Pollution traceability analysis method and system based on water quality sampling data

Abstract

The invention belongs to the technical field of pollution traceability, and provides a pollution traceability analysis method and system based on water quality sampling data. The method comprises the following steps: receiving multiple pieces of returned water quality monitoring data corresponding to a target sampling area, and obtaining geographic information, hydrological information and known pollution source information of the target sampling area; performing evaluation by adopting an analytic hierarchy process to obtain hydrological complexity of the target sampling area, and formulating a simulation strategy of the pollution diffusion simulation model; controlling the pollution diffusion simulation model to perform analogue simulation according to the simulation strategy to obtain a plurality of pollutant traceability path simulation results; and screening the plurality of pollutant traceability path simulation results by using a generative adversarial network to obtain an optimal pollutant traceability path simulation result. According to the invention, accurate and efficient pollution tracing can be realized, a reliable basis is provided for treatment, and the pertinence and effectiveness of pollution treatment can be remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of pollution source tracing, and in particular to a pollution source tracing analysis method and system based on water quality sampling data. Background Art

[0002] Traditional pollution source tracing methods often rely on manual judgment or simple physical and chemical analysis, which are subject to strong subjectivity, low efficiency, and limited coverage. For example, manual investigation of pollution sources often requires a large amount of manpower and time, and is limited by the professional level of investigators and on-site environmental conditions, making it easy to miss or misjudge. Analysis methods based on data from a single monitoring point are difficult to reflect the migration and transformation patterns of pollutants in water bodies, and cannot establish a complete traceability path.

[0003] In recent years, with the development of sensor technology and the Internet of Things, online water quality monitoring systems have become increasingly popular, enabling real-time data collection of parameters such as pollutant concentrations, pH, and dissolved oxygen. However, existing systems focus primarily on data collection and display, lacking the ability to deeply mine and integrate monitoring data for analysis. Furthermore, various factors can lead to insufficient deployment of monitoring nodes in online water quality monitoring systems, making it difficult to accurately mine these data using conventional data mining methods.

[0004] Therefore, how to conduct in-depth mining of multi-source monitoring data to accurately identify the sources and diffusion paths of pollutants has become a technical problem that needs to be solved urgently in the current water environment field. Summary of the Invention

[0005] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a pollution source tracing analysis method and system based on water quality sampling data.

[0006] The present invention provides a pollution source tracing analysis method based on water quality sampling data, which is applied to a data processing center and includes the following steps: Receive multiple pieces of water quality monitoring data corresponding to the target sampling area, each piece of water quality monitoring data including geographic information of the sampling point and water quality monitoring indicators; and obtain geographic information, hydrological information, and known pollution source information of the target sampling area; Using the analytic hierarchy process to evaluate the water quality monitoring data, geographic information, hydrological information, and known pollution source information to obtain the hydrological complexity of the target sampling area, and formulating a simulation strategy for the pollution diffusion simulation model based on the hydrological complexity; the simulation strategy includes the simulation start time, time step, and spatial grid accuracy; Inputting the water quality monitoring data, geographic information, hydrological information and known pollution source information into the pollution diffusion simulation model, performing simulation according to the simulation strategy, and obtaining multiple pollutant tracing path simulation results; the pollution diffusion simulation model has built-in convection-diffusion equations and pollutant degradation kinetics modules; Use a pre-trained generative adversarial network to screen multiple pollutant tracing path simulation results to obtain the optimal pollutant tracing path simulation result.

[0007] The present invention also provides a pollution source tracing analysis system based on water quality sampling data, which is applied to a data processing center and includes an acquisition unit, a simulation strategy formulation unit, a simulation unit, and a path screening unit; The acquisition unit receives a plurality of water quality monitoring data corresponding to the target sampling area, each of which includes geographic information of the sampling point and water quality monitoring indicators; and acquires geographic information, hydrological information and known pollution source information of the target sampling area; The simulation strategy formulation unit uses a hierarchical analysis method to evaluate the water quality monitoring data, geographic information, hydrological information, and known pollution source information to obtain the hydrological complexity of the target sampling area, and formulates a simulation strategy for the pollution diffusion simulation model based on the hydrological complexity; the simulation strategy includes a simulation start time, a time step, and a spatial grid accuracy; The simulation unit inputs the water quality monitoring data, geographic information, hydrological information, and known pollution source information into a pollution diffusion simulation model, performs simulation according to the simulation strategy, and obtains multiple pollutant tracing path simulation results; the pollution diffusion simulation model has a built-in convection-diffusion equation and a pollutant degradation kinetics module; The path screening unit uses a pre-trained generative adversarial network to screen multiple pollutant tracing path simulation results to obtain the optimal pollutant tracing path simulation result.

[0008] This method integrates multi-source water quality and environmental data, uses the Analytic Hierarchy Process (AHP) to assess hydrological complexity and develop adaptive simulation strategies. It then combines convection-diffusion and degradation modules to generate multiple pathways, and then uses a pre-trained generative adversarial network to select the optimal pathway. This method enables precise and efficient pollution source tracing, providing a reliable basis for pollution control, and significantly improving the targeted and effective nature of pollution control. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 This is a flow chart of a pollution source tracing analysis method based on water quality sampling data disclosed in an embodiment of the present invention; Figure 2 It is a schematic diagram of the system architecture of the solution of the present invention; Figure 3Schematic diagram of the structure of the pollution diffusion simulation model disclosed in an embodiment of the present invention; Figure 4 This is a structural diagram of a pollution source tracing and analysis system based on water quality sampling data disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0010] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0011] like Figure 1 As shown, an embodiment of the present invention discloses a pollution source tracing analysis method based on water quality sampling data, which is applied to a data processing center and includes the following steps: S10, receiving multiple water quality monitoring data corresponding to the target sampling area, each of the water quality monitoring data includes sampling point geographic information and water quality monitoring indicators; and obtaining geographic information, hydrological information and known pollution source information of the target sampling area.

