Groundwater pollution path microscopic analysis method based on graph neural network

Abstract

The invention discloses an underground water pollution path micro-analysis method based on a graph neural network. The method comprises the following steps: S1, acquiring a historical data set and a real-time monitoring data set of underground water pollution; s2, using a fuzzy membership function to represent the language level of the fuzzy variable; s3, constructing a fuzzy logic rule base based on the historical data set; s4, using a fuzzy inference system to process fuzzy variables in the historical data set; s5, constructing an underground water pollution path diagram structure model based on the historical data set and the real-time monitoring data set; s6, constructing an incidence matrix between the nodes; s7, based on the trained graph neural network model, inputting water flow velocity, pollutant concentration and geologic structure parameters in the real-time monitoring data set; s8, generating real-time prediction of a pollution diffusion path; and S9, outputting an analysis result of the groundwater pollution path. The problems that in the prior art, a pollution diffusion path is difficult to accurately predict under a complex geological structure, and real-time dynamic response is insufficient are solved.

Description

Technical Field

[0001] The present invention relates to the field of groundwater pollution technology, and in particular to a groundwater pollution path micro-analysis method based on graph neural network. Background Art

[0002] In the existing technology, the prediction of groundwater pollution paths mainly relies on physical models and empirical models. Physical models and empirical models usually describe the flow characteristics of groundwater and the migration and diffusion behavior of pollutants, and combine geological data and hydrological data to simulate the pollution diffusion paths in groundwater. However, traditional methods face many challenges in practical applications, especially when the geological structure is complex or multiple environmental factors are intertwined. Prediction errors often occur. In complex geological structures, different rock layers and soil permeability vary greatly, resulting in the diffusion path of pollutants in groundwater being highly nonlinear and unpredictable. In addition, traditional models have weak processing of time dynamics and usually rely on static data for prediction, and cannot update the model in real time to reflect the latest dynamics of pollution diffusion.

[0003] The main defects of the existing technology are reflected in the following aspects: First, the computational complexity of physical models and empirical models is high, and it is difficult to accurately describe the pollution diffusion process under complex geological conditions. In particular, when pollutants pass through different geological units, traditional methods are often unable to effectively capture the details of pollutant diffusion, resulting in insufficient accuracy of prediction results. Secondly, the existing technology has limited processing capabilities for real-time monitoring data. Groundwater pollution is usually a dynamic process. Pollutants will continue to diffuse over time, and existing methods cannot adjust the prediction model in time, resulting in a lag in pollution warning. In addition, traditional models rely heavily on artificially set parameters and assumptions, lack adaptability, and cannot learn the complex relationship of pollution diffusion through large-scale data. Especially when faced with the combined effects of multiple pollution sources and multiple types of pollutants, the prediction error will be further expanded.

[0004] In summary, the existing technology has the following shortcomings in predicting groundwater pollution paths: first, it is difficult for existing methods to accurately predict pollution diffusion paths in complex geological structures; second, the existing technology lacks the ability to dynamically adjust real-time data and cannot provide real-time pollution diffusion warnings; third, traditional models rely on fixed physical parameters, lack the ability to automatically learn pollution paths through big data, and cannot adapt to dynamically changing environmental factors. These shortcomings significantly affect the accuracy and timeliness of groundwater pollution path prediction. Summary of the invention

[0005] One purpose of the present invention is to propose a groundwater pollution path micro-analysis method based on graph neural network, which solves the problem in the prior art that it is difficult to accurately predict the pollution diffusion path under complex geological structures and the real-time dynamic response is insufficient.

[0006] A groundwater pollution path micro-analysis method based on a graph neural network according to an embodiment of the present invention comprises the following steps:

[0007] S1. Obtain historical data sets and real-time monitoring data sets of groundwater pollution, including geological structure permeability, pollutant types, pollution source locations, pollutant concentration distribution, and water flow velocity;

[0008] S2. Fuzzify the geological structure permeability, water flow velocity, pollutant type and pollutant concentration distribution parameters in the historical data set to convert the parameters into fuzzy variables, and use fuzzy membership functions to represent the language level of the fuzzy variables;

[0009] S3. Building a fuzzy logic rule base based on historical data sets to describe the fuzzy relationship of groundwater pollutant diffusion paths, the fuzzy logic rule base consisting of multiple fuzzy logic rules;

[0010] S4. Use a fuzzy inference system to process fuzzy variables in the historical data set, perform reasoning based on the fuzzy logic rule base, and obtain preliminary fuzzy prediction results of pollutant diffusion paths. The preliminary fuzzy prediction results represent the diffusion probability of pollutants under historical environmental conditions;

[0011] S5. Construct a groundwater pollution path diagram structure model based on historical data sets and real-time monitoring data sets. The attributes of nodes and edges include water flow velocity, pollutant diffusion rate, geological permeability, and diffusion probability obtained by the fuzzy inference system. The historical data set is used to construct the initial state of the groundwater pollution path diagram structure model, and the real-time monitoring data set is used to dynamically update the attributes of nodes and edges.

[0012] S6. Use the preliminary fuzzy prediction results as the input features of the graph neural network model to train the groundwater pollution path graph structure model, extract the relationship between geological structure and groundwater flow, construct the association matrix between nodes, learn the feature vectors of nodes and edges, and optimize the prediction model of pollutant diffusion path;

[0013] S7. Based on the trained graph neural network model, the water flow velocity, pollutant concentration and geological structure parameters in the real-time monitoring data set are input, the graph neural network model parameters are dynamically adjusted, and the results of the fuzzy inference system are combined to perform real-time prediction and analysis of the groundwater pollution diffusion path and its propagation range;

[0014] S8. By monitoring the sensor data in the data set in real time, dynamically update the node and edge attributes of the groundwater pollution path graph structure model, adjust the parameters of fuzzy logic reasoning and graph neural network in real time, generate real-time prediction of the pollution diffusion path, and update the prediction results when the pollution diffusion path changes;

[0015] S9. Output the analysis results of the groundwater pollution path and generate a visual diagram of the pollution diffusion path.

