Water pollution tracing method based on data reconstruction mode
By combining data reconstruction models and artificial intelligence algorithms with hydrodynamic-water quality models, the problem of pollution source tracing under conditions of sparse monitoring points and limited data quality was solved, enabling the simultaneous identification of pollution sources and diffusion coefficients, and improving the accuracy and applicability of source tracing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-11
- Publication Date
- 2026-03-20
AI Technical Summary
Existing pollution source tracing methods based on numerical models struggle to accurately identify pollution sources and diffusion coefficients when there are sparse monitoring stations and limited data quality, resulting in large errors in the tracing results and limited applicability.
By employing a data reconstruction model combined with artificial intelligence algorithms, a pollutant database is constructed, water quality monitoring equipment is deployed, the observation status is reconstructed, and a hydrodynamic-water quality forward model is built. Pollution sources and diffusion coefficients are identified simultaneously, and hydrodynamic conditions are calculated using the ELCIRC model and irregular grids. Parameters are then optimized using genetic algorithms or ensemble Kalman filtering methods.
Despite the limited number of monitoring points and the limited quality of data, this method enables rapid and effective source tracing of pollutants in water bodies, identifies pollution sources and diffusion coefficients, and improves the accuracy and applicability of source tracing.
Smart Images
Figure CN121709086A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of water environment monitoring and intelligent analysis technology, and in particular to a water pollution source tracing method based on data reconstruction patterns. Background Technology
[0002] Numerical model-based pollution source tracing methods have attracted attention due to their rapid response. These methods typically rely on monitoring data from a limited number of stations within a watershed, analyzing spatiotemporal variations in pollutant concentrations to infer the location of pollution sources. However, these methods still have significant limitations. First, the sparse spatial distribution of monitoring stations results in discrete data, making it difficult to comprehensively and continuously reflect the overall water quality situation, leading to biases in judging the migration path and diffusion range of pollution plumes. Second, observational data are prone to error fluctuations due to factors such as meteorological conditions, equipment performance, and human operation, making source tracing results highly sensitive to data gaps and noise interference. Furthermore, some hydrodynamic and water quality process parameters involved in the model are often simplified to assumed prior knowledge due to a lack of a priori information, limiting the applicability of this assumption in complex real-world environments. The ability to efficiently identify pollution sources in conjunction with hydrological and water quality parameters and pollution source information remains insufficient, hindering improvements in source tracing accuracy and model practicality. Summary of the Invention
[0003] Purpose of the invention: To propose a water pollution source tracing method based on data reconstruction model, which enables rapid and effective source tracing of water pollutants when observation points are sparse and data quality is limited, while simultaneously identifying the diffusion coefficient in the material transport model.
[0004] This invention proposes a water pollution source tracing method based on data reconstruction patterns, the steps of which are as follows: Investigate the distribution of potential pollution sources in the target area, construct a pollutant database, and determine pollution monitoring indicators for the target area based on the pollutant database; Based on the pollution monitoring indicators of the target area, deploy several water quality monitoring devices; Water quality monitoring equipment continuously collects pollutant concentration data to construct a relative pollution monitoring time series. The observation status is reconstructed based on the relative pollution monitoring time series, and the reconstructed data is embedded into an artificial intelligence algorithm. Collect topographic, meteorological, and hydrological data of the target area, construct a hydrodynamic-water quality forward model, and combine it with the aforementioned artificial intelligence algorithm to construct a water pollutant source tracing model; The water quality monitoring equipment collects pollutant monitoring concentration data in real time and updates the relative pollution monitoring time series. The updated relative pollution monitoring time series is then input into the water body pollutant source tracing model to update the pollution source location, emission time, and diffusion coefficient.
[0005] As a preferred option, the distribution of potential pollution sources in the target area includes industrial parks, residential areas, and cargo ships within a predetermined range around the target area. Pollution source characteristics are obtained from these sources to construct a pollutant database. The pollution source characteristics are described using three types of parameters: source location, time, and release amount.
[0006] As a preferred option, based on the pollution monitoring indicators of the target area, several water quality monitoring devices are deployed to ensure that the coverage of each water quality monitoring device overlaps, thus forming a water quality monitoring network.
[0007] As a preferred approach, water quality monitoring equipment collects pollutant concentration data from past periods to construct a relative pollution monitoring time series, which is expressed as follows:
[0008] In the formula, Let be the relative concentration at the i-th observation time step; t is the time length of the concentration time series. It is the directly observed or simulated concentration at the i-th observation time step.