[0012] like Figure 2 As shown, the solution of the present invention is applied to a data processing center connected to an online water quality monitoring system. The online water quality monitoring system includes distributed IoT sensors installed in various water areas within the target sampling area. Simultaneously, the data processing center is connected to various water quality monitoring stations or mobile water quality monitoring terminals to receive water quality monitoring data analyzed from water samples collected on-site by the water quality monitoring stations, as well as water quality monitoring data transmitted back by field sampling personnel after sampling and analyzing water quality at a small number of on-site sampling points. This water quality monitoring data includes the geographic information of the corresponding sampling points.

[0013] The data processing center receives multiple pieces of water quality monitoring data from various sampling points within the target sampling area. The geographic information of the sampling points represents the spatial location of the data, while water quality monitoring indicators (such as pollutant concentration, pH, and dissolved oxygen) reflect the pollution status of the water body. The data processing center also obtains geographic information about the target sampling area (including basin topography and river system distribution), hydrological information (such as flow rate, flow rate, and water temperature), and information on known pollution sources (such as the location and discharge status of industrial wastewater outlets and domestic sewage treatment plants).

[0014] S20, using the hierarchical analysis method to evaluate the water quality monitoring data, geographic information, hydrological information and known pollution source information to obtain the hydrological complexity of the target sampling area, and formulating a simulation strategy for the pollution diffusion simulation model based on the hydrological complexity; the simulation strategy includes the simulation start time, time step and spatial grid accuracy.

[0015] The analytic hierarchy process is used to comprehensively evaluate the various data obtained in step S10. The analytic hierarchy process can decompose complex evaluation problems into different levels. By determining the weights of various factors and performing comprehensive calculations, the hydrological complexity of the target sampling area is obtained.

[0016] Based on this hydrological complexity, a simulation strategy for the pollution diffusion simulation model was developed. For example, the timing of the simulation start should be determined based on the type of pollution event (e.g., sudden pollution starts at the time of data anomaly, while routine monitoring starts at a stable period). The simulation time step is adjusted based on complexity, with smaller step sizes used in high-complexity areas to capture subtle changes, while larger step sizes can be appropriately increased in low-complexity areas. The spatial grid accuracy is also matched to the complexity, with high-precision grids used in complex terrain and relatively coarse grids in flat areas.

[0017] Such a setting can ensure that the pollution diffusion simulation model can adapt to different hydrological environments and improve the pertinence and efficiency of the simulation.

[0018] S30, inputting the water quality monitoring data, geographic information, hydrological information and known pollution source information into the pollution diffusion simulation model, performing simulation according to the simulation strategy, and obtaining multiple pollutant tracing path simulation results; the pollution diffusion simulation model has built-in convection-diffusion equations and pollutant degradation kinetics modules.

[0019] After determining the appropriate simulation strategy, the water quality monitoring data, geographic information, hydrological information, and known pollution source information obtained above were input into the pollution diffusion simulation model. This pollution diffusion simulation model incorporates a built-in convection-diffusion equation to simulate the migration and diffusion of pollutants under the influence of water flow, while the pollutant degradation kinetics module considers the chemical and biological degradation of pollutants in the water.

[0020] When running simulations according to the aforementioned simulation strategy, the pollution diffusion simulation model will run based on different potential pollution source combinations and other scenarios, ultimately generating multiple pollutant source path simulation results. These results include information such as the possible source of the pollutants and the key nodes through which they diffuse.

[0021] S40, using a pre-trained generative adversarial network to screen multiple pollutant tracing path simulation results to obtain the optimal pollutant tracing path simulation result.

[0022] This step uses a pre-trained generative adversarial network to filter out the optimal pollutant tracing path simulation results that are closest to the actual situation from the above multiple simulation results.

[0023] The generative adversarial network, composed of a path generator and a path discriminator, is trained on historical source tracing case data to differentiate between simulated results and actual conditions. The network analyzes and evaluates multiple pollutant source tracing path simulations, ultimately deriving the optimal pollutant source tracing path simulation. This allows for accurate identification of pollution sources and diffusion paths, improving the effectiveness of pollution control.

[0024] This method integrates multi-source water quality and environmental data, uses the Analytic Hierarchy Process (AHP) to assess hydrological complexity and develop adaptive simulation strategies. It then combines convection-diffusion and degradation modules to generate multiple pathways, and then uses a pre-trained generative adversarial network to select the optimal pathway. This method enables precise and efficient pollution source tracing, providing a reliable basis for pollution control, and significantly improving the targeted and effective nature of pollution control.

[0025] In some embodiments, the use of the analytic hierarchy process to evaluate the water quality monitoring data, geographic information, hydrological information, and known pollution source information to obtain the hydrological complexity of the target sampling area includes: A hierarchical model was constructed to decompose the hydrological complexity assessment target into a criterion layer and an indicator layer. The criterion layer includes the complexity of water flow dynamics, the complexity of topography and geomorphology, the complexity of pollution source distribution, and the spatial heterogeneity of water quality. The indicator layer includes indicators corresponding to the complexity of water flow dynamics, the complexity of topography and geomorphology, the complexity of pollution source distribution, and the spatial heterogeneity of water quality. For each factor in the criterion layer, the 1-9 scaling method is used to compare the factors in the indicator layer in pairs to determine the relative importance and form a judgment matrix. The weight of each indicator in the indicator layer is obtained by solving the maximum eigenvalue and eigenvector of the judgment matrix, and a consistency test is performed. According to the measured or statistical values ​​of each indicator and combined with the weight vector, the weighted summation method is used to calculate the hydrological complexity of the target sampling area.