[0016] Optionally, S1 includes the following sub-steps:

[0017] S11. Constructing geological structure permeability data set P = {p1, p2, ..., p n}, where p i represents the permeability parameter of geological unit i, with a value range of [0,1], where 0 represents completely impermeable and 1 represents completely permeable;

[0018] S12, construct pollutant type data set C = {c1, c2, ..., c n}, where c i represents the pollutant type parameter of geological unit i, c i ∈{T o ,T m ,T c}, T o represents organic pollutants, T m represents heavy metal pollutants, T c Indicates chemical waste;

[0019] S13, construct the pollution source location data set L = {l1,l2,...,l n}, where l i The relative position of the pollution source in geological unit i is expressed using a two-dimensional coordinate system, l i =(x i ,y i ), x i and i is the coordinates of the pollution source on the map;

[0020] S14, construct pollutant concentration distribution data set = {z1,z2,...,z n}, where z i represents the pollutant concentration of geological unit i, and its value range is [0,1], where 0 means no pollutant and 1 means the pollutant concentration reaches the maximum value;

[0021] S15, construct water velocity data set V = {v1, v2, ..., v n}, where v i It represents the water velocity of geological unit i, in meters per second, with a range of v i ∈[0,V max ],V max is the maximum water flow velocity value in the area;

[0022] S16. Constructing historical data set D h, each data record d in the historical data set i Represented as a multidimensional vector d i =(p i ,c i ,l i ,z i ,v i ):

[0023]

[0024] S17. Build real-time monitoring data set D r , real-time monitoring of each data record r in the data set j Represented as a multidimensional vector r j =(p j ,c j ,l j ,z j ,v j ):

[0025]

[0026] Optionally, S2 includes the following sub-steps:

[0027] S21. For historical data set D h Medium geological structure permeability parameter p i Perform fuzzy processing and define the fuzzy membership function μ p (p i ), the geological permeability parameter p i Converted into fuzzy variables, the fuzzy membership function represents the geological permeability language level as low, medium, and high, and the corresponding membership functions are:

[0028]

[0029] in, Indicates that the geological permeability is low membership, represents the membership degree of medium permeability, The membership degree indicates high permeability;

[0030] S22. For historical data set D h Medium water velocity parameter v i Perform fuzzy processing and define the fuzzy membership function μ v (v i ), the water velocity parameter v i Converted into fuzzy variables, the fuzzy membership function represents the language level of water flow speed as slow, normal, and fast, and the corresponding membership functions are:

[0031]

[0032] in, Indicates the membership degree of slow water flow, represents the membership degree of normal speed, Indicates the membership of fast speed;

[0033] S23. For historical data set D h Pollutant type parameter c i Fuzzy processing is performed to classify pollutant types into organic pollutants T o , heavy metal pollutants T m and chemical waste c , define the fuzzy membership function μ of the pollutant type c (c i ), the corresponding membership functions are:

[0034]

[0035] in, represents the membership degree of organic pollutants, represents the membership degree of heavy metal pollutants, Indicates the membership degree of chemical waste;

[0036] S24. For historical data set D h The pollutant concentration parameter z i Perform fuzzy processing and define the fuzzy membership function μ z (z i ), the pollutant concentration parameter z i Converted into fuzzy variables, the fuzzy membership function represents the language level of pollutant concentration as low, medium, and high, and the corresponding membership functions are:

[0037]

[0038] in, represents the membership degree of low pollutant concentration, indicates a moderate concentration of membership, Indicates a high concentration of membership;

[0039] S25. Generate the corresponding fuzzy variable matrix F based on the fuzzy variables of geological structure permeability, water flow velocity, pollutant type and pollutant concentration h , the fuzzy variable matrix F h represents the preliminary fuzzy analysis results of groundwater pollution path, and the element f of the fuzzy variable matrix ij It is expressed as:

[0040]

[0041] in, and are the fuzzy membership function values ​​of geological structure permeability, water flow velocity, pollutant type and pollutant concentration, respectively, indicating the fuzzy level of each parameter; is the weight coefficient corresponding to each fuzzy variable, reflecting the influence of different variables on the pollution path; g(l ij ) is the distance attenuation function between geological unit i and geological unit j, which is used to correct the pollution diffusion effect between adjacent nodes and is defined as:

[0042]

[0043] Among them, α is the distance attenuation factor, d ij is the Euclidean distance between geological units i and j.

[0044] Optionally, the fuzzy logic rule is in the form of:

[0045] If the geological permeability is The water flow velocity is The pollutant type is The pollutant concentration is The probability of pollutant diffusion is P e , where P e A fuzzy value representing the probability of diffusion.

[0046] Optionally, the fuzzy logic rules specifically include:

[0047] If the geological permeability is high, the water flow velocity is fast, the pollutant type is organic pollutant, and the pollutant concentration is high, the probability of the pollutant diffusion path is high;

[0048] If the geological permeability is medium, the water flow rate is normal, the pollutant type is heavy metal pollutants, and the pollutant concentration is medium, then the possibility of the pollutant diffusion path is medium;

[0049] If the geological permeability is low, the water flow velocity is slow, the pollutant type is chemical waste, and the pollutant concentration is low, the probability of the pollutant diffusion path is low;

[0050] If the geological permeability is high, the water flow rate is slow, the pollutant type is heavy metal pollutants, and the pollutant concentration is high, the pollutant will diffuse slowly in the area but the pollution will remain for a long time;

[0051] If the geological permeability is medium, the water flow velocity is fast, the pollutant type is organic pollutants, and the pollutant concentration is low, the pollutants may spread to distant areas in a shorter time;

[0052] If the geological permeability is low, the water flow velocity is fast, the pollutant type is chemical waste, and the pollutant concentration is high, the probability of the pollutant diffusion path is medium and the diffusion process may be limited by geological conditions.