[0009] As a preferred approach, the observation state is reconstructed based on the relative pollution monitoring time series, resulting in a new state quantity that highlights different characteristics of the data series. The new state quantity is calculated as follows:
[0010]
[0011]
[0012] In the formula, Let be the relative concentration at the i-th observation time step; The concentration directly observed in step i; This represents the average concentration observed from step 1 to step t. Let be the simulated concentration in step i; This represents the average concentration from step 1 to step t. and These represent the relative concentrations observed and simulated in step i, respectively; , , These are the correlation coefficient, the sum of squares of relative deviations of relative concentrations, and the mean of relative concentrations at step t, respectively.
[0013] As a preferred approach, the new state variables are used as the basis for embedding artificial intelligence algorithms, including differential evolution algorithms or ensemble Kalman filter assimilation methods.
[0014] As a preferred embodiment, the hydrodynamic-water quality forward model includes a hydrodynamic model and a water quality model, wherein: The hydrodynamic model adopts the ELCIRC open-source model, and the computational grid adopts an irregular grid. The regional hydrodynamics are calculated as the dynamic conditions for material transport in the water quality model. The water quality model is constructed based on a two-dimensional convection-diffusion equation. An irregular grid is used for computation, and the finite volume method is employed for differential solution to calculate the material concentration process at all grid points.
[0015] As a preferred embodiment, a water pollutant source tracing model is constructed based on the aforementioned hydrodynamic-water quality forward model and combined with the reconstructed data, specifically including: An objective function is constructed, and the three types of pollution source parameters and the horizontal and vertical diffusion coefficients are simultaneously optimized by calculating the extreme value of the objective function. The extreme value of the objective function is related to the substance concentration simulated by the water quality model under different combinations of pollution source parameters.
[0016] Beneficial Effects: Compared with existing technologies, this invention proposes a water pollution source tracing method based on data reconstruction, incorporating intelligent optimization methods and the physical processes of pollutant transport using water quality monitoring IoT data. This method is applicable to tracing water pollution caused by point sources or continuous discharge sources, and simultaneously identifies unknown diffusion coefficients in the material transport model. This method can achieve simultaneous identification of multiple unknown parameters composed of pollution sources and model diffusion coefficients even when observation points are sparse and data quality is limited. It can solve the problem of water pollution source tracing under poor monitoring conditions. Furthermore, the proposed data reconstruction technology is adaptable to various intelligent optimization methods, exhibiting broad applicability. Attached Figure Description
[0017] Figure 1 This is a flowchart illustrating the workflow of the water pollution source tracing method based on data reconstruction model of the present invention.
[0018] Figure 2 The map shows the scope of the study and the distribution of pollution release points and water quality monitoring points in the examples.
[0019] Figure 3 The results of simultaneous identification of pollution sources and substance diffusion coefficients are shown in the example. Detailed Implementation
[0020] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.
[0021] This embodiment discloses a water pollution source tracing method based on data reconstruction patterns, and the specific implementation steps are as follows: 1) Investigate the distribution of potential pollution sources and construct a pollutant database for the study area; 2) Based on the characteristics of the study area and water pollution monitoring indicators, rationally deploy the online monitoring equipment network; 3) Construct a hydrodynamic-water quality forward model based on multi-source data such as topography, meteorology, and hydrology, and combine it with artificial intelligence optimization methods to construct a water pollutant source tracing model; 4) Construct relative concentration data sequences based on pollutant monitoring concentration time series data; 5) Reconstruct the observed state based on the relative concentration data sequence, and embed the data reconstruction pattern into an artificial intelligence optimization method; 6) Update the observation time series and iteratively optimize the pollution source parameter group based on the real-time monitoring data transmitted by the monitoring equipment network.
[0022] Furthermore, in the water pollution source tracing optimization method based on data reconstruction model, the pollution source characteristics are described by source location, time, and release amount. Pollution source tracing is to identify these three types of parameters of pollution source through monitoring information.
[0023] Furthermore, in step 1), the investigation of the distribution of potential pollution sources includes studying industrial parks, residential areas, and cargo ships around the water area that may cause water pollution incidents, and statistically analyzing the types and distribution of potential pollutants to construct a pollutant database.
[0024] Furthermore, in step 2), the online monitoring equipment needs to be determined based on the actual water pollution monitoring indicators. When deploying the equipment, it is necessary to take into account the regional terrain characteristics to ensure that the monitoring network has equipment capable of acquiring pollution concentration processes under all possible water pollution events. In addition, the sampling time interval of the online monitoring equipment should be set to 30 seconds or less as much as possible to ensure the comprehensiveness and completeness of data collection.
