Determination method, application method and device of prediction model of micro-plastic components in water body

Abstract

The invention discloses a determination method, an application method and a device of a prediction model of micro-plastic components in a water body, and relates to the field of water environment monitoring and pollutant modeling, and the determination method of the model comprises the following steps: obtaining water body sample data collected in a plurality of target areas; preprocessing the water body sample data to obtain preprocessed data; inputting the preprocessed hydrological environment factors into a multi-layer perceptron model to obtain the output of the multi-layer perceptron model; based on the output of the multi-layer perceptron model, the logarithmic concentration of various microplastics corresponding to the preprocessed hydrological environment factors and the determined loss function, combining an EarlyStopping strategy and a ReduceLROnPlateau mechanism, and dynamically adjusting the learning rate; and optimizing network parameters of the multi-layer perceptron model based on the learning rate to obtain a prediction model of the micro-plastic components in the water body. According to the invention, the accuracy, stability and generalizability of micro-plastic concentration prediction can be effectively improved.

Description

Technical Field

[0001] This application relates to the field of water environment monitoring and pollutant modeling technology, and in particular to a method, application method and device for determining a predictive model of microplastic composition in water. Background Technology

[0002] Microplastics, as an emerging pollutant, are widely present in water bodies, posing a potential threat to the ecological environment and human health. The concentration of microplastics in water bodies in different regions is influenced by various factors, such as hydrological characteristics, topographic features, climate, land cover, soil properties, and human activities. Existing studies typically collect microplastic samples through experimental monitoring and laboratory analysis, but these methods cannot provide rapid and accurate concentration predictions for large-scale or dynamically changing areas.

[0003] Traditional methods for predicting microplastic concentrations often employ single-objective regression or single machine learning models, which struggle to simultaneously handle the correlations and differences among various microplastic components (such as PP and PE). Furthermore, the lack of consideration for physical properties like particle size and density during data preprocessing can lead to significant prediction biases, failing to meet the demands of refined monitoring in water environment management. Therefore, there is an urgent need for a modeling method capable of efficiently and accurately predicting the concentrations of multiple types of microplastics to support the assessment and remediation of microplastic pollution in water bodies. Summary of the Invention

[0004] The purpose of this application is to provide a method, application method and device for determining a predictive model of microplastic composition in water bodies, which can effectively improve the accuracy, stability and scalability of microplastic concentration prediction and is applicable to pollution monitoring tasks in large-scale water bodies.

[0005] To achieve the above objectives, this application provides the following solution:

[0006] In a first aspect, this application provides a method for determining a prediction model for microplastic composition in water bodies, the method comprising:

[0007] Acquire water sample data collected from multiple target areas; the water sample data includes: microplastic abundance information and hydrological environmental factors; the microplastic abundance information includes: polypropylene (PP), polyethylene (PE), polystyrene (PS), polyethylene terephthalate (PET), and polyvinyl chloride (PVC), as well as the particle size distribution information corresponding to each component; the hydrological environmental factors include: hydrological characteristic data, topographic feature data, climate data, land cover data, soil characteristic data, and human activity data.

[0008] The water sample data is preprocessed to obtain preprocessed data; the preprocessing includes: outlier removal, particle size-based normalization, concentration conversion, feature screening of hydrological environmental factors, and standardization.

[0009] The preprocessed hydrological environmental factors are input into the multilayer perceptron model to obtain the output of the multilayer perceptron model; the multilayer perceptron model is used to realize multi-task regression prediction of various microplastic concentrations.

[0010] Based on the output of the multilayer perceptron model, the logarithmic concentrations of various microplastics corresponding to the preprocessed hydrological environmental factors, and the determined loss function, the learning rate is dynamically adjusted by combining the EarlyStopping strategy and the ReduceLROnPlateau mechanism.

[0011] Based on the learning rate, the network parameters of the multilayer perceptron model are optimized to obtain a prediction model for microplastic composition in water.

[0012] Secondly, this application provides a method for applying a prediction model of microplastic composition in water bodies, the method comprising:

[0013] The study aims to acquire hydrological environmental factors and the concentrations of various microplastics that have no predictive effect on the aquatic ecosystem in the area to be tested. The hydrological environmental factors include hydrological characteristic data, topographic feature data, climate data, land cover data, soil characteristic data, and human activity data.

