Simulation-Optimization Method for Groundwater Pumping-Processing Pollution Remediation System in Contaminated Sites

Abstract

The present invention discloses a simulation-optimization method for a groundwater pumping and treatment pollution remediation system in a contaminated site based on a multi-form evolutionary search paradigm, including: constructing a groundwater flow model and a pollutant solute transport model; determining the management objective function and constraint conditions for hydraulic control of the groundwater pollution range, establishing a groundwater pollutant treatment model, optimizing the groundwater pumping and treatment in the contaminated site by using BLMFO, solving the interdependence problem between the upper-layer global objective and the lower-layer local objective through nested optimization, and constructing an auxiliary task by using MFO; updating the state variables through the groundwater flow model and the pollutant solute transport model, calculating the value of the management objective function of the groundwater pollutant treatment model, and determining whether the constraint conditions are satisfied; and simultaneously selecting decision variables by optimizing the groundwater pollutant treatment model and returning to update the state variables in the groundwater flow model and the pollutant solute transport model.

Description

Technical Field

[0001] The present invention relates to the field of groundwater pollution control engineering and environmental science applications, and particularly to a simulation-optimization method for a groundwater pumping-injection-treatment pollution remediation system at a contaminated site based on a multi-form evolutionary search paradigm. Background Art

[0002] In the field of groundwater pollution control engineering and environmental science applications, certain progress has been made in the decision-making management of groundwater remediation at contaminated sites. For example, patent application CN102992417A discloses a decision-making method for in-situ remediation of petroleum-contaminated groundwater. Based on the in-situ extraction and treatment technology, by collecting relevant parameters of the contaminated site, using numerical simulation, regression methods, and nonlinear optimization techniques, a surrogate simulation model of the remediation plan is obtained, and then an optimal groundwater remediation plan is produced. Patent application CN105718632A proposes an uncertainty groundwater remediation multi-objective optimization management method. Based on the multi-objective stochastic tabu search algorithm PEMOTS with elitist retention strategy, combined with sequential Gaussian conditional simulation SGSIM to reduce the uncertainty of aquifer system parameters, it is coupled with the groundwater flow program MODFLOW and the solute transport program MT3DMS, and has strong reliability and robustness. Patent application CN105819616A relates to an integrated remediation system for shallow groundwater pollutants under multi-objective conditions, including a groundwater pollution migration monitoring system, a diagnosis system, a groundwater remediation system, and a post-treatment and maintenance system. By real-time monitoring pollutant data and simulating and calculating the migration and transformation paths of pollutants, etc., a groundwater pollution remediation management plan is established. Patent application CN113240282A discloses a method for determining mercury pollution soil remediation standards hierarchically with multiple objectives. Based on the theory of human health risk assessment and soil pollution assessment theory, the mercury remediation target values at different levels are calculated, providing a basis for the selection of mercury remediation target values in soil. Patent application CN115659719A is an optimization method for in-situ bioremediation of chlorinated hydrocarbon pollution in an aquifer. By constructing a simulation model and a multi-objective optimization model, and using the NSGA-II algorithm to solve, an optimal set of remediation plans is obtained, improving the simulation accuracy and weighing the contradictory relationship between remediation cost and remediation effect.

[0003] These existing technologies have played an important role in the field of groundwater pollution remediation and have certain advantages. For example, some technologies can combine multiple advanced technologies, such as numerical simulation, optimization algorithms, etc., to provide a scientific basis for the formulation of remediation plans, improving the reliability and efficiency of remediation decisions; some methods consider the complexity and uncertainty of groundwater pollution, and by introducing corresponding models and algorithms, reduce the uncertainty of the models and enhance the adaptability and robustness of the remediation plans; there are also technologies that have constructed a relatively complete groundwater pollution remediation system, covering multiple links such as monitoring, diagnosis, remediation, and post-maintenance, achieving all-round control and remediation of groundwater pollution.

[0004] However, there are also some deficiencies in the existing technologies. In terms of the applicability of algorithms, some of the optimization algorithms adopted by certain methods may have problems with insufficient diversity and convergence of solutions when dealing with complex multi-objective optimization problems such as the groundwater extraction-treatment pollution remediation system for chlorinated hydrocarbon contaminated sites. Although the NSGA-II algorithm performs well in general multi-objective optimization problems, for the multiple decision variables and complex constraint conditions involved in this system, it may be difficult to effectively explore the solution space, maintain the diversity of solutions, and is prone to falling into local optimal solutions.