[0025] Further, in step 3), the hydrodynamic-water quality forward model is constructed based on the topographic, meteorological, and hydrological data of the study area, and is a two-dimensional planar model. The hydrodynamic model uses the ELCIRC open-source model, with an irregular grid for computation, calculating the regional hydrodynamics as the dynamic conditions for material transport in the water quality model. The water quality model is constructed based on two-dimensional convection-diffusion equations, using an irregular grid for computation, and employing the finite volume method for differential solution to calculate the material concentration process at all grid points.
[0026] Furthermore, in step 3), the construction of a water pollutant source tracing model using artificial intelligence optimization methods specifically refers to the construction of the water pollutant source tracing model based on artificial intelligence optimization methods and a water quality model. In this model, the lateral and longitudinal diffusion coefficients directly affect the pollutant concentration distribution, and these coefficients are typically impossible to determine in advance. An objective function is constructed using the artificial intelligence optimization method, and the three types of pollution source parameters and the lateral and longitudinal diffusion coefficients are simultaneously optimized through objective function extremum calculation. The extremum of the objective function is related to the substance concentration simulated by the water quality model under different combinations of pollution source parameters.
[0027] Further, in step 4), the relative concentration is constructed as shown in equation (1), and the constructed relative concentration... It is a dimensionless number and is independent of the mass released from the source, thus decoupling the amount of pollution released from the source from the parameter set consisting of three types of parameters of the pollution source.
[0028]
[0029] In the formula: the subscript i represents the observation time step; Let be the relative concentration at the i-th observation time step; t is the time length of the concentration time series. It is the directly observed or simulated concentration at the i-th observation time step.
[0030] Furthermore, in step 5), the observation state reconstruction is based on the analysis of the relative concentration data sequence, deriving new state quantities that can highlight different characteristics of the data sequence. The calculation of the new state quantities is as shown in equations (2a) to (2c):
[0031]
[0032]
[0033] In the formula: C represents the pollutant concentration; the superscripts s and ob represent the simulated concentration and observed concentration, respectively; the superscript ' represents the relative concentration; The concentration directly observed in step i; This represents the average concentration observed from step 1 to step t. Let be the simulated concentration in step i; This represents the average concentration from step 1 to step t. and These are the relative concentrations observed and simulated in step i, respectively. , , These are the correlation coefficient, the sum of squares of relative deviations of relative concentrations, and the mean of relative concentrations at step t, respectively.
[0034] In addition, in step 5), the data reconstruction mode is applicable to different artificial intelligence optimization methods, such as genetic algorithms, differential evolution algorithms, ensemble Kalman filter assimilation methods, etc.
[0035] Taking the differential evolution algorithm as an example, the calculation steps after embedding the data reconstruction pattern are as follows: (1) Initialize the ensemble population. Determine the number of parameters to be optimized, M, the ensemble population size, N, and the upper and lower bounds of the parameters. , The initial generation set of individuals is set according to the following formula. (i=1,2…N, where i is the individual number in the population, Gen is the number of evolutionary iterations, φ is the parameter vector to be optimized, and j is the element number of the parameter vector, j=1,2…M):
[0036] In the formula: rand(0,1) means generating random numbers in the range [0,1].
[0037] (2) Population variation. The variation set of individuals is generated according to the following formula:
[0038] In the formula: This represents the mutation probability, typically ranging from [0.5, 1]. , , All are within the range [1, N], and , , None of them are i; if the mutated population is... [ , Then, the mutant individuals are recalculated using the above formula. Until the requirements are met.
[0039] (3) Population crossover. Individuals in the crossover population The generation process is as follows:
[0040] In the formula: The crossover probability is typically in the range of [0.8, 1]; jrand is a random integer generated when the j-th parameter population of the parameter group undergoes crossover, and it is within the range of [1, M].
[0041] (4) Population selection. Fitness functions were calculated for the population before mutation and the population after crossover, ensuring the non-negativity of the fitness function. The smaller the fitness function value, the better the fitness of the individual. The selection criteria for the Gen+1 generation population are as follows:
[0042] In the formula This is the fitness function.
[0043] The above differential evolution algorithm calculation steps are iterated repeatedly until a certain number of iterations are met or the evolution of the ensemble population converges to the set conditions. The average value of the last generation of the ensemble population is taken as the final optimal result.