[0053] Optionally, S4 includes the following sub-steps:

[0054] S41, the fuzzy variable F in the historical data set h Input into the fuzzy inference system, the fuzzy inference system performs inference processing based on the fuzzy logic rule base, and infers the input fuzzy variables according to the fuzzy logic rule base in the fuzzy inference system to calculate the activation degree λ of each fuzzy logic rule k :

[0055]

[0056] Among them, λ k represents the activation degree of the kth fuzzy logic rule, and They are the fuzzy membership of geological permeability, water flow velocity, pollutant type and pollutant concentration respectively;

[0057] S43, weighted average all activated fuzzy logic rules, and calculate the preliminary fuzzy prediction result P of the pollutant diffusion path f :

[0058]

[0059] Among them, P f is the preliminary fuzzy prediction result of the pollutant diffusion path, indicating the diffusion probability of pollutants under historical environmental conditions. is the pollutant diffusion probability corresponding to the kth rule;

[0060] S44, according to the output result P of the fuzzy inference system f Generate preliminary fuzzy prediction maps of pollutant diffusion paths.

[0061] Optionally, S5 includes the following sub-steps:

[0062] S51. Based on historical data set D h and real-time monitoring dataset D r Constructing the groundwater pollution pathway diagram structural model:

[0063] G = (N, E);

[0064] Among them, N represents the node set of groundwater monitoring points or geological units, and E represents the edge set of groundwater flow paths and pollutant migration relationships;

[0065] S52, node n i∈N includes the geological permeability p i , water velocity v i 、Pollutant type c i , pollutant concentration z i And the diffusion probability P obtained by the fuzzy inference system f ;

[0066] S53, Edge ij The attributes of ∈E include node n i and node n j The water flow transfer rate between ij and the pollutant migration rate t ij , where the water flow rate r ij Depends on node n i and n j The water velocity and the geological permeability are defined as:

[0067]

[0068] Pollutant migration rate t ij Depends on node n i and n j The pollutant concentration and diffusion probability P f , defined as:

[0069]

[0070] S54, the historical data set D h The attributes of the nodes and edges in are initialized to the initial state of the groundwater pollution path graph structure model, and the association matrix A between the nodes is established. The association matrix indicates whether there is a water flow path between the nodes:

[0071]

[0072] S55. Using real-time monitoring data set D r Dynamically update the properties of nodes and edges in the groundwater pollution path graph structure model.

[0073] Optionally, S6 includes the following sub-steps:

[0074] S61, the preliminary fuzzy prediction result P f As input features, it is input into the nodes and edges of the groundwater pollution path graph structure model G = (N, E), node n i The eigenvector x i It is expressed as:

[0075] x i =[p i ,v i ,c i ,zi ,P f (i)];

[0076] Among them, p i 、v i 、c i 、z i represent geological permeability, water flow velocity, pollutant type and pollutant concentration respectively, P f (i) is node n i The diffusion probability of

[0077] S62. Use the graph neural network model to train the groundwater pollution path graph structure model, and combine the information of neighboring nodes to perform feature propagation to update the feature representation of each node

[0078]

[0079] in, For node n i The neighbor node set, A ij For node n i and node n j The incidence matrix element, W (l) and is the weight matrix of the graph neural network model, σ is the activation function, is the node n in the lth layer j The feature representation of , l is the number of layers of the graph neural network;

[0080] S63. Through multi-layer graph neural network training, the relationship between geological structure and groundwater flow is extracted iteratively, the feature representation of nodes and edges is updated, the prediction model of pollutant diffusion path is optimized, and finally the prediction result P of pollutant diffusion path is obtained. final .

[0081] The beneficial effects of the present invention are:

[0082] (1) The present invention forms a graph structure model of groundwater pollution path by representing groundwater monitoring points or geological units as nodes and groundwater flow paths and pollutant migration relationships as edges. Compared with traditional physical models, the graph structure model can fully capture the interaction relationship between nodes under complex geological structures, and combine the geological permeability, water flow velocity and pollutant type attributes extracted by the fuzzy reasoning system to improve the modeling accuracy of pollutant diffusion paths. By using graph neural networks to train and optimize the graph structure, the complex dynamic behaviors of groundwater flow and pollutant diffusion are effectively learned, which significantly improves the accuracy of pollution path prediction.

[0083] (2) The present invention introduces a fuzzy reasoning system into the traditional data-driven prediction model, and uses a fuzzy logic rule base to fuzzify the geological permeability, water flow velocity, pollutant type and concentration parameters in the historical data to generate a fuzzy variable matrix, which is then trained through a graph neural network. The fuzzy reasoning system can make a preliminary fuzzy prediction of the groundwater pollution diffusion path, while the graph neural network further optimizes the prediction results through the synergy of historical data and real-time monitoring data. The double-layer prediction mechanism not only makes up for the uncertainty defects of the existing model in dealing with complex geological environments, but also greatly improves the system's adaptability to the groundwater pollution diffusion path.

[0084] (3) The present invention constructs a groundwater pollution path graph structure model based on historical data sets and real-time monitoring data sets. After the real-time monitoring data is accessed, the attributes of the nodes and edges in the graph structure model can be dynamically updated, and the model parameters can be adjusted in real time in combination with the fuzzy reasoning system and the graph neural network. According to the latest dynamics of pollution diffusion, the prediction results can be quickly responded to and adjusted in real time, which significantly improves the timeliness of groundwater pollution path analysis and the ability to respond to sudden pollution incidents. Compared with the traditional static data prediction model, the present invention can provide more accurate real-time pollution diffusion path prediction and early warning in sudden pollution incidents, thereby enhancing the response efficiency of pollution control. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0086] Figure 1 This is a flow chart of a groundwater pollution path micro-analysis method based on graph neural network proposed by the present invention;

[0087] Figure 2 This is a groundwater pollution path graph structure model updated based on real-time monitoring data in a groundwater pollution path micro-analysis method based on graph neural network proposed in the present invention. DETAILED DESCRIPTION

[0088] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0089] refer to Figure 1-2 , a groundwater pollution path micro-analysis method based on graph neural network, including the following steps:

[0090] S1. Obtain historical data sets and real-time monitoring data sets of groundwater pollution, including geological structure permeability, pollutant types, pollution source locations, pollutant concentration distribution, and water flow velocity;