[0014] The hydrological environmental factors of the area to be measured are normalized to obtain normalized data.

[0015] The normalized data is input into the water body microplastic composition prediction model to obtain the predicted logarithmic concentration of various microplastics; the water body microplastic composition prediction model is a model obtained based on the above-described method for determining the water body microplastic composition prediction model.

[0016] The predicted logarithmic concentrations of the various microplastics are subjected to inverse concentration transformation to obtain the predicted actual concentrations of the various microplastics.

[0017] Based on the predicted actual concentrations of the various microplastics and the predicted ineffective concentrations of the various microplastics on the aquatic ecosystem, the risk quotients of the various microplastics are calculated.

[0018] Based on the risk quotients of the various microplastics and the set pollution risk level thresholds, the pollution risk level of the area to be tested is determined; the pollution risk level includes: safe zone, key monitoring zone, and warning and control zone.

[0019] Based on the pollution risk level of the area to be tested, an interpolation method is used to generate a spatial risk distribution map, and control recommendations are given in conjunction with pollution source information.

[0020] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to implement the above-described method for determining the microplastic composition prediction model in water or the method for applying the microplastic composition prediction model in water.

[0021] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for determining the prediction model of microplastic composition in water or the method for applying the prediction model of microplastic composition in water as described above.

[0022] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method for determining the microplastic composition prediction model in water or the method for applying the microplastic composition prediction model in water as described above.

[0023] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0024] This application provides a method, application method, and apparatus for determining a prediction model of microplastic composition in water bodies. The method for determining the prediction model of microplastic composition in water bodies includes: acquiring water sample data collected from multiple target areas; the water sample data includes: microplastic abundance information and hydrological environmental factors; the microplastic abundance information includes: polypropylene, polyethylene, polystyrene, polyethylene terephthalate, and polyvinyl chloride, as well as the particle size distribution information corresponding to each component; the hydrological environmental factors include: hydrological characteristic data, topographic feature data, climate data, land cover data, soil characteristic data, and human activity data. The data from "multiple target areas" covers scenarios with different geographical, climatic, and human activity intensities, avoiding the limitations of single-area data, improving sample diversity, and providing support for the generalization ability of the subsequent model. The water sample data is preprocessed to obtain preprocessed data; the preprocessing includes: outlier removal, particle size-based normalization, concentration conversion, feature screening of hydrological environmental factors, and standardization, which can improve data quality and provide standardized and highly relevant features for model input. Preprocessed hydrological environmental factors are input into a multilayer perceptron (MLP) model to obtain its output. This MLP model is used for multi-task regression prediction of various microplastic concentrations. Multi-task regression, by sharing a bottom-level feature extraction layer, leverages the correlation between components to improve prediction accuracy for each task, while reducing redundant computations from individual modeling and improving efficiency. Based on the MLP model's output, the logarithmic concentrations of various microplastics corresponding to the preprocessed hydrological environmental factors, and a determined loss function, the learning rate is dynamically adjusted using the EarlyStopping strategy and the ReduceLROnPlateau mechanism. This effectively prevents overfitting and improves training efficiency and accuracy. Based on the learning rate, the network parameters of the MLP model are optimized to obtain a microplastic composition prediction model for water bodies. This model can directly predict the concentrations of various microplastic components, providing clear targets and basis for pollution control, and possessing both scientific research and application value. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is an application environment diagram of a method for determining a prediction model of microplastic composition in water, as described in one embodiment of this application.

[0027] Figure 2This is a flowchart illustrating a method for determining a predictive model of microplastic composition in water, provided as an embodiment of this application.

[0028] Figure 3 This is a flowchart illustrating the application method of a microplastic composition prediction model in water, provided in one embodiment of this application.

[0029] Figure 4 This is a flowchart illustrating a multi-task neural network modeling method for predicting the concentration of microplastic components in water, provided as an embodiment of this application.

[0030] Figure 5 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] This application employs a scientific data processing strategy and a multi-task modeling structure, which can effectively improve the accuracy, stability, and generalizability of microplastic concentration prediction, and is applicable to pollution monitoring tasks in large-scale water bodies.