[0026] When constructing the hierarchical model, the overall assessment objective is first determined to be the hydrological complexity of the target sampling area, which is then broken down into four criterion layers: hydrodynamic complexity, topographic complexity, pollution source distribution complexity, and spatial heterogeneity of water quality.

[0027] For each criterion layer, further determine the specific indicator layer factors: The index layer of water flow dynamic complexity includes: velocity variation coefficient (velocity standard deviation / velocity average), Reynolds number ( ,in is the flow rate, is the hydraulic diameter, is the kinematic viscosity), Froude number ( ,in is the acceleration due to gravity, is water depth); The indicator layers of topographic complexity include: topographic relief (highest elevation - lowest elevation), water system fractal dimension (calculated by box counting method, formula is ,in The side length required to cover the water system is The number of boxes), riverbed roughness (through the Manning formula, the formula is ,in is the hydraulic slope, is the flow velocity, such as the average flow velocity of water); The indicator layers of pollution source distribution complexity include: pollution source density (number of pollution sources / regional area), spatial variation of emission intensity (standard deviation of emission intensity / average emission intensity), number of pollutant types (number of pollutant types detected in the statistical area); The indicator layers of water quality spatial heterogeneity include: coefficient of variation of pollutant concentration (concentration standard deviation / average concentration), dissolved oxygen gradient ( ,in is the difference in dissolved oxygen between adjacent sampling points, is the distance), pH spatial variability ( , is the pH value of the i-th sampling point, is the average pH value of all sampling points, and n is the number of sampling points).

[0028] When constructing the judgment matrix and calculating the weights, for each criterion layer factor, the 1-9 scaling method is used to compare the corresponding indicator layer factors in pairs to form a judgment matrix. For example, for the water flow dynamic complexity criterion layer, its judgment matrix is for:

[0029] Among them, the matrix elements Indicator Relative to the indicator The importance of the matrix is ​​determined by solving the maximum eigenvalue of the matrix. and its corresponding eigenvector, and obtain the weight vector of each indicator in the indicator layer For the above judgment matrix, we can calculate: ,

[0030] Then the consistency test is performed and the consistency index is calculated , average random consistency index (n=3), consistency ratio , indicating that the judgment matrix has satisfactory consistency.

[0031] When calculating the comprehensive hydrological complexity index, the measured values ​​of each indicator are first standardized and mapped to the interval [0,1]. For positive indicators (the larger the value, the more complex it is), the formula is used. For negative indicators (smaller values ​​indicate more complex), use the formula Then, based on the weight vector of each indicator, the weighted summation formula is used to calculate the score of each criterion layer:

[0032] in, For the The weight of the indicator, For the The standardized value of an indicator.

[0033] Finally, the weighted sum of the scores of the four criteria layers is used to obtain the hydrological complexity :

[0034] in, For the The weight of the criterion layer, For the The score of the criterion layer.

[0035] According to the hydrological complexity The size of the target sampling area is divided into four levels of hydrological complexity: simple ( ),medium( ),complex( ), extremely complex ( ).

[0036] In some embodiments, formulating a simulation strategy for a pollution diffusion simulation model based on the hydrological complexity includes: A multi-layer perceptron was used to construct an attention mechanism. The complexity of water flow dynamics, topographic complexity, pollution source distribution complexity, and water quality spatial heterogeneity were input into the attention mechanism, and their initial impact weights on the simulation start time, time step, and spatial grid accuracy were calculated respectively. generating an adjustment coefficient matrix based on the pollution incident urgency parameter, and adjusting the preliminary impact weight to the target impact weight based on the adjustment coefficient matrix; Dynamically adjust simulation strategy parameters based on the target impact weight and the hydrological complexity.

[0037] Due to the essential differences in the impact of different hydrological factors on the simulation strategy parameters, for example, the complexity of water flow dynamics dominates the diffusion speed of pollutants and has a more significant impact on the time step; the complexity of topography determines the diffusion path and is closely related to the accuracy of the spatial grid; the distribution of pollution sources and the heterogeneity of water quality have special impacts on the timing of simulation start and dynamic adjustment of parameters respectively. Traditional fixed weights cannot reflect this difference, while the attention mechanism can automatically assign preliminary weights through the nonlinear learning of the multi-layer perceptron, allowing key factors to obtain a higher proportion, providing a basic basis for subsequent strategy formulation that fits the hydrological characteristics. Therefore, this embodiment further introduces the attention mechanism to formulate the simulation strategy of the pollution diffusion simulation model.

[0038] Specifically, if Figure 3 As shown, the pollution diffusion simulation model includes an attention mechanism, an adjustment coefficient generation module and a correction module.

[0039] A multi-layer perceptron (MLP) attention mechanism was constructed, using standardized values ​​of flow dynamic complexity, topographic complexity, pollution source distribution complexity, and water quality spatial heterogeneity as input feature vectors. Through nonlinear transformation and softmax normalization in a three-layer MLP network, a preliminary influence weight matrix was output. Each row in this matrix corresponds to a preliminary weight for the simulation start time, time step, and spatial grid accuracy, reflecting the fundamental influence of factors at each criterion layer on the simulation parameters. For example, topographic complexity generally has a higher preliminary weight for spatial grid accuracy because the terrain directly determines the fineness of the grid division.