[0091] S2. Fuzzify the geological structure permeability, water flow velocity, pollutant type and pollutant concentration distribution parameters in the historical data set to convert the parameters into fuzzy variables, and use fuzzy membership functions to represent the language level of the fuzzy variables;

[0092] S3. Building a fuzzy logic rule base based on historical data sets to describe the fuzzy relationship of groundwater pollutant diffusion paths, the fuzzy logic rule base consisting of multiple fuzzy logic rules;

[0093] S4. Use a fuzzy inference system to process fuzzy variables in the historical data set, perform reasoning based on the fuzzy logic rule base, and obtain preliminary fuzzy prediction results of pollutant diffusion paths. The preliminary fuzzy prediction results represent the diffusion probability of pollutants under historical environmental conditions;

[0094] S5. Construct a groundwater pollution path diagram structure model based on historical data sets and real-time monitoring data sets. The attributes of nodes and edges include water flow velocity, pollutant diffusion rate, geological permeability, and diffusion probability obtained by the fuzzy inference system. The historical data set is used to construct the initial state of the groundwater pollution path diagram structure model, and the real-time monitoring data set is used to dynamically update the attributes of nodes and edges.

[0095] S6. Use the preliminary fuzzy prediction results as the input features of the graph neural network model to train the groundwater pollution path graph structure model, extract the relationship between geological structure and groundwater flow, construct the association matrix between nodes, learn the feature vectors of nodes and edges, and optimize the prediction model of pollutant diffusion path;

[0096] S7. Based on the trained graph neural network model, the water flow velocity, pollutant concentration and geological structure parameters in the real-time monitoring data set are input, the graph neural network model parameters are dynamically adjusted, and the results of the fuzzy inference system are combined to perform real-time prediction and analysis of the groundwater pollution diffusion path and its propagation range;

[0097] S8. By monitoring the sensor data in the data set in real time, dynamically update the node and edge attributes of the groundwater pollution path graph structure model, adjust the parameters of fuzzy logic reasoning and graph neural network in real time, generate real-time prediction of the pollution diffusion path, and update the prediction results when the pollution diffusion path changes;

[0098] S9. Output the analysis results of the groundwater pollution path and generate a visual diagram of the pollution diffusion path.

[0099] In this embodiment, S1 includes the following sub-steps:

[0100] S11. Constructing geological structure permeability data set P = {p1, p2, ..., p n}, where pi represents the permeability parameter of geological unit i, with a value range of [0,1], where 0 represents completely impermeable and 1 represents completely permeable;

[0101] S12, construct pollutant type data set C = {c1, c2, ..., c n}, where c i represents the pollutant type parameter of geological unit i, c i ∈{T o ,T m ,T c}, T o represents organic pollutants, T m represents heavy metal pollutants, T c Indicates chemical waste;

[0102] S13, construct the pollution source location data set L = {l1,l2,...,l n}, where l i The relative position of the pollution source in geological unit i is expressed using a two-dimensional coordinate system, l i =(x i ,y i ), x i and i is the coordinates of the pollution source on the map;

[0103] S14, construct pollutant concentration distribution data set = {z1,z2,...,z n}, where z i represents the pollutant concentration of geological unit i, and its value range is [0,1], where 0 means no pollutant and 1 means the pollutant concentration reaches the maximum value;

[0104] S15, construct water velocity data set V = {v1, v2, ..., v n}, where v i It represents the water velocity of geological unit i, in meters per second, with a range of v i ∈[0,V max ],V max is the maximum water flow velocity value in the area;

[0105] S16. Constructing historical data set D h , each data record d in the historical data set i Represented as a multidimensional vector d i =(p i ,c i ,l i ,z i ,v i ):

[0106]

[0107] S17. Build real-time monitoring data set D r , real-time monitoring of each data record r in the data set j Represented as a multidimensional vector r j =(p j ,c j ,l j ,z j ,v j ):

[0108]

[0109] In this embodiment, S2 includes the following sub-steps:

[0110] S21. For historical data set D h Medium geological structure permeability parameter p i Perform fuzzy processing and define the fuzzy membership function μ p (p i ), the geological permeability parameter p i Converted into fuzzy variables, the fuzzy membership function represents the geological permeability language level as low, medium, and high, and the corresponding membership functions are:

[0111]

[0112] in, Indicates that the geological permeability is low membership, represents the membership degree of medium permeability, The membership degree indicates high permeability;

[0113] S22. For historical data set D h Medium water velocity parameter v i Perform fuzzy processing and define the fuzzy membership function μ v (v i ), the water velocity parameter v i Converted into fuzzy variables, the fuzzy membership function represents the language level of water flow speed as slow, normal, and fast, and the corresponding membership functions are:

[0114]

[0115] in, Indicates the membership degree of slow water flow, represents the membership degree of normal speed, Indicates the membership of fast speed;

[0116] S23. For historical data set D h Pollutant type parameter ci Fuzzy processing is performed to classify pollutant types into organic pollutants T o , heavy metal pollutants T m and chemical waste c , define the fuzzy membership function μ of the pollutant type c (c i ), the corresponding membership functions are:

[0117]

[0118] in, represents the membership degree of organic pollutants, represents the membership degree of heavy metal pollutants, Indicates the membership degree of chemical waste;

[0119] S24. For historical data set D h The pollutant concentration parameter z i Perform fuzzy processing and define the fuzzy membership function μ z (z i ), the pollutant concentration parameter z i Converted into fuzzy variables, the fuzzy membership function represents the language level of pollutant concentration as low, medium, and high, and the corresponding membership functions are:

[0120]

[0121] in, represents the membership degree of low pollutant concentration, indicates a moderate concentration of membership, Indicates a high concentration of membership;

[0122] S25. Generate the corresponding fuzzy variable matrix F based on the fuzzy variables of geological structure permeability, water flow velocity, pollutant type and pollutant concentration h , the fuzzy variable matrix F h represents the preliminary fuzzy analysis results of groundwater pollution path, and the element f of the fuzzy variable matrix ij It is expressed as:

[0123]

[0124] in, and are the fuzzy membership function values ​​of geological structure permeability, water flow velocity, pollutant type and pollutant concentration, respectively, indicating the fuzzy level of each parameter; is the weight coefficient corresponding to each fuzzy variable, reflecting the influence of different variables on the pollution path; g(l ij) is the distance attenuation function between geological unit i and geological unit j, which is used to correct the pollution diffusion effect between adjacent nodes and is defined as:

[0125]

[0126] Among them, α is the distance attenuation factor, d ij is the Euclidean distance between geological units i and j.

