored in the memory and running on the processor. The microplastic transport path prediction program based on a complex hydrodynamic environment is configured to implement the steps of the microplastic transport path prediction method based on a complex hydrodynamic environment as described in any one of claims 1-7.

10. A medium, characterized in that The medium stores a microplastic transport path prediction program based on a complex hydrodynamic environment. When the microplastic transport path prediction program based on a complex hydrodynamic environment is executed by the processor, the steps of the microplastic transport path prediction method based on a complex hydrodynamic environment as described in any one of claims 1 to 7 are implemented.

Images