Search

CN-122021474-A - DNAPL pollution source area structure identification method and device based on Bayesian monitoring design

CN122021474ACN 122021474 ACN122021474 ACN 122021474ACN-122021474-A

Abstract

The invention discloses a DNAPL pollution source area structure identification method and device based on Bayesian monitoring design, and belongs to the technical field of polluted hydrogeological exploration. The method comprises the steps of generating an initial parameter field through a random percolation model, running a forward model to obtain concentration simulation data, further adopting Bayesian monitoring design to calculate relative entropy to determine an optimal observation point and obtain actual measurement data, then iteratively updating the parameter field based on a set smoother multi-data assimilation method, circularly executing monitoring design and data assimilation processes, and finally circularly repeating the forward modeling and inner layer optimization processes through an outer layer until the optimal monitoring design is obtained and a final estimation result of a pollution source area structure is output. The invention realizes more accurate pollution source area structure identification under the condition of limited cost or limited monitoring data by utilizing the Bayesian principle, and obviously improves the identification precision and efficiency of DNAPL pollution source area structures in a strong heterogeneous field.

Inventors

  • HAN ZHENG
  • WU YEQI
  • WU JICHUN
  • SHI XIAOQING

Assignees

  • 南京大学

Dates

Publication Date
20260512
Application Date
20260414

Claims (10)