[0127] In this implementation mode, the fuzzy logic rule form is:

[0128] If the geological permeability is The water flow velocity is The pollutant type is The pollutant concentration is The probability of pollutant diffusion is P e , where P e A fuzzy value representing the probability of diffusion.

[0129] In this implementation mode, the fuzzy logic rules specifically include:

[0130] If the geological permeability is high, the water flow velocity is fast, the pollutant type is organic pollutant, and the pollutant concentration is high, the probability of the pollutant diffusion path is high;

[0131] If the geological permeability is medium, the water flow rate is normal, the pollutant type is heavy metal pollutants, and the pollutant concentration is medium, then the possibility of the pollutant diffusion path is medium;

[0132] If the geological permeability is low, the water flow velocity is slow, the pollutant type is chemical waste, and the pollutant concentration is low, the probability of the pollutant diffusion path is low;

[0133] If the geological permeability is high, the water flow rate is slow, the pollutant type is heavy metal pollutants, and the pollutant concentration is high, the pollutant will diffuse slowly in the area but the pollution will remain for a long time;

[0134] If the geological permeability is medium, the water flow velocity is fast, the pollutant type is organic pollutants, and the pollutant concentration is low, the pollutants may spread to distant areas in a shorter time;

[0135] If the geological permeability is low, the water flow velocity is fast, the pollutant type is chemical waste, and the pollutant concentration is high, the probability of the pollutant diffusion path is medium and the diffusion process may be limited by geological conditions.

[0136] In this embodiment, S4 includes the following sub-steps:

[0137] S41, the fuzzy variable F in the historical data set hInput into the fuzzy inference system, the fuzzy inference system performs inference processing based on the fuzzy logic rule base, and infers the input fuzzy variables according to the fuzzy logic rule base in the fuzzy inference system to calculate the activation degree λ of each fuzzy logic rule k :

[0138]

[0139] Among them, λ k represents the activation degree of the kth fuzzy logic rule, and They are the fuzzy membership of geological permeability, water flow velocity, pollutant type and pollutant concentration respectively;

[0140] S43, weighted average all activated fuzzy logic rules, and calculate the preliminary fuzzy prediction result P of the pollutant diffusion path f :

[0141]

[0142] Among them, P f is the preliminary fuzzy prediction result of the pollutant diffusion path, indicating the diffusion probability of pollutants under historical environmental conditions. is the pollutant diffusion probability corresponding to the kth rule;

[0143] S44, according to the output result P of the fuzzy inference system f Generate preliminary fuzzy prediction maps of pollutant diffusion paths.

[0144] In this embodiment, S5 includes the following sub-steps:

[0145] S51. Based on historical data set D h and real-time monitoring dataset D r Constructing the groundwater pollution pathway diagram structural model:

[0146] G = (N, E);

[0147] Among them, N represents the node set of groundwater monitoring points or geological units, and E represents the edge set of groundwater flow paths and pollutant migration relationships;

[0148] S52, node n i ∈N includes the geological permeability p i , water velocity v i 、Pollutant type c i , pollutant concentration z i And the diffusion probability P obtained by the fuzzy inference system f ;

[0149] S53, Edge ijThe attributes of ∈E include node n i and node n j The water flow transfer rate between ij and the pollutant migration rate t ij , where the water flow rate r ij Depends on node n i and n j The water velocity and the geological permeability are defined as:

[0150]

[0151] Pollutant migration rate t ij Depends on node n i and n j The pollutant concentration and diffusion probability P f , defined as:

[0152]

[0153] S54, the historical data set D h The attributes of the nodes and edges in are initialized to the initial state of the groundwater pollution path graph structure model, and the association matrix A between the nodes is established. The association matrix indicates whether there is a water flow path between the nodes:

[0154]

[0155] S55. Using real-time monitoring data set D r Dynamically update the properties of nodes and edges in the groundwater pollution path graph structure model.

[0156] In this embodiment, S6 includes the following sub-steps:

[0157] S61, the preliminary fuzzy prediction result P f As input features, it is input into the nodes and edges of the groundwater pollution path graph structure model G = (N, E), node n i The eigenvector x i It is expressed as:

[0158] x i =[p i ,v i ,c i ,z i ,P f (i)];

[0159] Among them, p i 、v i 、c i 、z i represent geological permeability, water flow velocity, pollutant type and pollutant concentration respectively, P f(i) is node n i The diffusion probability of

[0160] S62. Use the graph neural network model to train the groundwater pollution path graph structure model, and combine the information of neighboring nodes to perform feature propagation to update the feature representation of each node

[0161]

[0162] in, For node n i The neighbor node set, A ij For node n i and node n j The incidence matrix element, W (l) and is the weight matrix of the graph neural network model, σ is the activation function, is the node n in the lth layer j The feature representation of , l is the number of layers of the graph neural network;

[0163] S63. Through multi-layer graph neural network training, the relationship between geological structure and groundwater flow is extracted iteratively, the feature representation of nodes and edges is updated, the prediction model of pollutant diffusion path is optimized, and finally the prediction result P of pollutant diffusion path is obtained. final .

[0164] Embodiment 1:

[0165] In order to demonstrate the feasibility and effectiveness of the present invention, this paper simulates the application process of the groundwater pollution path micro-analysis method based on graph neural network under complex geological conditions in combination with an actual groundwater pollution incident.

[0166] In June 2022, during the environmental monitoring of an industrial zone in North China, the monitoring personnel found through the real-time monitoring system that the data of a groundwater monitoring point in the area was abnormal and the pollutant concentration exceeded the warning value. Due to the long-standing industrial wastewater discharge problem in the industrial zone, the spread of groundwater pollution has always been the focus of attention of the local environmental protection department. The regional geological structure is complex, with multiple layers of clay and gravel rock geological units, and the pollutant diffusion paths are diverse and unpredictable. To solve the problem, the local environmental protection department decided to use a groundwater pollution path micro-analysis method based on graph neural network to monitor the spread of pollution in real time and warn of potential risks.