[0044] In equation (6), the fitness function F is embedded with a data reconstruction pattern:
[0045]
[0046] In the formula: Reconstruct variable numbers for the data. The absolute value of the difference between the simulation and observation of the reconstructed variable can be expressed as:
[0047] In the formula: i is the observation point number, Here, d represents the predicted value of the reconstructed quantity, and N represents the observed value of the reconstructed quantity. ob This represents the total number of observation points.
[0048] Taking the ensemble Kalman filter algorithm as an example, the calculation steps after embedding the data reconstruction pattern are as follows: (1) Construct a set of parameters to be identified. Generate a set containing the parameters to be identified: X, Y, T, and K. L K N The parameter set contains N. r One implementation:
[0049] (2) Predicting the state. Substitute the parameter set vector updated at the t-1 assimilation step into the convection-diffusion model to simulate the concentration state at time step t. Simultaneously, the simulated concentrations at time steps 1 to t-1 can be updated at the same time:
[0050] In the formula: As a composite operator, it can be written as ,in Represents the convection-diffusion operator. This represents the conversion operator that transforms the directly observed concentration C into a reconstructed quantity; The initial concentration distribution of pollutants; This is the parameter set vector updated at time t-1; To predict the reconstructed vector, i.e. .
[0051] (3) Parameter update.
[0052]
[0053] In the formula: superscripts a and f represent the updated and unupdated states, respectively; superscripts ob and s represent the observed and simulated states, respectively; The observation error of the reconstructed quantity is set to 0 here. To update the damping factor, which can handle the set collapse problem during assimilation, its value ranges from (0,1), and here it can be 0.3. The Kalman gain matrix K is calculated as follows:
[0054] In the formula: The covariance matrix between the parameter set to be identified and the reconstructed quantity; To reconstruct the autovariance matrix of the quantity vector; The observation error vector The covariance matrix. and The following formulas can be used to calculate the results:
[0055]
[0056] (4) Repeat steps (2) and (3) until assimilation is complete.
[0057] Furthermore, in step 6), iteratively optimizing the parameter group based on the updated observation time series data is the iterative step in the above optimization algorithm. New monitoring data is added in each iteration, and the sequence from the previous step is repeated when there is no effective new monitoring data.
[0058] With Quanzhou Bay as the specific implementation area, see Figure 2 The steps are as follows: Step 1: Investigate the distribution of potential pollution sources and construct a pollutant database for the study area. According to recent environmental status bulletins and marine environmental monitoring reports, the main pollutants in Quanzhou Bay are inorganic nitrogen, reactive phosphate, and chemical oxygen demand (COD), mainly from land-based sources. In addition, the aquatic environment in this area is also affected to some extent by heavy metals and organic pollutants.
[0059] Step 2: Based on the characteristics of the study area and water pollution monitoring indicators, rationally deploy an online monitoring equipment network. Three monitoring points will be set up in the area, distributed as follows: Figure 2 As shown, the water pollution monitoring indicator is COD.
[0060] Step 3: Construct a hydrodynamic-water quality forward model based on multi-source data including topography, meteorology, and hydrology, and combine this with artificial intelligence optimization methods to build a water pollutant source tracing model. The scope and initial water depth of the hydrodynamic-water quality model are as follows: Figure 2 As shown, in this embodiment, the ensemble Kalman filter method is used as an intelligent optimization algorithm, and the pollutants are emitted in a continuous emission mode.
[0061] Step 4: Construct a relative concentration data series based on the pollutant monitoring concentration time series data.
[0062] Step 5: Reconstruct the observation state based on the relative concentration data sequence, and embed this data reconstruction pattern into the ensemble Kalman filter algorithm. Using this method, simultaneously identify the pollution source parameters (XY plane coordinates, initial discharge time T1, continuous discharge time T2) and the water quality model diffusion coefficients (lateral and longitudinal diffusion coefficients α and β).
[0063] Step 6: Update the observed time series based on the real-time monitoring data transmitted by the monitoring equipment network, and iteratively optimize the pollution source parameter set. The calculation results are as follows: Figure 3 As shown, by the 70th iteration, the characteristics of each parameter set had basically stabilized. (Except for the parameters...) Outside of the set, the mean values of the other parameter sets are very close to the actual values, especially in the identification of pollution source locations, where the deviation is minimal. This can provide reliable technical support for subsequent remedial and early warning work.
[0064] The logical ideas behind the methods disclosed in the above embodiments can be implemented, in whole or in part, through software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs.