[0040] The preliminary impact weights derived above only reflect the fundamental role of hydrological factors and do not consider real-time pollution trends. However, the urgency of a pollution event (e.g., the extent of pollutant exceedances or a sudden drop in dissolved oxygen) directly alters the priority of each factor. For example, in the case of sudden high-concentration pollution, the influence of pollution source distribution on the timing of the simulation start should be emphasized; in the case of slowly accumulating pollution, the influence of water flow dynamics on the time step should be emphasized. To address this, this embodiment further employs an adjustment coefficient generation module to generate an adjustment coefficient matrix based on the pollution event urgency parameter and uses this adjustment coefficient matrix to modify the preliminary impact weights derived above.

[0041] Specifically, the pollution incident urgency parameter is calculated based on the magnitude of the pollutant concentration exceeding the standard, the drop in dissolved oxygen, and abnormal changes in water temperature. A higher urgency indicates a greater risk of further spread of the pollution. It is understood that this pollution source analysis can be conducted in response to the pollution incident. This disclosure does not specifically limit the specific calculation formula or mapping table.

[0042] The adjustment coefficient matrix is ​​dynamically generated based on the pollution event urgency parameter. For example, when a sudden high-concentration pollution event occurs, the adjustment coefficient of the pollution source distribution complexity on the simulation start timing is increased to strengthen its weight. The specific adjustment method can be referred to the example shown in Table 1 below: Table 1

[0043] After multiplying the preliminary impact weights and the adjustment coefficient matrix element by element and normalizing them, the target impact weights are obtained, so that the weight distribution is more in line with the real-time pollution situation.

[0044] Therefore, when the correction module dynamically adjusts simulation strategy parameters based on target impact weights and hydrological complexity, the target impact weights determine the priority of each criterion layer's parameter adjustments, while hydrological complexity determines the magnitude of parameter adjustments. For example, if topographic complexity has a high target weight for spatial grid accuracy and hydrological complexity is at the complex level, the spatial grid accuracy is refined to 5 meters; if hydrodynamic complexity has a high target weight for time step and hydrological complexity is high, the time step is shortened to 10 minutes, thereby accurately adapting the simulation strategy to actual hydrological and pollution conditions.

[0045] In some embodiments, the simulation is performed according to the simulation strategy to obtain multiple pollutant tracing path simulation results, including: According to the simulation start timing, time step and spatial grid accuracy in the simulation strategy, the geographic information, hydrological information and known pollution source information of the target sampling area are loaded into the pollution diffusion simulation model; Construct multiple simulation scenarios, including baseline scenarios, sensitive scenarios, and extreme scenarios. Under each scenario, the convection-diffusion equation is used to calculate the migration process of pollutants within the grid cells. At the same time, the pollutant degradation dynamics module is called to dynamically adjust the pollutant degradation coefficient according to the pollutant type. The pollutant concentration distribution of each grid cell is output at each time step, and the simulation results of the corresponding scenario are constructed based on this. Based on the simulation results of each scenario, a reverse tracing algorithm is used. Starting from the pollutant concentration monitoring values ​​of each sampling point, the pollutant diffusion trajectory is reversed along the time axis, marking the upstream grid unit with the largest concentration contribution in each time step, forming a tracing path including candidate pollution source points, diffusion path nodes and confidence levels, and outputting several differentiated pollutant tracing path simulation results.

[0046] When loading data according to the simulation strategy, the geographic information of the target sampling area (such as terrain undulation, river direction), hydrological information (such as flow velocity, spatiotemporal distribution of flow direction) and known pollution source information (such as location, emission volume, and pollutant type) must be imported into the pollution diffusion simulation model according to the determined simulation start time, time step, and spatial grid accuracy.

[0047] Next, multiple simulation scenarios are constructed and pollutant concentration distributions are calculated to cover the possibility of pollution diffusion under different environmental conditions. Benchmark scenario: Directly adopt the parameters determined by the simulation strategy, including the simulation start time (such as the first time an anomaly is detected), time step (such as 10 minutes), spatial grid accuracy (such as 5 meters), input basic hydrological data of the target area (flow velocity 2m / s, flow direction due east) and known pollution source parameters (such as chemical company A's emission of 50kg / h), without introducing additional disturbance factors, to reflect the diffusion process under normal circumstances.

[0048] Sensitive scenarios: Parameter perturbations are performed on the criterion layer factors with the highest target impact weights. For example, if topographic complexity has the highest target weight for spatial grid accuracy (0.6), the grid accuracy is adjusted within ±20% of the baseline value (i.e., between 4 and 6 meters). If hydrodynamic complexity has the highest weight, the flow velocity is adjusted within ±15% of the baseline value (between 1.7 and 2.3 m / s), while all other parameters remain consistent with the baseline scenario.

[0049] Extreme scenarios: Two types of extreme conditions are set: ① Sudden hydrological changes, such as simulating heavy rain that causes the flow velocity to suddenly increase to twice the baseline value (4m / s), while adding the flow impact of tributaries (an additional 1m³ / s); ② Activation of potential pollution sources. Based on the historical distribution of pollution points, a virtual pollution source (emissions of 30kg / h) is added 1km upstream that is not covered by sensitive scenarios to simulate the impact of unknown pollution sources.

[0050] In each of the above scenarios, the convection-diffusion equation is used to calculate pollutant migration. At the same time, the degradation kinetics module dynamically adjusts the degradation coefficient according to the characteristics of the pollutants (such as organic matter is easily degradable and heavy metals are difficult to degrade), ensuring that the concentration distribution calculation at each time step is accurate. The details are as follows: The finite difference method is used to solve the three-dimensional convection-diffusion equation:

[0051] in, 、 、 for 、 、 Directional diffusion coefficient (valued according to the hydrological complexity level, 0.8m² / s for complex scenarios and 0.3m² / s for simple scenarios), 、 、 is the three-dimensional velocity component, is the degradation coefficient. At each time step (e.g., 10 minutes), the concentration change is calculated within each grid cell. For example, if the initial concentration in a grid cell is 10 mg / L, after convection and diffusion, it is affected by the high-concentration grid cell to the east (15 mg / L), and the concentration is updated to 12 mg / L.