[0167] On June 15, 2022, the groundwater pollution monitoring system in an industrial zone in North China detected abnormal groundwater pollution concentration at the monitoring point ID "NDW-017". The pollutant was the organic substance benzene, and the concentration value reached 0.75mg / L, exceeding the standard limit. The monitoring data further showed that the pollutant gradually appeared on the diffusion path of the adjacent monitoring point NDW-018, and the pollutant concentration of NDW-018 increased from 0.45mg / L to 0.6mg / L. The system quickly modeled the graph structure of the pollution path through fuzzy processing of historical data and combined it with real-time monitoring data.

[0168] Based on the analysis method of the present invention, the system preliminarily processes the monitoring data of NDW-017 and NDW-018 through the fuzzy inference system. After the key parameters of geological permeability, water flow velocity, and pollutant concentration are fuzzified, a diffusion graph model of pollutants is constructed. In the model, NDW-017 and NDW-018 are modeled as nodes, and the edges between the nodes represent the water flow path and the migration rate of pollutants. Through the training of the graph neural network, the system predicts the diffusion path of pollutants in real time. The system analysis shows that the diffusion rate of pollutants will accelerate in the next 24 hours, which may affect the more distant groundwater monitoring point NDW-020.

[0169] The system immediately generated a forecast report on the spread of pollution. The report showed that within the next 24 hours, the pollutant concentration of NDW-020 is expected to rise from the current 0.25 mg / L to 0.55 mg / L, posing a potential pollution risk. The system sent the forecast results to the local environmental protection department and issued an early warning signal. The monitoring personnel quickly took preventive measures based on the pollution path prediction map generated by the system and launched a pollution interception project near NDW-020 to prevent pollutants from further spreading to the local residents' drinking water area.

[0170] On June 17, 2022, the system updated the monitoring data again. Real-time monitoring data showed that the pollutant concentration of NDW-020 did reach 0.52 mg / L, which basically met the system's prediction results. By updating the graph structure model in real time, the system continued to monitor the path of pollution diffusion and adjusted the prediction results in time. New analysis shows that the trend of pollutant diffusion has been effectively controlled, and the pollution concentration will gradually decrease in the next 48 hours.

[0171] In order to verify the effectiveness and superiority of the method of the present invention, this paper compares it with the traditional groundwater pollutant migration model. Taking the data from June 15 to June 17, 2022 as an example, the following comparison table shows the efficiency and accuracy of the two methods in dealing with groundwater pollution diffusion problems:

[0172] Table 1 Comparison of prediction time and error

[0173]

[0174] As can be seen from Table 1 above, the traditional physical model takes a long time to predict the pollution diffusion path and has a high error rate. Since the traditional method relies on static data models and cannot dynamically update parameters, it is impossible to make accurate predictions and responses in a short time. In contrast, the present invention can quickly and accurately predict the diffusion path of pollutants by combining real-time monitoring data with historical data, and significantly improves the prediction accuracy through the dynamic adjustment capability of the graph neural network. Especially in the event of sudden pollution, the method of the present invention can issue a warning signal within 30 minutes, providing sufficient response time for the environmental protection department in advance.

[0175] In order to further verify the effectiveness of the method of the present invention, we compared the prediction errors using the traditional model and the method of the present invention from June 15 to June 17, 2022:

[0176] Table 2 Comparison of monitoring data and prediction results

[0177]

[0178]

[0179] As can be seen from Table 2 above, the traditional physical model has a large error when dealing with complex geological structures and multiple environmental factors, especially at the NDW-020 monitoring point, where the prediction error of the traditional method is as high as 33%, while the prediction error of the method of the present invention is controlled within 3% through real-time data update and graph neural network training. This result further proves the high accuracy and adaptability of the method of the present invention in complex environments.

[0180] This embodiment verifies the superiority of the groundwater pollution path micro-analysis method based on graph neural network in solving the problem of pollution diffusion prediction under complex geological conditions through the simulation of actual groundwater pollution events. Through the method of the present invention, the system can achieve high-precision prediction of pollution diffusion paths and quickly respond to dynamic changes in pollution diffusion with the support of real-time monitoring data, providing timely and reliable early warning signals for environmental protection departments. Compared with traditional physical models, the present invention not only shows obvious advantages in prediction accuracy and response time, but also can update the model in real time, ensuring accurate control and effective governance of complex groundwater pollution problems.

[0181] The present invention forms a graph structure model of groundwater pollution path by representing groundwater monitoring points or geological units as nodes and groundwater flow paths and pollutant migration relationships as edges. Compared with traditional physical models, the graph structure model can fully capture the interaction relationship between nodes under complex geological structures, and combine the geological permeability, water flow velocity and pollutant type attributes extracted by the fuzzy reasoning system to improve the modeling accuracy of pollutant diffusion paths. By using graph neural networks to train and optimize the graph structure, the complex dynamic behaviors of groundwater flow and pollutant diffusion are effectively learned, which significantly improves the accuracy of pollution path prediction.

[0182] The present invention introduces a fuzzy inference system into the traditional data-driven prediction model, and uses the fuzzy logic rule base to fuzzify the geological permeability, water flow velocity, pollutant type and concentration parameters in the historical data to generate a fuzzy variable matrix, which is then trained through a graph neural network. The fuzzy inference system can make a preliminary fuzzy prediction of the groundwater pollution diffusion path, and the graph neural network further optimizes the prediction results through the synergy of historical data and real-time monitoring data. The double-layer prediction mechanism not only makes up for the uncertainty defects of the existing model in dealing with complex geological environments, but also greatly improves the system's adaptability to the groundwater pollution diffusion path.