[0005] In terms of the difficulty of convergent solution, due to the complexity of the models and large computational amounts in some technologies, the convergence speed is slow and the solution difficulty is high. For example, the in-situ bioremediation scheme optimization method for chlorinated hydrocarbon pollution in aquifers mentioned in patent application CN115659719A involves the coupling of multiple models, including the in-situ bioremediation simulation model for chlorinated hydrocarbon pollution in aquifers and the multi-objective optimization model for the in-situ bioremediation scheme of chlorinated hydrocarbon pollution in aquifers. Although this coupling method can improve the simulation accuracy and optimization effect, in practical applications, the complexity of model coupling is high, the computational amount is large, and the requirements for computing resources and time are also high, affecting the practicality and promotion of the method.

[0006] In terms of the high complexity of simulation-optimization coupling, some methods in the existing technologies have problems with high coupling complexity when coupling the simulation model with the optimization model. For example, the multi-objective optimization management method for uncertain groundwater remediation mentioned in patent application CN105718632A, although coupled with the groundwater flow program MODFLOW and the solute transport program MT3DMS, has strong reliability and robustness, but in the coupling process, a large amount of data and complex model relationships need to be processed, increasing the difficulty and complexity of coupling.

[0007] In terms of high system integration complexity, although the groundwater pollution remediation systems constructed by some technologies are comprehensive in function, their system integration complexity is high. For example, the integrated remediation system for shallow groundwater pollutants under multi-objective conditions mentioned in patent application CN105819616A covers multiple links such as monitoring, diagnosis, remediation, and post-maintenance, forming a relatively complete system. However, this system integration involves various different treatment processes and equipment, such as coagulation sedimentation, heavy metal separation, biological denitrification, ultrafiltration, and reverse osmosis. In practical applications, there may be certain difficulties in the coordinated cooperation between these processes and equipment, and it is necessary to further optimize the system integration scheme to reduce complexity and improve the stability and reliability of the system. Summary of the Invention

[0008] The present invention aims to solve various deficiencies existing in the simulation-optimization methods of the groundwater pumping-treatment pollution remediation system for contaminated sites such as chlorinated hydrocarbons. In the prior art, when the optimization algorithm deals with complex multi-objective problems, the diversity and convergence of solutions are insufficient, and it is easy to fall into local optimal solutions, unable to provide a comprehensive and reasonable remediation plan for decision-makers. At the same time, due to the high model complexity and large computational amount of the simulation-optimization method, the convergence speed is slow and the solution difficulty is high, making it difficult to meet the requirements of rapid and efficient decision-making in actual projects. In addition, when the prior art couples the groundwater flow and pollutant transport simulation model with the optimization management model, the coupling complexity is high, and data interaction and collaborative work are difficult, affecting the reliability and stability of the model. In terms of system integration, it involves various treatment processes and equipment, and the coordinated cooperation is difficult, with high integration complexity, reducing the stability and reliability of the system. In the actual remediation process, there is a lack of a real-time feedback and adjustment mechanism, and the remediation plan cannot be adjusted in a timely manner according to real-time monitoring data, affecting the remediation effect and resource utilization efficiency. In terms of multi-objective coordinated optimization, the prior art does not sufficiently balance and coordinate different objectives, and it is difficult to effectively control the remediation cost and resource consumption while meeting the remediation effect. In view of these deficiencies, the present invention proposes a simulation-optimization method for the groundwater pumping-injection-treatment pollution remediation system of contaminated sites based on a multi-form evolutionary search paradigm. By introducing the multi-form evolutionary search paradigm (MFO) and the double-layer multi-objective optimization framework (BLMFO), the optimization efficiency and the diversity of solutions are improved, the model coupling complexity is reduced, the real-time feedback and adjustment ability of the system is enhanced, and the coordinated optimization between multiple objectives is realized, providing scientific, efficient, and economic decision-making support for the groundwater remediation of contaminated sites.

[0009] The specific technical solutions are as follows:

[0010] A simulation-optimization method for the groundwater pumping-injection-treatment pollution remediation system of contaminated sites based on a multi-form evolutionary search paradigm, comprising:

[0011] Construct a groundwater flow model for the contaminated site. On this basis, set the initial pumping and injection well scheme, and establish a solute transport model for groundwater pollutants in the contaminated site;

[0012] Determine the management objective function and constraint conditions for the hydraulic control of the groundwater pollution range in the contaminated site, establish a groundwater pollutant treatment model with pumping and injection water treatment as the core, optimize the groundwater pumping and injection treatment in the contaminated site using a bilayer multi-objective optimization framework (BLMFO), solve the mutual dependence problem between the upper-layer global objective and the lower-layer local objective through nested optimization, and use a multi-form evolutionary search paradigm (MFO) to construct auxiliary tasks; the upper-layer global objective, that is, the management objective function, includes minimizing the remediation cost and maximizing the pollutant removal efficiency; the lower-layer local objectives include the pumping and injection water volume and the operation time of each well;

[0013] Update the state variables through the groundwater flow model and the solute transport model of groundwater pollutants in the contaminated site, calculate the value of the management objective function of the groundwater pollutant treatment model, and determine whether the constraint conditions are met; at the same time, select decision variables through the optimized groundwater pollutant treatment model and return to the groundwater flow model and the solute transport model of groundwater pollutants in the contaminated site to update the state variables.