  1. 1. The DNAPL pollution source area structure identification method based on Bayesian monitoring design is characterized by comprising the following steps of: Generating an initial saturation field and an initial permeability coefficient field through a random percolation model, and operating an underground water flow and solute migration forward model based on the initial saturation field and the initial permeability coefficient field to obtain pollutant concentration simulation data; Based on the initial saturation field, the initial permeability coefficient field and the pollutant concentration simulation data, calculating relative entropy by adopting a Bayesian monitoring design method, determining the position with the maximum relative entropy as an optimal observation point, and acquiring observation data at the optimal observation point, wherein the observation data comprises saturation data, permeability coefficient data and concentration data; based on the initial saturation field, the initial permeability coefficient field and the observed data, carrying out iterative updating by adopting a data assimilation ESMDA method of a set smoother until the maximum iterative times are met, and obtaining an updated saturation field and an updated permeability coefficient field; and taking the updated saturation field and permeability coefficient field as new initial parameter fields, re-performing forward modeling, optimizing monitoring points and data assimilation updating until the target well position number is reached, obtaining an optimal monitoring design, and obtaining a corresponding estimation result of the saturation field and the effective permeability coefficient field.
  2. 2. The method of claim 1, wherein the forward model of groundwater flow and solute transport is as follows: The presence of DNAPL will be considered in the groundwater flow model by the effective permeability coefficient K eff = K i ·K r (S N ), where K i is the intrinsic hydraulic conductivity, K r is the relative hydraulic conductivity, S N is the saturation, assuming the groundwater flow is a steady flow, its control equation is: , , The boundary conditions are: , , Wherein, the Is gradient, q is vector flow rate, h is head, Γ D is Dirichlet boundary, h D is given head on Γ D , Γ N is Neumann boundary, n represents the outer unit vector perpendicular to Γ N ; Based on a stable groundwater flow field, the DNAPL pollution source region can be used as a Dirichlet boundary in the research region on the assumption that DNAPL reaches local equilibrium, so that the migration problem of DNAPL is solved by a convection dispersion equation of the stable flow: , The initial conditions and boundary conditions are: , , , Wherein, the In order to achieve a degree of porosity, the porous material, C is the concentration of dissolved phase DNAPL, C 0 is the concentration of DNAPL in the initial state, t is the solute transport time, For effective porosity, q s is the solute flux at Neumann boundary Γ N , C S is the solubility of DNAPL, For a preset concentration at the x k position, Ω represents the entire investigation region, and the inner Dirichlet boundary region Γ D is the DNAPL contamination source region.
  3. 3. The method according to claim 1, characterized in that the bayesian monitoring design method calculates the relative entropy specifically by the following formula: , Wherein, m is pollution source parameter, which comprises effective permeability coefficient lnK eff and saturation S N , d is pollution source parameter observation value, which comprises effective permeability coefficient lnK eff , saturation S N and concentration C; for the relative entropy of the sampling scheme c, For the posterior distribution of the parameters, And selecting a sampling optimization scheme with the maximum relative entropy as an optimal sampling position.
  4. 4. The method according to claim 1, characterized in that the specific method of determining the optimal sampling position comprises the steps of: Setting an initial iteration state, wherein the iteration times j=1, and inputting all possible horizontal positions and uniformly selected vertical position parameters; calculating the relative entropy of each horizontal position, and selecting a position c j with the maximum relative entropy from the calculated relative entropy as the current optimal horizontal sampling point; Acquiring posterior samples and inversion analysis, namely acquiring an observation value at a selected optimal horizontal position c j , performing ESMDA inversion, and updating parameter posterior distribution; Applying spatial weight decay, namely applying weight decay in the selected position and the adjacent area; And (3) performing loop iteration, namely adding 1 to the iteration times, repeatedly selecting an optimal horizontal sampling position, acquiring a posterior sample, performing inversion analysis and applying a space weight attenuation step until a termination condition is met, and finally outputting a self-adaptive single-layer monitoring network design.
  5. 5. The method according to claim 1, characterized in that the specific method of determining the optimal sampling position comprises the steps of: Setting an initial iteration state, enabling the iteration times j=1, and inputting all possible horizontal and vertical sampling position parameters; the self-adaptive optimization vertical sampling position is that for each horizontal position, the optimal vertical sampling point is dynamically selected from high to low according to the index of relative entropy, and a weight attenuation mechanism is applied to the selected position and the adjacent area; Selecting optimal horizontal sampling positions, namely dynamically selecting the optimal sampling positions from high to low based on the sum of the relative entropy of all vertical points on each horizontal position, and finally determining the globally optimal horizontal well distribution position; And (3) performing loop iteration, namely adding 1 to the iteration times, repeating the steps of adaptively optimizing the vertical sampling position and selecting the optimal horizontal sampling position until the termination condition is met, and finally outputting the adaptive multi-level monitoring network design.
  6. 6. The method of claim 1, wherein the set smoother multiple data assimilation ESMDA method comprises the steps of: Selecting the number of data assimilation times N a and the corresponding expansion coefficient of each assimilation iteration Wherein t=1, 2, carrying out N a ; extracting N samples from the prior distribution to form an initial sample parameter set; starting iterative operation from t=1, and operating a forward model to obtain a simulation value set corresponding to a sample based on model parameters updated in the previous iteration in each iteration: , In the formula, The model parameters obtained in the t-th iteration are obtained; f (m) is a forward model for the simulation value obtained by the t-th iteration; The error covariance matrix is calculated according to the following formula: , Where C MD is the covariance matrix between the model parameters and the analog values, C DD is the auto-covariance matrix of the analog values, As the mean value of the parameters of the model, The superscript T represents transposition for the average value of the analog values; iterating N a times to update the sample parameter set and obtain an inversion estimate of the posterior distribution of the parameters according to the following formula: , Where C D is the covariance matrix of the observation error and d obs is the observation value after adding the disturbance.
  7. 7. A DNAPL pollution source area structure identification system based on bayesian monitoring design, comprising: the forward modeling module is used for generating an initial saturation field and an initial permeability coefficient field through the random percolation model, and operating the underground water flow and solute migration forward modeling module based on the initial saturation field and the initial permeability coefficient field to obtain pollutant concentration simulation data; The monitoring point optimizing module is used for calculating relative entropy by adopting a Bayesian monitoring design method based on the initial saturation field, the initial permeability coefficient field and the pollutant concentration simulation data, determining the position with the maximum relative entropy as an optimal observation point, and acquiring observation data at the optimal observation point, wherein the observation data comprises saturation data, permeability coefficient data and concentration data; the data assimilation updating module is used for carrying out iterative updating by adopting a data assimilation ESMDA method of the set smoother based on the initial saturation field, the initial permeability coefficient field and the observed data until the maximum iterative times are met, so as to obtain an updated saturation field and an updated permeability coefficient field; And the iteration control and result output module is used for taking the updated saturation field and the updated permeability coefficient field as new initial parameter fields, re-carrying out forward modeling, optimizing and data assimilating updating on monitoring points until the target well position number is reached, obtaining the optimal monitoring design, and obtaining the estimation results of the corresponding saturation field and the effective permeability coefficient field.
  8. 8. An electronic device comprising one or more processors, memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, which when executed by the processor implement the steps of the bayesian monitoring design based DNAPL pollution source structure identification method of any one of claims 1 to 6.
  9. 9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the DNAPL pollution source structure identification method based on bayesian monitoring design as claimed in any one of claims 1-6.
  10. 10. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the steps of the DNAPL pollution source structure identification method based on bayesian monitoring design according to any of claims 1-6.