[0183] The present invention constructs a groundwater pollution path graph structure model based on historical data sets and real-time monitoring data sets. After the real-time monitoring data is accessed, the attributes of the nodes and edges in the graph structure model can be dynamically updated, and the model parameters can be adjusted in real time in combination with the fuzzy reasoning system and the graph neural network. The prediction results can be quickly responded to and adjusted in real time according to the latest dynamics of pollution diffusion, which significantly improves the timeliness of groundwater pollution path analysis and the ability to respond to sudden pollution incidents. Compared with the traditional static data prediction model, the present invention can provide more accurate real-time pollution diffusion path prediction and early warning in sudden pollution incidents, thereby enhancing the response efficiency of pollution control.

[0184] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A groundwater pollution path micro-analysis method based on graph neural network, characterized in that: The steps include: S1. Obtain historical data sets and real-time monitoring data sets of groundwater pollution, including geological structure permeability, pollutant types, pollution source locations, pollutant concentration distribution, and water flow velocity; S2. Fuzzify the geological structure permeability, water flow velocity, pollutant type and pollutant concentration distribution parameters in the historical data set to convert the parameters into fuzzy variables, and use fuzzy membership functions to represent the language level of the fuzzy variables; S3. Building a fuzzy logic rule base based on historical data sets to describe the fuzzy relationship of groundwater pollutant diffusion paths, the fuzzy logic rule base consisting of multiple fuzzy logic rules; S4. Use a fuzzy inference system to process fuzzy variables in the historical data set, perform reasoning based on the fuzzy logic rule base, and obtain preliminary fuzzy prediction results of pollutant diffusion paths. The preliminary fuzzy prediction results represent the diffusion probability of pollutants under historical environmental conditions; S5. Construct a groundwater pollution path diagram structure model based on historical data sets and real-time monitoring data sets. The attributes of nodes and edges include water flow velocity, pollutant diffusion rate, geological permeability, and diffusion probability obtained by the fuzzy inference system. The historical data set is used to construct the initial state of the groundwater pollution path diagram structure model, and the real-time monitoring data set is used to dynamically update the attributes of nodes and edges. S6. Use the preliminary fuzzy prediction results as the input features of the graph neural network model to train the groundwater pollution path graph structure model, extract the relationship between geological structure and groundwater flow, construct the association matrix between nodes, learn the feature vectors of nodes and edges, and optimize the prediction model of pollutant diffusion path; S7. Based on the trained graph neural network model, the water flow velocity, pollutant concentration and geological structure parameters in the real-time monitoring data set are input, the graph neural network model parameters are dynamically adjusted, and the results of the fuzzy inference system are combined to perform real-time prediction and analysis of the groundwater pollution diffusion path and its propagation range; S8. By monitoring the sensor data in the data set in real time, dynamically update the node and edge attributes of the groundwater pollution path graph structure model, adjust the parameters of fuzzy logic reasoning and graph neural network in real time, generate real-time prediction of the pollution diffusion path, and update the prediction results when the pollution diffusion path changes; S9. Output the analysis results of the groundwater pollution path and generate a visual diagram of the pollution diffusion path.

2. According to claim 1, a groundwater pollution path micro-analysis method based on graph neural network is characterized in that: The S1 comprises the following sub-steps: S11. Constructing geological structure permeability data set P = {p1, p2, ..., p n }, where p i represents the permeability parameter of geological unit i, with a value range of [0,1], where 0 represents completely impermeable and 1 represents completely permeable; S12, construct pollutant type data set C = {c1, c2, ..., c n }, where c i represents the pollutant type parameter of geological unit i, c i ∈{T o ,T m ,T c }, T o represents organic pollutants, T m represents heavy metal pollutants, T c Indicates chemical waste; S13, construct the pollution source location data set L = {l1,l2,...,l n }, where l i The relative position of the pollution source in geological unit i is expressed using a two-dimensional coordinate system, l i =(x i ,y i ), x i and i is the coordinates of the pollution source on the map; S14, construct pollutant concentration distribution data set = {z1,z2,...,z n }, where z i represents the pollutant concentration of geological unit i, and its value range is [0,1], where 0 means no pollutant and 1 means the pollutant concentration reaches the maximum value; S15, construct water velocity data set V = {v1, v2, ..., v n }, where v i It represents the water velocity of geological unit i, in meters per second, with a range of v i ∈[0,V max ],V max is the maximum water flow velocity value in the area; S16. Constructing historical data set D h , each data record d in the historical data set i Represented as a multidimensional vector d i =(p i ,c i ,l i ,z i ,v i ): S17. Build real-time monitoring data set D r , real-time monitoring of each data record r in the data set j Represented as a multidimensional vector r j =(p j ,c j ,l j ,z j ,v j ):

3. According to claim 1, a groundwater pollution path micro-analysis method based on graph neural network is characterized in that: The S2 comprises the following sub-steps: S21. For historical data set D h Medium geological structure permeability parameter p i Perform fuzzy processing and define the fuzzy membership function μ p (p i ), the geological permeability parameter p i Converted into fuzzy variables, the fuzzy membership function represents the geological permeability language level as low, medium, and high, and the corresponding membership functions are: in, Indicates that the geological permeability is low membership, represents the membership degree of medium permeability, The membership degree indicates high permeability; S22. For historical data set D h Medium water velocity parameter v i Perform fuzzy processing and define the fuzzy membership function μ v (v i ), the water velocity parameter v i Converted into fuzzy variables, the fuzzy membership function represents the language level of water flow speed as slow, normal, and fast, and the corresponding membership functions are: in, Indicates the membership degree of slow water flow, Indicates the degree of membership of normal speed, Indicates the membership of fast speed; S23. For historical data set D h Pollutant type parameter c i Fuzzy processing is performed to classify pollutant types into organic pollutants T o , heavy metal pollutants T m and chemical waste c , define the fuzzy membership function μ of the pollutant type c (c i ), the corresponding membership functions are: in, represents the membership degree of organic pollutants, represents the membership degree of heavy metal pollutants, Indicates the membership degree of chemical waste; S24. For historical data set D h The pollutant concentration parameter z i Perform fuzzy processing and define the fuzzy membership function μ z (z i ), the pollutant concentration parameter z i Converted into fuzzy variables, the fuzzy membership function represents the language level of pollutant concentration as low, medium, and high, and the corresponding membership functions are: in, represents the membership degree of low pollutant concentration, indicates a moderate concentration of membership, Indicates a high concentration of membership; S25. Generate the corresponding fuzzy variable matrix F based on the fuzzy variables of geological structure permeability, water flow velocity, pollutant type and pollutant concentration h , the fuzzy variable matrix F h represents the preliminary fuzzy analysis results of groundwater pollution path, and the element f of the fuzzy variable matrix ij It is expressed as: in, and are the fuzzy membership function values ​​of geological structure permeability, water flow velocity, pollutant type and pollutant concentration, respectively, indicating the fuzzy level of each parameter; is the weight coefficient corresponding to each fuzzy variable, reflecting the influence of different variables on the pollution path; g(l ij ) is the distance attenuation function between geological unit i and geological unit j, which is used to correct the pollution diffusion effect between adjacent nodes and is defined as: Among them, α is the distance attenuation factor, d ij is the Euclidean distance between geological units i and j.