[0014] In some embodiments, for the method of simulating and optimizing the groundwater pumping-injection treatment and pollution remediation system in the contaminated site based on the multi-form evolutionary search paradigm, the construction of the groundwater flow model for the contaminated site includes: statistically analyzing the borehole formation data of the contaminated site and the distribution data of pollutants in the groundwater, using the MODFLOW (seepage flow) module of GMS (Groundwater Modeling Systems) software to determine the aquifer structure, source-sink terms, and boundary conditions of the contaminated site, constructing the groundwater flow model for the contaminated site, and adjusting the groundwater flow model parameters to the optimal through groundwater observation data.

[0015] In some embodiments, for the method of simulating and optimizing the groundwater pumping-injection treatment and pollution remediation system in the contaminated site based on the multi-form evolutionary search paradigm, use the MT3DMS (Modular Three-Dimensional Solute Transport Model) module to establish a solute transport model for groundwater pollutants in the contaminated site.

[0016] In some embodiments, for the method of simulating and optimizing the groundwater pumping-injection treatment and pollution remediation system in the contaminated site based on the multi-form evolutionary search paradigm, the state variables in the groundwater pumping-injection treatment and pollution remediation system in the contaminated site are the groundwater level and the pollutant concentration value in the groundwater in the contaminated site, and the decision variables include any one or more of the following: the pumping and injection water volume of the pumping and injection wells, the location and number of the pumping and injection wells, the operation time, and the scheduling, which determines the specific time sequence of pumping and injecting water.

[0017] In some embodiments, for the simulation-optimization method of the groundwater pumping and treatment pollution remediation system based on the multi-form evolutionary search paradigm, the decision variables include the pumping volume of the injection well, and the following groundwater pollutant treatment model is established:

[0018] ;

[0019] ;

[0020] ;

[0021] ;

[0022] Where: is the remediation cost; represents minimizing the remediation cost; is the pumping cost coefficient of the injection well, which can be set to 0.15 yuan / (m 3 ·m), etc.; is the pumping volume of the th injection well (the unit can be m 3 / d); is the continuous pumping time of the th injection well (the unit can be d); is the number of injection wells controlled by the pumping and treatment pollution remediation system; is the number of iterative optimizations, is the percentage of the remaining groundwater pollutant mass after the th iterative optimization (the unit can be kg) to the initial groundwater pollutant mass (the unit can be kg); represents iterative optimization to maximize the pollutant removal efficiency; is the pollutant concentration in the groundwater of the polluted site after the th iterative optimization; is the maximum allowable concentration value of pollutants in the groundwater of the polluted site; is the area of the pollutants with excessive groundwater concentration in the polluted site after pumping and treatment remediation; is the initial area of the pollutants with excessive groundwater concentration in the polluted site;

[0023] Optimize the pumping volume to meet the constraints of the groundwater flow field and pollutant concentration.

[0024] In some embodiments, for the simulation-optimization method of the groundwater pumping-treatment pollution remediation system for contaminated sites based on the multi-form evolutionary search paradigm, multiple groups of initial pumping well schemes are generated in the upper layer as initial solutions, and each solution in the upper layer, that is, each group of pumping well schemes, corresponds to an independent optimization in the lower layer. The MT3DMS module is called to simulate the migration of groundwater pollutants;

[0025] The lower layer optimizes the pumping volume, generates the optimal lower layer solution set (Pareto front) corresponding to the upper layer solution, and feeds it back to the upper layer to calculate the global fitness value.

[0026] Furthermore, the NSGA-II algorithm is used in the lower layer to optimize the pumping volume. The NSGA-II algorithm randomly initializes the population individuals, where each individual represents the pumping volume vector group of all pumping wells:

[0027] ;

[0028] ;

[0029] where, represents the total number of individuals in the population generated by NSGA-II, and the individual represents a pumping scheme, .