[0033] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] The method for determining the prediction model of microplastic composition in water provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on other servers. Terminal 102 can send water sample data collected from multiple target areas to server 104. The water sample data includes: microplastic abundance information and hydrological environmental factors; the microplastic abundance information includes: polypropylene, polyethylene, polystyrene, polyethylene terephthalate, and polyvinyl chloride, as well as the particle size distribution information corresponding to each component; the hydrological environmental factors include: hydrological characteristic data, topographic feature data, climate data, land cover data, soil characteristic data, and human activity data; after receiving the water sample data collected from multiple target areas, server 104 preprocesses the water sample data to obtain preprocessed data. The preprocessing includes: outlier removal, particle size-based normalization, concentration conversion, feature screening of hydrological environmental factors, and standardization. The preprocessed hydrological environmental factors are input into a multilayer perceptron model to obtain its output. This multilayer perceptron model is used for multi-task regression prediction of various microplastic concentrations. Based on the output of the multilayer perceptron model, the logarithmic concentrations of various microplastics corresponding to the preprocessed hydrological environmental factors, and a determined loss function, the learning rate is dynamically adjusted using an EarlyStopping strategy and a ReduceLROnPlateau mechanism. Based on the learning rate, the network parameters of the multilayer perceptron model are optimized to obtain a microplastic composition prediction model for water bodies. Server 104 can feed back the obtained microplastic composition prediction model for water bodies to terminal 102. In addition, in some embodiments, the method for determining the prediction model of microplastic composition in water can also be implemented by the server 104 or the terminal 102 separately. For example, the terminal 102 can directly determine the prediction model of microplastic composition in water based on water sample data collected in multiple target areas, or the server 104 can obtain water sample data collected in multiple target areas from the data storage system and determine the prediction model of microplastic composition in water based on the water sample data collected in multiple target areas.

[0035] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, and tablets. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers, or it can be a cloud server.

[0036] In one exemplary embodiment, such as Figure 2As shown, a method for determining a predictive model for microplastic composition in water is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the following steps are included:

[0037] S1: Acquire water sample data collected from multiple target areas; the water sample data includes: microplastic abundance information and hydrological environmental factors; the microplastic abundance information includes: polypropylene, polyethylene, polystyrene, polyethylene terephthalate and polyvinyl chloride, as well as the particle size distribution information corresponding to each component; the hydrological environmental factors include: hydrological characteristic data, topographic feature data, climate data, land cover data, soil characteristic data and human activity data.

[0038] S2: Preprocess the water sample data to obtain preprocessed data; the preprocessing includes: outlier removal, particle size-based normalization, concentration conversion, feature screening of hydrological environmental factors, and standardization.

[0039] S3: Input the preprocessed hydrological environmental factors into the multilayer perceptron model to obtain the output of the multilayer perceptron model; the multilayer perceptron model is used to realize multi-task regression prediction of various microplastic concentrations.

[0040] S4: Based on the output of the multilayer perceptron model, the logarithmic concentrations of various microplastics corresponding to the preprocessed hydrological environmental factors, and the determined loss function, the learning rate is dynamically adjusted by combining the EarlyStopping strategy and the ReduceLROnPlateau mechanism.

[0041] S5: Based on the learning rate, optimize the network parameters of the multilayer perceptron model to obtain a prediction model for microplastic composition in water.

[0042] By implementing steps S1 to S5 above, this application employs a scientific data processing strategy and a multi-task modeling structure, which effectively improves the accuracy, stability, and generalizability of microplastic concentration prediction, making it suitable for large-scale water pollution monitoring tasks.

[0043] As an optional implementation, in step S1, the particle size distribution information is used for subsequent concentration conversion and normalization processing.

[0044] The hydrological characteristic data include: flow rate, surface runoff, lake area, lake volume, river area, river volume, and groundwater depth.

[0045] The terrain feature data includes: elevation, terrain slope, and river slope.

[0046] The climate data includes: temperature, precipitation, potential evaporation, actual evaporation, and humidity.

[0047] The land cover data includes: land cover type, land cover area, potential plant type, potential plant area, forest cover area, farmland cover area, pasture cover area, irrigated cover area, and glacier cover area.

[0048] The soil characteristic data include: clay content, silt content, gravel content, organic carbon content, rock type, and soil erosion area.