[0065] When computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives (SSDs).
[0066] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments have been described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. Although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the invention itself. Various changes in form and detail may be made without departing from the spirit and scope of the invention as defined in the appended claims.
Claims
1. A water pollution source tracing method based on data reconstruction patterns, characterized in that, Includes the following steps: Investigate the distribution of potential pollution sources in the target area, construct a pollutant database, and determine pollution monitoring indicators for the target area based on the pollutant database; Based on the pollution monitoring indicators of the target area, deploy several water quality monitoring devices; Water quality monitoring equipment continuously collects pollutant concentration data to construct a relative pollution monitoring time series. The observation status is reconstructed based on the relative pollution monitoring time series, and the reconstructed data is embedded into an artificial intelligence algorithm. Collect topographic, meteorological, and hydrological data of the target area, construct a hydrodynamic-water quality forward model, and combine it with the aforementioned artificial intelligence algorithm to construct a water pollutant source tracing model; The water quality monitoring equipment collects pollutant monitoring concentration data in real time and updates the relative pollution monitoring time series. The updated relative pollution monitoring time series is then input into the water body pollutant source tracing model to update the pollution source location, emission time, and diffusion coefficient.
2. The water pollution source tracing method based on data reconstruction pattern according to claim 1, characterized in that, The distribution of potential pollution sources in the target area includes industrial parks, residential areas, and cargo ships within a predetermined range around the target area. Pollution source characteristics are obtained from these sources to construct a pollutant database. The pollution source characteristics are described using three types of parameters: source location, time, and release amount.
3. The water pollution source tracing method based on data reconstruction pattern according to claim 1, characterized in that, Based on the pollution monitoring indicators of the target area, deploy several water quality monitoring devices to ensure that the coverage of each water quality monitoring device overlaps, and establish a water quality monitoring network.
4. The water pollution source tracing method based on data reconstruction pattern according to claim 1, characterized in that, Pollutant concentration data from past periods are collected by water quality monitoring equipment to construct a relative pollution monitoring time series, which is expressed as follows: In the formula, Let be the relative concentration at the i-th observation time step; t is the time length of the concentration time series. It is the directly observed or simulated concentration at the i-th observation time step.
5. A water pollution source tracing method based on data reconstruction patterns according to claim 4, characterized in that, Based on the reconstructed observation state of the relative pollution monitoring time series, a new state quantity that can highlight different characteristics of the data series is derived. The new state quantity is calculated as follows: In the formula, Let be the relative concentration at the i-th observation time step; The concentration directly observed in step i; This represents the average concentration observed from step 1 to step t. Let be the simulated concentration in step i; This represents the average concentration from step 1 to step t. and These represent the relative concentrations observed and simulated in step i, respectively; , , These are the correlation coefficient, the sum of squares of relative deviations of relative concentrations, and the mean of relative concentrations at step t, respectively.
6. The water pollution source tracing method based on data reconstruction pattern according to claim 5, characterized in that, Based on the new state variables, an artificial intelligence algorithm is embedded, which includes a differential evolution algorithm or an ensemble Kalman filter assimilation method.
7. A water pollution source tracing method based on data reconstruction patterns according to claim 1 or 6, characterized in that, The hydrodynamic-water quality forward model includes a hydrodynamic model and a water quality model, wherein: The hydrodynamic model adopts the ELCIRC open-source model, and the computational grid adopts an irregular grid. The regional hydrodynamics are calculated as the dynamic conditions for material transport in the water quality model. The water quality model is constructed based on a two-dimensional convection-diffusion equation. An irregular grid is used for computation, and the finite volume method is employed for differential solution to calculate the material concentration process at all grid points.
8. The water pollution source tracing method based on data reconstruction pattern according to claim 2, characterized in that, Based on the aforementioned hydrodynamic-water quality forward model, and combined with the reconstructed data, a water pollutant source tracing model is constructed, specifically including: An objective function is constructed, and the three types of pollution source parameters and the horizontal and vertical diffusion coefficients are simultaneously optimized by calculating the extreme value of the objective function. The extreme value of the objective function is related to the substance concentration simulated by the water quality model under different combinations of pollution source parameters.
9. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the water pollution source tracing method based on data reconstruction pattern as described in any one of claims 1 to 8.
10. A computer-readable storage medium, characterized in that, The storage medium stores at least one executable instruction, which, when executed on an electronic device, causes the electronic device to perform the water pollution source tracing method based on data reconstruction pattern as described in any one of claims 1 to 8.