[0052] For easily degradable organic matter (such as BOD), The value is dynamically adjusted according to the water temperature. The formula is:

[0053] in, , T is water temperature, when the water temperature is 25℃ Increased to 0.25d⁻¹; for heavy metals (such as Pb) (difficult to degrade) and is not affected by water temperature; updated every 3 time steps values ​​to ensure that concentration calculations reflect the natural attenuation characteristics of pollutants.

[0054] After each scenario is completed, the spatial concentration distribution matrix of each time node (such as the concentration value of a 100×100 grid) is output, and a concentration contour map is generated, marking the areas that exceed the environmental standard value (such as 5 mg / L).

[0055] Finally, a reverse tracing algorithm is used to generate a traceability path, that is, to reversely deduce the pollution source from the measured concentration at the sampling point. By tracing back along the time axis, the upstream grid that contributes most to the downstream concentration within each step is marked to lock the propagation trajectory of the pollutant. Each path contains a candidate pollution source point (potential source location), a path node (key diffusion location and concentration attenuation), and a confidence level (the degree of consistency between the simulated concentration and the measured value). The multiple differentiated paths finally output can fully reflect the traceability possibilities under different scenarios. The details are as follows: (1) Initialize the reverse tracking starting point: Taking the pollutant concentration monitoring value of each sampling point as the starting point, determine the initial conditions for reverse tracing: time starting point: the moment when the pollutant concentration is monitored at the sampling point (such as t=24h); spatial starting point: the grid unit corresponding to the sampling point (such as grid coordinates (50,30)); initial concentration: the measured concentration value at the sampling point (such as 8mg / L).

[0056] (2) Deducing backward along the time axis: Based on the forward simulation results of the pollution diffusion simulation model (concentration distribution at each time step), reverse calculation is performed according to the time step: Go back from the current time step (e.g., t=24h) to the previous time step (t=23h50min, assuming a 10-minute step). For the current grid cell, calculate the upstream grid cells that may have contributed to its concentration in the previous time step by inverse calculation of the convection-diffusion equation (e.g., based on the flow direction, lock the three adjacent grids on the west side). Based on the principle of concentration conservation, calculate the concentration contribution rate of each upstream grid cell to the current grid (e.g., the contribution rate of grid (49,30) is 60%, and the contribution rate of (50,29) is 30%).

[0057] (3) Mark the key path nodes: During the reverse tracking process, key information within each time step is marked: Path node: Select the upstream grid cell with the highest concentration contribution rate as the next reverse tracking node (e.g., prioritize tracking the grid (49,30) with a contribution rate of 60%). Concentration decay rate: Calculate the concentration ratio of adjacent nodes (e.g., the concentration of grid (49,30) is 10 mg / L at t=23h50min, compared to 8 mg / L at t=24h, the decay rate is 20%). Candidate pollution source: When tracing back to a grid cell close to a known pollution source (e.g., within 100 meters of the pollution source), mark the pollution source as a candidate source and record its concentration match with the path node (e.g., the deviation between the simulated concentration and the pollution source emission concentration is <10%).

[0058] (4) Calculate path confidence: After each reverse tracing (e.g., 10 time steps back), the path reliability is evaluated using the following methods: Compare the reverse-calculated concentration to the forward-simulated concentration (e.g., the reverse-calculated concentration at t = 23h is 10 mg / L, while the forward-simulated result is 9.5 mg / L, with a 5% deviation). Confidence = 1 - |reverse concentration - forward concentration| / forward concentration. For example, in the above example, confidence = 95%. If the confidence falls below a threshold (e.g., 70%), adjust the tracking path and reselect the upstream contributing grid.

[0059] Backward tracing stops when any of the following conditions are met: tracing back to the start of the simulation (e.g., t=0); tracing to a known pollution source with a confidence level of ≥90%; or tracing to a confidence level of <50% for three consecutive time steps (path interruption). The final output includes the candidate pollution source, the complete diffusion path nodes (including time, coordinates, and concentration), and the average confidence level for the entire process.

[0060] In some embodiments, the use of a pre-trained generative adversarial network to screen multiple pollutant source tracing path simulation results to obtain an optimal pollutant source tracing path simulation result includes: The simulation results of each pollutant traceability path are converted into a feature matrix in a unified format. The feature matrix is ​​input into the pre-trained path discriminator to obtain the authenticity score of each path. The path with the highest authenticity score is selected as the optimal result. Furthermore, if there are multiple paths with the same score, they are further screened according to the following rules: paths that contain known pollution sources and have a matching degree higher than the preset matching degree are given priority; if all paths contain known pollution sources, the path with the least number of nodes is selected.

[0061] The simulation results of each pollutant source traceability path are converted into a unified feature matrix to eliminate interference caused by differences in path data formats across different scenarios. The feature matrix integrates core path information, including spatial coordinates (such as longitude and latitude), pollutant concentration, confidence level, and corresponding timestamp at each time step, enabling the path discriminator to perform consistency assessments based on standardized data.

[0062] For example, a path is output in the form of "grid coordinates + concentration" in the baseline scenario and in the form of "latitude and longitude + concentration" in the extreme scenario. Through feature matrix conversion, it can be unified into a matrix structure of "spatial coordinates (normalized to the range of 0-1) + concentration (normalized to the range of 0-1) + time step number" to ensure that the input format of the path discriminator is consistent.