[0049] The human activity data includes: population size, population density, urbanized area, road density, human footprint, and human development index.

[0050] As an optional implementation, in step S2, the water sample data is preprocessed to obtain preprocessed data, specifically including:

[0051] S21: Remove outliers from the water sample data to obtain the data after removal.

[0052] S22: Based on the data after elimination, the particle size distribution information corresponding to each component is normalized based on the particle size distribution to obtain the normalized data.

[0053] S23: Based on the normalized data and the density of various microplastics, the concentration is converted to obtain the logarithmic concentration of various microplastics.

[0054] S24: The hydrological environmental factors in the removed data are subjected to feature screening and standardization processing through statistical correlation and multicollinearity analysis to obtain the feature-screened factors.

[0055] Specifically, the preprocessing includes: removing outliers from the microplastic data, including removing extreme values ​​within the top and bottom 5%; normalizing the microplastic abundance based on particle size distribution; converting the microplastic abundance into concentration values ​​for various microplastics based on particle size and density; and performing feature filtering on the environmental factor data, including removing highly collinear and highly correlated variables based on variance inflation factor (VIF) and Spearman rank correlation coefficient, while standardizing all numerical features.

[0056] As an optional implementation, in step S3, the multilayer perceptron model includes: an input layer, several nonlinear hidden layers, and a linear output layer.

[0057] The input layer is used to input the preprocessed hydrological environmental factors.

[0058] The aforementioned nonlinear hidden layers are used to extract features from the preprocessed hydrological environmental factors and, in conjunction with L2 regularization, BatchNormalization, and Dropout mechanisms, to obtain the predicted logarithmic concentrations of various microplastics.

[0059] A linear output layer is used to output the predicted logarithmic concentrations of various microplastics.

[0060] Specifically, the multilayer perceptron model consists of an input layer, multiple nonlinear hidden layers, and a linear output layer. The input layer corresponds to multiple hydrological environmental factor variables. The nonlinear hidden layers achieve complex feature extraction through nonlinear activation functions and combine L2 regularization, Batch Normalization, and Dropout mechanisms to enhance the model's stability and generalization ability. The linear output layer is used to simultaneously predict the logarithmic concentration of multiple types of microplastics, realizing multi-task regression modeling.

[0061] In this embodiment, the multilayer perceptron model includes three hidden layers, each employing the ReLU activation function, and is combined with L2 regularization, BatchNormalization, and Dropout mechanisms to effectively improve the model's nonlinear expressive power and generalization performance.

[0062] It should be noted that the model input has been standardized to ensure that the scale of different feature variables is consistent, avoiding the impact of numerical differences on the training process. At the same time, the hidden layer output is normalized to stabilize the network training and promote the model convergence speed.

[0063] As an optional implementation, before inputting the preprocessed hydrological environmental factors into the multilayer perceptron model, the method for determining the microplastic composition prediction model in the water body further includes:

[0064] The preprocessed hydrological environmental factors are then augmented to obtain augmented data.

[0065] Specifically, the model training uses mean squared error (MSE) as the loss function, combined with the EarlyStopping strategy, which automatically stops training to prevent overfitting when the validation set loss does not improve significantly within a set number of rounds.

[0066] The ReduceLROnPlateau mechanism is introduced to dynamically adjust the learning rate based on the changes in the validation set loss, which helps the model escape local optima and improves training stability. During training, data augmentation methods (Gaussian noise perturbation and linear interpolation) can enhance the robustness of the model.

[0067] As an optional implementation, the method for determining the microplastic composition prediction model in the water body further includes: evaluating the microplastic composition prediction model in the water body using mean square error and coefficient of determination.

[0068] Specifically, a ten-fold cross-validation method is used to evaluate the performance of the training set. The model weights are reinitialized at each fold of training, and the results are evaluated based on the mean squared error (MSE) and coefficient of determination (R²). 2 The model's prediction accuracy and generalization ability are comprehensively evaluated. If the evaluation metrics are not met, the model parameters are fine-tuned and the model is retrained to ensure optimized performance.

[0069] After training, the model is applied to data from unknown areas. The input data is first normalized using standardized parameters stored in the training set to predict the logarithmic value of microplastic concentration. Finally, the actual concentration is obtained through inverse transformation, enabling a quantitative assessment of pollution in the new area.