4. According to claim 1, a groundwater pollution path micro-analysis method based on graph neural network is characterized in that: The fuzzy logic rule form is: If the geological permeability is The water flow velocity is The pollutant type is The pollutant concentration is The probability of pollutant diffusion is P e , where P e A fuzzy value representing the probability of diffusion.

5. A method for microscopic analysis of groundwater pollution paths based on graph neural networks according to claim 4, characterized in that: The fuzzy logic rules specifically include: If the geological permeability is high, the water flow velocity is fast, the pollutant type is organic pollutant, and the pollutant concentration is high, the probability of the pollutant diffusion path is high; If the geological permeability is medium, the water flow rate is normal, the pollutant type is heavy metal pollutants, and the pollutant concentration is medium, then the possibility of the pollutant diffusion path is medium; If the geological permeability is low, the water flow velocity is slow, the pollutant type is chemical waste, and the pollutant concentration is low, the probability of the pollutant diffusion path is low; If the geological permeability is high, the water flow rate is slow, the pollutant type is heavy metal pollutants, and the pollutant concentration is high, the pollutant will diffuse slowly in the area but the pollution will remain for a long time; If the geological permeability is medium, the water flow velocity is fast, the pollutant type is organic pollutants, and the pollutant concentration is low, the pollutants may spread to distant areas in a shorter time; If the geological permeability is low, the water flow velocity is fast, the pollutant type is chemical waste, and the pollutant concentration is high, the probability of the pollutant diffusion path is medium and the diffusion process may be limited by geological conditions.

6. The method for microscopic analysis of groundwater pollution paths based on graph neural network according to claim 1 is characterized in that: The S4 comprises the following sub-steps: S41, the fuzzy variable F in the historical data set h Input into the fuzzy inference system, the fuzzy inference system performs inference processing based on the fuzzy logic rule base, and infers the input fuzzy variables according to the fuzzy logic rule base in the fuzzy inference system to calculate the activation degree λ of each fuzzy logic rule k : Among them, λ k represents the activation degree of the kth fuzzy logic rule, and They are the fuzzy membership of geological permeability, water flow velocity, pollutant type and pollutant concentration respectively; S43, weighted average all activated fuzzy logic rules, and calculate the preliminary fuzzy prediction result P of the pollutant diffusion path f : Among them, P f is the preliminary fuzzy prediction result of the pollutant diffusion path, indicating the diffusion probability of pollutants under historical environmental conditions. is the pollutant diffusion probability corresponding to the kth rule; S44, according to the output result P of the fuzzy inference system f Generate preliminary fuzzy prediction maps of pollutant diffusion paths.

7. The method for microscopic analysis of groundwater pollution paths based on graph neural networks according to claim 1 is characterized in that: The S5 comprises the following sub-steps: S51. Based on historical data set D h and real-time monitoring dataset D r Constructing the groundwater pollution pathway diagram structural model: G = (N, E); Among them, N represents the node set of groundwater monitoring points or geological units, and E represents the edge set of groundwater flow paths and pollutant migration relationships; S52, node n i ∈N includes the geological permeability p i , water velocity v i , Pollutant type c i , pollutant concentration z i And the diffusion probability P obtained by the fuzzy inference system f ; S53, Edge ij The attributes of ∈E include node n i and node n j The water flow transfer rate between ij and the pollutant migration rate t ij , where the water flow rate r ij Depends on node n i and n j The water velocity and the geological permeability are defined as: Pollutant migration rate t ij Depends on node n i and n j The pollutant concentration and diffusion probability P f , defined as: S54, the historical data set D h The attributes of the nodes and edges in are initialized to the initial state of the groundwater pollution path graph structure model, and the association matrix A between the nodes is established. The association matrix indicates whether there is a water flow path between the nodes: S55. Using real-time monitoring data set D r Dynamically update the properties of nodes and edges in the groundwater pollution pathway graph structure model.

8. The method for microscopic analysis of groundwater pollution paths based on graph neural network according to claim 1 is characterized in that: The S6 comprises the following sub-steps: S61, the preliminary fuzzy prediction result P f As input features, it is input into the nodes and edges of the groundwater pollution path graph structure model G = (N, E), node n i The eigenvector x i It is expressed as: x i =[p i ,v i ,c i ,z i ,P f (i)]; Among them, p i 、v i 、c i 、z i represent geological permeability, water flow velocity, pollutant type and pollutant concentration respectively, P f (i) is node n i The diffusion probability of S62. Use the graph neural network model to train the groundwater pollution path graph structure model, and combine the information of neighboring nodes to perform feature propagation to update the feature representation of each node in, For node n i The neighbor node set A ij For node n i and node n j The incidence matrix element, W (l) and is the weight matrix of the graph neural network model, σ is the activation function, is the node n in the first layer j The feature representation of , l is the number of layers of the graph neural network; S63. Through multi-layer graph neural network training, the relationship between geological structure and groundwater flow is extracted iteratively, the feature representation of nodes and edges is updated, the prediction model of pollutant diffusion path is optimized, and finally the prediction result P of pollutant diffusion path is obtained. final .

Images