[0063] The feature matrix is ​​fed into a pre-trained path discriminator to obtain a plausibility score. This uses the real-world traceability patterns learned by the generative adversarial network during training (such as upstream diffusion of pollutants and concentration decay with distance) to assess the plausibility of the path. The pre-trained path discriminator has learned key features from a large number of historical real-world and invalid paths. For example, it can identify paths that do not conform to hydrological patterns, such as upstream diffusion and no concentration decay, and assign them low scores. It also assigns high scores to paths that conform to the velocity-concentration decay relationship and closely match the measured data at the sampling points. The plausibility score comprehensively reflects the degree to which the path conforms to the real-world pollution diffusion patterns.

[0064] The path with the highest authenticity score is selected as the optimal result. For example, if a path's authenticity score is 0.92 (out of a maximum score of 1), significantly higher than the 0.75 of other paths, it indicates that it performs best in terms of spatial rationality (such as downstream direction) and temporal consistency (such as matching concentration changes with water flow velocity), and can be directly determined as the optimal path.

[0065] In addition, when multiple paths with the same score exist, further rule-based screening is used to select the result that best matches the actual pollution scenario from these equally reasonable paths. Paths that include known pollution sources and have a matching degree higher than a preset matching degree (e.g., 90%) are preferred because known pollution sources (e.g., factory sewage outlets) are high-incidence sources of pollution. A high matching degree indicates that the path is more consistent with the emission characteristics (e.g., emission volume, pollutant type) of the known pollution source, and thus has higher traceability reliability. If both paths include known pollution sources, the path with the fewest nodes is selected. This is based on the principle of simplicity: a path with fewer nodes means a more direct diffusion process, less interference from complex hydrological processes, and a clearer location of the corresponding pollution source. For example, if two paths both lead to a chemical plant, one with 10 nodes and the other with 20 nodes, the former will be preferred due to its fewer nodes and clearer logic.

[0066] The pre-training process of the generative adversarial network is roughly as follows: the training dataset consists of the real traceability paths of historical pollution events (including the pollution source location and diffusion trajectory verified by monitoring data) and artificially generated invalid paths (such as upstream paths that violate hydrological laws); the path generator generates pseudo-real paths that conform to hydrological diffusion laws and trains the path discriminator together with the real paths, so that the path discriminator learns the characteristic patterns of the real traceability paths (such as downstream diffusion, and the positive correlation between the concentration decay rate and the flow velocity); the training termination condition is that the path discriminator's recognition accuracy of the real path is stable at above 95%, and the average characteristic deviation between the path generated by the path generator and the real path is ≤5%.

[0067] like Figure 4 As shown, the embodiment of the present invention also discloses a pollution source tracing analysis system 100 based on water quality sampling data, which is applied to a data processing center and includes an acquisition unit 101, a simulation strategy formulation unit 102, a simulation unit 103, and a path screening unit 104; The acquisition unit 101 receives a plurality of water quality monitoring data corresponding to a target sampling area, each of which includes geographic information of a sampling point and water quality monitoring indicators; and acquires geographic information, hydrological information, and known pollution source information of the target sampling area; The simulation strategy formulation unit 102 uses the analytic hierarchy process to evaluate the water quality monitoring data, geographic information, hydrological information, and known pollution source information to obtain the hydrological complexity of the target sampling area, and formulates a simulation strategy for the pollution diffusion simulation model based on the hydrological complexity; the simulation strategy includes the simulation start time, time step, and spatial grid accuracy; The simulation unit 103 inputs the water quality monitoring data, geographic information, hydrological information, and known pollution source information into a pollution diffusion simulation model, performs simulation according to the simulation strategy, and obtains multiple pollutant tracing path simulation results; the pollution diffusion simulation model has built-in convection-diffusion equations and pollutant degradation kinetics modules; The path screening unit 104 uses a pre-trained generative adversarial network to screen multiple pollutant tracing path simulation results to obtain the optimal pollutant tracing path simulation result.

[0068] In some embodiments, the simulation strategy formulation unit 102 is configured to: A hierarchical model was constructed to decompose the hydrological complexity assessment target into a criterion layer and an indicator layer. The criterion layer includes the complexity of water flow dynamics, the complexity of topography and geomorphology, the complexity of pollution source distribution, and the spatial heterogeneity of water quality. The indicator layer includes indicators corresponding to the complexity of water flow dynamics, the complexity of topography and geomorphology, the complexity of pollution source distribution, and the spatial heterogeneity of water quality. For each factor in the criterion layer, the 1-9 scaling method is used to compare the factors in the indicator layer in pairs to determine the relative importance and form a judgment matrix. The weight of each indicator in the indicator layer is obtained by solving the maximum eigenvalue and eigenvector of the judgment matrix, and a consistency test is performed. According to the measured or statistical values ​​of each indicator and combined with the weight vector, the weighted summation method is used to calculate the hydrological complexity of the target sampling area.

[0069] In some embodiments, the simulation strategy formulation unit 102 is configured to: A multi-layer perceptron was used to construct an attention mechanism. The complexity of water flow dynamics, topographic complexity, pollution source distribution complexity, and water quality spatial heterogeneity were input into the attention mechanism, and their initial impact weights on the simulation start time, time step, and spatial grid accuracy were calculated respectively. generating an adjustment coefficient matrix based on the pollution incident urgency parameter, and adjusting the preliminary impact weight to the target impact weight based on the adjustment coefficient matrix; Dynamically adjust simulation strategy parameters based on the target impact weight and the hydrological complexity.