Description

DNAPL pollution source area structure identification method and device based on Bayesian monitoring design Technical Field The invention relates to the technical field of polluted hydrogeological exploration, in particular to a DNAPL pollution source area structure identification method and equipment based on Bayesian monitoring design. Background In groundwater environment, heavy nonaqueous phase liquid (DNAPL) pollutants have the characteristics of high density, low viscosity, high volatility, strong penetrability, difficult degradation and the like. Compared with Light non-aqueous liquid (LNAPL) pollutants, the Light non-aqueous liquid (5326) has wider migration range in the underground environment and more complex occurrence structure, so that the corresponding repair and treatment difficulty is obviously increased. Therefore, the DNAPL pollution site can be effectively repaired, and the key premise is that the space distribution structure of the pollution source region and the pollution plumes is accurately described, and the distribution form of the pollution plumes is mainly determined by the internal structure of the pollution source region. The structure of the pollution source area essentially describes the complex mass distribution characteristics of DNAPL in space, and particularly relates to the breadth of a pollution range, the saturation degree of pollutants, the occurrence form and the spatial position distribution of the pollution source area. The structure is significantly controlled by the heterogeneity (such as spatial distribution of permeability coefficients) of the aquifer medium, and small changes in permeability coefficients can significantly affect the structure of the pollution source region. Therefore, accurately describing the structure of a pollution source region and the spatial heterogeneity of permeability coefficients has become the most critical technical link in DNAPL pollution site repair. In order to realize high-precision identification of DNAPL pollution source area structures, the traditional investigation method generally relies on intensive drilling and sampling. Such methods are costly, long in implementation period, and are less viable in practical complex sites. To solve the problem of limited monitoring data, it is important to design an effective monitoring network to obtain monitoring data capable of maximizing information gain. However, the conventional optimization method (such as a statistical method, a spatial interpolation method, an optimization-based method, a hybrid method and the like) is mainly designed for optimizing the monitoring network for identifying the pollution of the underground water point source, and is difficult to be suitable for sampling optimization of the complicated DNAPL pollution source area structure. Therefore, in order to improve the depiction precision of the pollution source area structure on the premise of limited cost (or limited monitoring data), it is necessary to develop a method for adaptively optimizing sampling strategy for identifying the DNAPL pollution source area structure, so as to realize the optimal identification of the pollution source area structure, thereby providing reliable technical support for the accurate restoration of the pollution site. Disclosure of Invention Aiming at the defects of high cost and long implementation period of the traditional uniform sampling scheme, the invention provides a Bayesian monitoring design-based adaptive multi-level sampling optimization method which is difficult to adapt to the defects of irregular form, obvious influence of the heterogeneity of an aquifer medium on the sampling optimization of a DNAPL pollution source region structure, and the like, so as to realize more accurate pollution source region structure identification under the condition of limited cost (or limited monitoring data), and the Bayesian monitoring design-based adaptive multi-level sampling optimization framework can provide reliable decision basis for risk assessment and restoration of DNAPL pollution sites. The technical scheme is that in order to achieve the aim of the invention, the invention adopts the following technical scheme: a DNAPL pollution source area structure identification method based on Bayesian monitoring design comprises the following steps: Generating an initial saturation field and an initial permeability coefficient field through a random percolation model, and operating an underground water flow and solute migration forward model based on the initial saturation field and the initial permeability coefficient field to obtain pollutant concentration simulation data; Based on the initial saturation field, the initial permeability coefficient field and the pollutant concentration simulation data, calculating relative entropy by adopting a Bayesian monitoring design method, determining the position with the maximum relative entropy as an optimal observation point, and acquiring