[0070] like Figure 3 As shown, an application method for a microplastic composition prediction model in water bodies is provided. The application method of the microplastic composition prediction model in water bodies includes:

[0071] A1: Obtain hydrological environmental factors and the concentrations of various microplastics that have no predictive effect on the aquatic ecosystem in the area to be tested; the hydrological environmental factors include: hydrological characteristic data, topographic feature data, climate data, land cover data, soil characteristic data, and human activity data.

[0072] A2: Normalize the hydrological environmental factors of the area to be measured to obtain normalized data.

[0073] A3: Input the normalized data into the water body microplastic composition prediction model to obtain the predicted logarithmic concentration of various microplastics; the water body microplastic composition prediction model is a model obtained based on the determination method of water body microplastic composition prediction model described above.

[0074] A4: Perform an inverse concentration transformation on the predicted logarithmic concentrations of the various microplastics to obtain the predicted actual concentrations of the various microplastics.

[0075] A5: Based on the predicted actual concentrations of the various microplastics and the predicted ineffective concentrations of the various microplastics on the aquatic ecosystem, the risk quotient (RQ) of each type of microplastic is calculated.

[0076] A6: Based on the risk quotient of the various microplastics and the set pollution risk level threshold, determine the pollution risk level of the area to be tested; the pollution risk level includes: safe zone, key monitoring zone and warning and control zone.

[0077] A7: Based on the pollution risk level of the area to be tested, an interpolation method is used to generate a spatial risk distribution map, and control recommendations are given in combination with pollution source information.

[0078] As an optional implementation, based on the concentration results of five types of microplastics predicted by the model, a risk assessment and response module is introduced, which specifically includes the following steps:

[0079] B1: The concentration values ​​of various microplastics predicted by the model are used as the predicted environmental concentration (PEC). The ineffective concentration (PNEC) of the same type of plastic for the aquatic ecosystem obtained from the literature is introduced to calculate the risk quotient of each type of microplastic.

[0080] B2: Based on the RQ values ​​of various microplastics, set pollution risk level thresholds to classify the pollution level of the predicted area.

[0081] B3: Combining microplastic concentration prediction with risk level information, the target area is gridded using the Kriging interpolation method to obtain the pollution risk level of each grid unit.

[0082] B4: The system matches a rule library of pollution source control recommendations based on various types and risk levels of microplastic pollution.

[0083] Preferably, in step B1, the formula for calculating the risk quotient is:

[0084] RQ = PEC / PNEC;

[0085] Where RQ is the risk quotient; PEC is the predicted concentration; and PNEC is the safe threshold concentration for this type of microplastic (without visible ecotoxicity effects).

[0086] Preferably, in step B2, the pollution risk level is classified as follows:

[0087] RQ<0.1: Low risk, classified as a safe zone.

[0088] 0.1≤RQ<1: Medium risk, identified as a key monitoring area.

[0089] RQ≥1: High risk, designated as a restricted area.

[0090] The risk levels of each microplastic can be superimposed to form a total pollution risk map, which can be used to identify areas with high incidence of compound pollution and support the coordinated treatment and priority assessment of multiple pollutants.

[0091] Based on the above principles, this application will be further elaborated as follows:

[0092] To improve the accuracy and regional adaptability of microplastic concentration prediction in water bodies, this application mainly combines multi-source heterogeneous environmental factor data with information on various microplastic components, and uses a multi-task neural network model for joint modeling. Through scientific data preprocessing, feature selection, and multi-task structure design, the influence mechanisms of various environmental factors on different microplastic concentrations can be effectively explored, thereby improving the accuracy and stability of concentration prediction.