[0070] In some embodiments, the simulation unit 103 is configured to: According to the simulation start timing, time step and spatial grid accuracy in the simulation strategy, the geographic information, hydrological information and known pollution source information of the target sampling area are loaded into the pollution diffusion simulation model; Construct multiple simulation scenarios, including baseline scenarios, sensitive scenarios, and extreme scenarios. Under each scenario, the convection-diffusion equation is used to calculate the migration process of pollutants within the grid cells. At the same time, the pollutant degradation dynamics module is called to dynamically adjust the pollutant degradation coefficient according to the pollutant type. The pollutant concentration distribution of each grid cell is output at each time step, and the simulation results of the corresponding scenario are constructed based on this. Based on the simulation results of each scenario, a reverse tracing algorithm is used. Starting from the pollutant concentration monitoring values ​​of each sampling point, the pollutant diffusion trajectory is reversed along the time axis, marking the upstream grid unit with the largest concentration contribution in each time step, forming a tracing path including candidate pollution source points, diffusion path nodes and confidence levels, and outputting several differentiated pollutant tracing path simulation results.

[0071] In some embodiments, the path screening unit 104 is configured to: The simulation results of each pollutant traceability path are converted into a feature matrix in a unified format. The feature matrix is ​​input into the pre-trained path discriminator to obtain the authenticity score of each path. The path with the highest authenticity score is selected as the optimal result. Furthermore, if there are multiple paths with the same score, they are further screened according to the following rules: paths that contain known pollution sources and have a matching degree higher than the preset matching degree are given priority; if all paths contain known pollution sources, the path with the least number of nodes is selected.

[0072] An embodiment of the present invention further discloses an electronic device, comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the computer program implements any of the above methods when executed by the processor.

[0073] An embodiment of the present invention further discloses a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method described in any of the above items is executed.

[0074] An embodiment of the present invention further discloses a computer program product, which includes a computer program stored in a computer storage medium. When the computer program is executed by a processor of an electronic device, it implements any of the above methods.

[0075] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A pollution source tracing analysis method based on water quality sampling data, characterized in that: The steps include: Receive multiple pieces of water quality monitoring data corresponding to the target sampling area, each piece of water quality monitoring data including geographic information of the sampling point and water quality monitoring indicators; and obtain geographic information, hydrological information, and known pollution source information of the target sampling area; Using the analytic hierarchy process to evaluate the water quality monitoring data, geographic information, hydrological information, and known pollution source information to obtain the hydrological complexity of the target sampling area, and formulating a simulation strategy for the pollution diffusion simulation model based on the hydrological complexity; the simulation strategy includes the simulation start time, time step, and spatial grid accuracy; Inputting the water quality monitoring data, geographic information, hydrological information and known pollution source information into the pollution diffusion simulation model, performing simulation according to the simulation strategy, and obtaining multiple pollutant tracing path simulation results; the pollution diffusion simulation model has built-in convection-diffusion equations and pollutant degradation kinetics modules; Use a pre-trained generative adversarial network to screen multiple pollutant tracing path simulation results to obtain the optimal pollutant tracing path simulation result.

2. The pollution source tracing analysis method based on water quality sampling data according to claim 1 is characterized by: The water quality monitoring data, geographic information, hydrological information and known pollution source information are evaluated using the analytic hierarchy process to obtain the hydrological complexity of the target sampling area, including: A hierarchical model was constructed to decompose the hydrological complexity assessment target into a criterion layer and an indicator layer. The criterion layer includes the complexity of water flow dynamics, the complexity of topography and geomorphology, the complexity of pollution source distribution, and the spatial heterogeneity of water quality. The indicator layer includes indicators corresponding to the complexity of water flow dynamics, the complexity of topography and geomorphology, the complexity of pollution source distribution, and the spatial heterogeneity of water quality. For each factor in the criterion layer, the 1-9 scaling method is used to compare the factors in the indicator layer in pairs to determine the relative importance and form a judgment matrix. The weight of each indicator in the indicator layer is obtained by solving the maximum eigenvalue and eigenvector of the judgment matrix, and a consistency test is performed. According to the measured or statistical values ​​of each indicator and combined with the weight vector, the weighted summation method is used to calculate the hydrological complexity of the target sampling area.

3. The pollution source tracing analysis method based on water quality sampling data according to claim 2 is characterized by: A simulation strategy for a pollution diffusion simulation model is formulated based on the hydrological complexity, including: A multi-layer perceptron was used to construct an attention mechanism. The complexity of water flow dynamics, topographic complexity, pollution source distribution complexity, and water quality spatial heterogeneity were input into the attention mechanism, and their initial impact weights on the simulation start time, time step, and spatial grid accuracy were calculated respectively. generating an adjustment coefficient matrix based on the pollution incident urgency parameter, and adjusting the preliminary impact weight to the target impact weight based on the adjustment coefficient matrix; Dynamically adjust simulation strategy parameters based on the target impact weight and the hydrological complexity.

4. The pollution source tracing analysis method based on water quality sampling data according to claim 3 is characterized by: According to the simulation strategy, multiple pollutant source traceability path simulation results were obtained, including: According to the simulation start timing, time step and spatial grid accuracy in the simulation strategy, the geographic information, hydrological information and known pollution source information of the target sampling area are loaded into the pollution diffusion simulation model; Construct multiple simulation scenarios, including baseline scenarios, sensitive scenarios, and extreme scenarios. Under each scenario, the convection-diffusion equation is used to calculate the migration process of pollutants within the grid cells. At the same time, the pollutant degradation dynamics module is called to dynamically adjust the pollutant degradation coefficient according to the pollutant type. The pollutant concentration distribution of each grid cell is output at each time step, and the simulation results of the corresponding scenario are constructed based on this. Based on the simulation results of each scenario, a reverse tracing algorithm is used. Starting from the pollutant concentration monitoring values ​​of each sampling point, the pollutant diffusion trajectory is reversed along the time axis, marking the upstream grid unit with the largest concentration contribution in each time step, forming a tracing path including candidate pollution source points, diffusion path nodes and confidence levels, and outputting several differentiated pollutant tracing path simulation results.

5. The pollution source tracing analysis method based on water quality sampling data according to claim 1 is characterized in that: Use a pre-trained generative adversarial network to screen multiple pollutant source path simulation results to obtain the optimal pollutant source path simulation results, including: The simulation results of each pollutant traceability path are converted into a feature matrix in a unified format. The feature matrix is ​​input into the pre-trained path discriminator to obtain the authenticity score of each path. The path with the highest authenticity score is selected as the optimal result. Furthermore, if there are multiple paths with the same score, they are further screened according to the following rules: paths that contain known pollution sources and have a matching degree higher than the preset matching degree are given priority; if all paths contain known pollution sources, the path with the least number of nodes is selected.

6. A pollution source tracing and analysis system based on water quality sampling data, applied to a data processing center, characterized by: It includes an acquisition unit, a simulation strategy formulation unit, a simulation unit, and a path screening unit; The acquisition unit receives a plurality of water quality monitoring data corresponding to the target sampling area, each of which includes geographic information of the sampling point and water quality monitoring indicators; and acquires geographic information, hydrological information and known pollution source information of the target sampling area; The simulation strategy formulation unit uses a hierarchical analysis method to evaluate the water quality monitoring data, geographic information, hydrological information, and known pollution source information to obtain the hydrological complexity of the target sampling area, and formulates a simulation strategy for the pollution diffusion simulation model based on the hydrological complexity; the simulation strategy includes a simulation start time, a time step, and a spatial grid accuracy; The simulation unit inputs the water quality monitoring data, geographic information, hydrological information, and known pollution source information into a pollution diffusion simulation model, performs simulation according to the simulation strategy, and obtains multiple pollutant tracing path simulation results; the pollution diffusion simulation model has a built-in convection-diffusion equation and a pollutant degradation kinetics module; The path screening unit uses a pre-trained generative adversarial network to screen multiple pollutant tracing path simulation results to obtain the optimal pollutant tracing path simulation result.

7. A pollution source tracing analysis system based on water quality sampling data according to claim 6, characterized in that: The simulation strategy formulation unit is used to: A hierarchical model was constructed to decompose the hydrological complexity assessment target into a criterion layer and an indicator layer. The criterion layer includes the complexity of water flow dynamics, the complexity of topography and geomorphology, the complexity of pollution source distribution, and the spatial heterogeneity of water quality. The indicator layer includes indicators corresponding to the complexity of water flow dynamics, the complexity of topography and geomorphology, the complexity of pollution source distribution, and the spatial heterogeneity of water quality. For each factor in the criterion layer, the 1-9 scaling method is used to compare the factors in the indicator layer in pairs to determine the relative importance and form a judgment matrix. The weight of each indicator in the indicator layer is obtained by solving the maximum eigenvalue and eigenvector of the judgment matrix, and a consistency test is performed. According to the measured or statistical values ​​of each indicator and combined with the weight vector, the weighted summation method is used to calculate the hydrological complexity of the target sampling area.

8. The pollution source tracing and analysis system based on water quality sampling data according to claim 7 is characterized by: The simulation strategy formulation unit is used to: A multi-layer perceptron was used to construct an attention mechanism. The complexity of water flow dynamics, topographic complexity, pollution source distribution complexity, and water quality spatial heterogeneity were input into the attention mechanism, and their initial impact weights on the simulation start time, time step, and spatial grid accuracy were calculated respectively. generating an adjustment coefficient matrix based on the pollution incident urgency parameter, and adjusting the preliminary impact weight to the target impact weight based on the adjustment coefficient matrix; Dynamically adjust simulation strategy parameters based on the target impact weight and the hydrological complexity.

9. A pollution source tracing analysis system based on water quality sampling data according to claim 8, characterized in that: The simulation unit is used to: According to the simulation start timing, time step and spatial grid accuracy in the simulation strategy, the geographic information, hydrological information and known pollution source information of the target sampling area are loaded into the pollution diffusion simulation model; Construct multiple simulation scenarios, including baseline scenarios, sensitive scenarios, and extreme scenarios. Under each scenario, the convection-diffusion equation is used to calculate the migration process of pollutants within the grid cells. At the same time, the pollutant degradation dynamics module is called to dynamically adjust the pollutant degradation coefficient according to the pollutant type. The pollutant concentration distribution of each grid cell is output at each time step, and the simulation results of the corresponding scenario are constructed based on this. Based on the simulation results of each scenario, a reverse tracing algorithm is used. Starting from the pollutant concentration monitoring values ​​of each sampling point, the pollutant diffusion trajectory is reversed along the time axis, marking the upstream grid unit with the largest concentration contribution in each time step, forming a tracing path including candidate pollution source points, diffusion path nodes and confidence levels, and outputting several differentiated pollutant tracing path simulation results.

10. The pollution source tracing analysis system based on water quality sampling data according to claim 6 is characterized in that: The path screening unit is used to: The simulation results of each pollutant traceability path are converted into a feature matrix in a unified format. The feature matrix is ​​input into the pre-trained path discriminator to obtain the authenticity score of each path. The path with the highest authenticity score is selected as the optimal result. Furthermore, if there are multiple paths with the same score, they are further screened according to the following rules: paths that contain known pollution sources and have a matching degree higher than the preset matching degree are given priority; if all paths contain known pollution sources, the path with the least number of nodes is selected.

Images