[0093] Specifically, such as Figure 4 As shown, historical microplastic data, environmental factor data, and concentration data of the predicted ineffectiveness of various microplastics on the aquatic ecosystem were first collected from multiple water sampling points within the target area. The microplastic data included abundance information and particle size distribution characteristics of five common microplastics: polypropylene, polyethylene, polystyrene, polyethylene terephthalate, and polyvinyl chloride. This data was obtained through a systematic search and screening of high-impact literature from the past decade. Environmental data include hydrological characteristics (flow, surface runoff, lake area, lake volume, river area, river volume, groundwater depth), topographic features (elevation, slope, river slope), climate data (temperature, precipitation, potential evaporation, actual evaporation, humidity), land cover (land cover type, land cover area, potential vegetation type, potential vegetation area, forest cover area, farmland cover area, pasture cover area, irrigated cover area, glacier cover area), soil characteristics (clay content, silt content, gravel content, organic carbon content, rock type, soil erosion area), and human activity data (population size, population density, urbanized area, road density, human footprint, human development index). All environmental factors were uniformly extracted from the globally authoritative hydrogeographic database HydroATLAS to ensure data consistency, spatial accuracy, and scientific comparability. PNEC values ​​for various microplastics were compiled from recently cited domestic and international literature and ecotoxicological research databases to ensure the authority and validity of the risk assessment parameters.

[0094] After obtaining the microplastic abundance data and environmental factor data, the raw data needs to be preprocessed. This preprocessing includes handling outliers and missing values, and standardizing and normalizing the feature variables, as detailed below:

[0095] (1) Data anomaly handling: Extreme values ​​in microplastic abundance data are identified and removed. The upper and lower 5% quantiles are used as thresholds to remove abnormal observations that deviate significantly from the overall distribution, so as to ensure the stability and representativeness of the data.

[0096] (2) Particle size normalization and concentration conversion: Based on the particle size distribution characteristics of each type of microplastic, the abundance of each type of microplastic is normalized using a standard particle size range. Combined with material density and volume parameters, the abundance of each type of microplastic is converted into a concentration value under a unified unit, which facilitates cross-regional comparison and unified modeling.

[0097] (3) Missing value handling: For individual missing items in the environmental factor data, spatial proximity interpolation or time series interpolation methods are used to fill in the missing items to ensure the integrity of the modeling features.

[0098] (4) Feature selection and dimensionality reduction: Based on Spearman's rank correlation coefficient and variance inflation factor, correlation analysis and collinearity diagnosis of environmental factors are performed to eliminate highly correlated or redundant feature variables, thereby improving the stability and computational efficiency of the model.

[0099] (5) Feature standardization: Z-score standardization is performed on all numerical feature variables to ensure that each feature dimension has the same mean and variance, thus avoiding deviations in the neural network training process due to differences in units.

[0100] In the implementation of this application, a multilayer perceptron neural network was used as the core modeling framework to perform multi-task regression prediction of the concentrations of five common microplastic components (polypropylene PP, polyethylene PE, polystyrene PS, polyethylene terephthalate PET, and polyvinyl chloride PVC) in water. The network structure includes an input layer, three hidden layers, and a linear output layer, which can effectively capture the complex nonlinear mapping relationships between multiple variables.

[0101] The model structure is configured as follows: The input layer has 14 neurons, corresponding to 14 environmental and human factor feature variables (such as watershed distance DIS, flow velocity RUN, water volume VOL, etc.); the first hidden layer has 128 neurons, using the ReLU activation function, and applying L2 regularization (regularization coefficient λ = 0.01) to prevent overfitting; the second hidden layer has 64 neurons, using the ReLU activation function, and incorporating Batch Normalization to stabilize the training process; the third hidden layer has 32 neurons, and in addition to the ReLU activation function and Batch Normalization, Dropout (dropout rate set to 0.3) is further introduced to enhance the robustness of the model; the output layer has 5 neurons, corresponding to five types of microplastic concentrations (after logarithmic transformation), using a linear activation function to preserve regression properties.

[0102] During model training, the optimizer was Adam, and the initial learning rate was set to 0.001. To improve model generalization performance and avoid overfitting, the EarlyStopping strategy (training stops if the validation loss does not improve within 50 training epochs) and the ReduceLROnPlateau mechanism (automatically decreasing the learning rate by a factor of 0.2 when the validation loss continues to fail to decrease, with a minimum learning rate of 1e-6) were introduced. Before model training, StandardScaler was used to standardize all input features, and the mean and standard deviation parameters were saved to ensure feature consistency during deployment.

[0103] After data preprocessing, the original dataset contained 14 input variables (environmental factors) and 5 target variables (microplastic species). Considering the small initial values ​​of microplastic concentration, the target variable Y underwent the following logarithmic transformation before modeling to enhance numerical stability and regression performance:

[0104] Y' = log 10 (Y×10 6 +1);

[0105] 10 of them 6 1 is the amplification factor, and 1 is the smoothing term used to avoid logarithmic calculation errors caused by zero values.

[0106] To improve the model's robustness and learning ability on small sample distributions, this application introduces two data augmentation strategies:

[0107] Gaussian noise perturbation method: Gaussian white noise with a standard deviation of 0.005 is added to the input features and target variables to simulate sampling error.

[0108] Linear interpolation: Two sets of sample data are randomly selected and a new sample is synthesized by weighted averaging, which enhances the continuity and richness of the sample distribution.

[0109] Through enhancement techniques, the number of original samples was significantly increased, thereby significantly improving the model's performance under complex environmental inputs.

[0110] The dataset was divided into training and test sets in an 8:2 ratio. A 10-fold cross-validation method was further used on the training set to evaluate model performance: the data was divided into ten subsets, one of which was used as the validation set in turn, while the remaining nine subsets were used as the training set. In each round, the coefficient of determination and mean squared error of the model on the validation set were calculated, and the final average was used as the evaluation metric to measure the model's generalization ability and stability.

[0111] During the testing phase, the trained final model is used to predict the test set samples, and the R-squared values ​​for the five target variables are calculated. 2The prediction accuracy differences for each target microplastic were plotted using the MSE (Mean Sequence Estimation) metric as an evaluation indicator. Simultaneously, the loss curves from the entire training process were saved to analyze model convergence speed and overfitting trends.

[0112] After deployment, the model can perform feature standardization and prediction operations on water sample data collected from any region, and output real-time estimated concentrations of five types of microplastics. Combined with the pollution risk assessment module, the system automatically calculates the risk quotient based on the predicted concentrations and classifies the pollution level according to international water quality risk standards (e.g., low risk, medium risk, high risk). The spatial risk map is generated using Kriging interpolation technology to identify spatial distribution patterns of microplastic pollution. The system supports the integration of a pollution source information database to automatically propose site-specific pollution source control recommendations, ultimately assisting water environment management departments in formulating precise governance strategies.

[0113] This application employs a multi-task neural network structure to jointly model and predict the concentrations of multiple microplastic components. Combined with scientific data preprocessing strategies and feature selection mechanisms, it enhances the model's ability to learn the distribution patterns of microplastics under complex hydrological environments. By introducing mechanisms such as regularization, normalization, and Dropout, the model's stability and generalization ability are improved. Furthermore, the use of EarlyStopping and dynamic learning rate adjustment strategies during training effectively prevents overfitting. In addition, this application utilizes risk quotient calculation and spatial interpolation methods to achieve quantitative ecological risk assessment and spatial identification of the prediction results. This supports accurate identification of high-risk areas and early warning and control of key water bodies, effectively improving the scientific rigor, accuracy, and applicability of microplastic pollution monitoring, and is particularly suitable for intelligent prediction and risk management tasks related to large-scale water pollution.

[0114] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 5As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores water sample data collected from multiple target areas. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for determining a predictive model of microplastic composition in water or a method for applying a predictive model of microplastic composition in water.

[0115] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0116] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method embodiments.

[0117] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the above-described method embodiments.

[0118] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the above-described method embodiments.

[0119] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0120] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0121] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0122] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0123] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for determining a predictive model for microplastic composition in water bodies, characterized in that, The method for determining the prediction model for microplastic composition in water includes: Acquire water sample data collected from multiple target areas; the water sample data includes: microplastic abundance information and hydrological environmental factors; the microplastic abundance information includes: polypropylene, polyethylene, polystyrene, polyethylene terephthalate and polyvinyl chloride, as well as the particle size distribution information corresponding to each component; the hydrological environmental factors include: hydrological characteristic data, topographic feature data, climate data, land cover data, soil characteristic data and human activity data; The water sample data is preprocessed to obtain preprocessed data; the preprocessing includes: outlier removal, particle size-based normalization, concentration conversion, feature screening of hydrological environmental factors, and standardization. The preprocessed hydrological environmental factors are input into the multilayer perceptron model to obtain the output of the multilayer perceptron model; the multilayer perceptron model is used to realize multi-task regression prediction of various microplastic concentrations. Based on the output of the multilayer perceptron model, the logarithmic concentrations of various microplastics corresponding to the preprocessed hydrological environmental factors, and the determined loss function, the learning rate is dynamically adjusted by combining the EarlyStopping strategy and the ReduceLROnPlateau mechanism. Based on the learning rate, the network parameters of the multilayer perceptron model are optimized to obtain a prediction model for microplastic composition in water.

2. The method for determining the prediction model of microplastic composition in water bodies according to claim 1, characterized in that, The water sample data is preprocessed to obtain preprocessed data, specifically including: Outliers in the water sample data are removed to obtain the data after removal; Based on the data after the removal, the particle size distribution information corresponding to each component is normalized based on the particle size distribution to obtain the normalized data. Based on the normalized data and the density of various microplastics, the concentration is converted to obtain the logarithmic concentration of various microplastics. The hydrological environmental factors in the removed data were subjected to feature screening and standardization by statistical correlation and multicollinearity analysis to obtain the feature-screened factors.

3. The method for determining the prediction model of microplastic composition in water bodies according to claim 1, characterized in that, The multilayer perceptron model includes: an input layer, several nonlinear hidden layers, and a linear output layer; The input layer is used to input the preprocessed hydrological environmental factors; The aforementioned nonlinear hidden layers are used to extract features from the preprocessed hydrological environmental factors and combine L2 regularization, Batch Normalization and Dropout mechanisms to obtain the predicted logarithmic concentrations of various microplastics. A linear output layer is used to output the predicted logarithmic concentrations of various microplastics.

4. The method for determining the prediction model of microplastic composition in water bodies according to claim 1, characterized in that, Before inputting the preprocessed hydrological environmental factors into the multilayer perceptron model, the method for determining the microplastic composition prediction model in the water body further includes: The preprocessed hydrological environmental factors are then augmented to obtain augmented data.

5. The method for determining the prediction model of microplastic composition in water bodies according to claim 1, characterized in that, The method for determining the prediction model of microplastic composition in water bodies further includes: evaluating the prediction model of microplastic composition in water bodies using mean square error and coefficient of determination.

6. A method for applying a predictive model for microplastic composition in water bodies, characterized in that, The application methods of the microplastic composition prediction model in water bodies include: The study aims to acquire hydrological environmental factors and the predictive ineffective concentrations of various microplastics on the aquatic ecosystem in the area under test. The hydrological environmental factors include: hydrological characteristic data, topographic feature data, climate data, land cover data, soil characteristic data, and human activity data. The hydrological environmental factors of the area to be measured are normalized to obtain normalized data. The normalized data is input into the water body microplastic composition prediction model to obtain the predicted logarithmic concentration of various microplastics; the water body microplastic composition prediction model is a model obtained based on the method for determining the water body microplastic composition prediction model according to any one of claims 1-5; The predicted logarithmic concentrations of the various microplastics are inversely transformed to obtain the predicted actual concentrations of the various microplastics. Based on the predicted actual concentrations of the various microplastics and the predicted ineffective concentrations of the various microplastics on the aquatic ecosystem, the risk quotients of the various microplastics are calculated. Based on the risk quotients of the various microplastics and the set pollution risk level thresholds, the pollution risk level of the area to be tested is determined; the pollution risk level includes: safe zone, key monitoring zone, and warning and control zone; Based on the pollution risk level of the area to be tested, an interpolation method is used to generate a spatial risk distribution map, and control recommendations are given in conjunction with pollution source information.

7. The application method of the microplastic composition prediction model in water bodies according to claim 6, characterized in that, The formula for calculating the risk quotient is: RQ = PEC / PNEC; Where RQ is the risk quotient; PEC is the predicted concentration; and PNEC is the safe threshold concentration for this type of microplastic.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the method for determining the microplastic composition prediction model in water as claimed in any one of claims 1-5 or the method for applying the microplastic composition prediction model in water as claimed in any one of claims 6-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for determining the prediction model of microplastic composition in water as described in any one of claims 1-5, or the method for applying the prediction model of microplastic composition in water as described in any one of claims 6-7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method for determining the prediction model of microplastic composition in water as described in any one of claims 1-5, or the method for applying the prediction model of microplastic composition in water as described in any one of claims 6-7.

Images