[0030] In some embodiments, for the simulation-optimization method of the groundwater pumping-treatment pollution remediation system for contaminated sites based on the multi-form evolutionary search paradigm, the construction of auxiliary tasks using the multi-form evolutionary search paradigm includes:

[0031] Construct auxiliary task 1: First, ignore the lower layer optimization constraints, simplify the multi-objective problem, and regard the lower layer optimization as a penalty term added to the upper layer objective function:

[0032] ;

[0033] ;

[0034] where: represents the objective function of auxiliary task 1; represents the upper layer objective function, that is, minimizing the remediation cost and maximizing the pollutant removal efficiency; is the remediation cost; is the percentage of the remaining pollutants in the groundwater after pumping treatment and remediation to the initial pollutants in the groundwater; represents the decision variable vector; is the penalty coefficient, used to control the penalty degree for violating the constraints. For example, it can be set to 0.1 in the initial stage, increased by 0.1 every 50 generations, and finally stabilized at 1.0; Indicates the degree of violation of the lower - layer constraints, , is the water injection and pumping volume constraint for a single well, is the hydraulic gradient constraint;

[0035] Construct auxiliary task 2: Integrate the lower - layer local objectives into the upper - layer to form a single - layer multi - objective optimization problem, and through this task, discover the trade - off relationship between the water injection and pumping volume and the removal efficiency of groundwater pollutants; the lower - layer objectives include the relationship between the water injection and pumping volume and the groundwater pollutant concentration;

[0036] The lower - layer optimization problem is used to simulate the migration of groundwater pollutants to calculate the final residual amount of groundwater pollutants:

[0037] ;

[0038] Where: is the pollutant degradation rate, is the hydrodynamic dispersion coefficient, is the groundwater flow velocity, and are respectively the second - order gradient term (diffusion term) and the first - order gradient term (convection term) of the pollutant concentration , represents the total number of grid cells after the spatial discretization of the polluted site, represents the label of the grid cell, , represents the set of optimal lower - layer decision variables under the given upper - layer decision variables ;

[0039] The two - layer multi - objective optimization framework uses explicit knowledge transfer (KT) to interact data between the two auxiliary tasks and the original problem to enhance the optimization effect, that is, perform an explicit knowledge transfer across tasks after iterating a certain number of times (such as 10 times, etc.) to ensure the co - evolution of the main task and the auxiliary tasks.

[0040] In some embodiments, for the method for simulating and optimizing the groundwater pumping - treatment pollution remediation system of the polluted site based on the multi - form evolutionary search paradigm, the solutions belonging to the Pareto front in the auxiliary tasks are preferentially selected. The fitness of the solutions of the auxiliary tasks in the main task needs to be higher than the current population mean, and the solutions are screened by the crowding distance to avoid population aggregation and ensure that the layout of the pumping wells covers different sub - regions of the polluted site;

[0041] When the solutions of the auxiliary tasks are transferred to the main task, the lower - layer decision variables including the water injection and pumping volume are supplemented, and the extension is completed by randomly extracting the lower - layer parameters of the historical optimal solutions;

[0042] The feasible solutions of the main task are fed back to the auxiliary task to guide its search direction, including initializing the auxiliary population using the verified well location scheme.

[0043] A computer device includes a memory and a processor. The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory. When the computer program runs, the processor executes the simulation-optimization method for the groundwater pumping and treatment pollution remediation system of a contaminated site based on the multi-form evolutionary search paradigm.

[0044] A computer-readable storage medium stores a program or instructions. When the program or instructions are executed by a computer device, the computer device executes the simulation-optimization method for the groundwater pumping and treatment pollution remediation system of a contaminated site based on the multi-form evolutionary search paradigm.

[0045] By introducing a double-layer multi-objective optimization framework (BLMFO) and a multi-form evolutionary search paradigm (MFO), the present invention solves the key problems existing in traditional groundwater remediation optimization methods, significantly improves the remediation efficiency, reduces the cost, and overcomes multiple technical obstacles in the prior art. Compared with the prior art, the present invention has achieved breakthrough innovations in the following aspects:

[0046] (1) The contradiction between solution diversity and convergence

[0047] When dealing with complex multi-objective problems, traditional methods often struggle to balance solution diversity and convergence and are prone to falling into local optimal solutions. The present invention constructs an auxiliary task (such as ignoring the lower-layer optimization or transforming it into a single-layer problem), quickly generates a diverse solution set in a low-complexity space, and combines a knowledge transfer mechanism to dynamically feed back the non-dominated solutions in the auxiliary task to the main task to ensure the global distribution of solutions. For example, in a chlorinated hydrocarbon contaminated site case in Jiangsu, the Pareto solution set generated by the present invention contains 6 different solutions, while traditional methods only generate 3 homogeneous solutions, significantly improving the flexibility of decision-making and the remediation effect.

[0048] (2) The high computational cost of simulation-optimization coupling

[0049] During the simulation-optimization coupling process of traditional methods, complex numerical simulation modules (such as MT3DMS) need to be repeatedly called, resulting in high computational costs and slow convergence speeds. Instead of relying on surrogate models (such as Kriging models), the present invention uses a multi-form evolutionary search paradigm and a knowledge transfer mechanism to replace part of the simulation calculations, only calls the numerical simulation (MT3DMS) for promising solutions, avoids repeated calls on a global scale, and combines a parallel optimization architecture to significantly reduce the computational amount. Under the same case, the present invention is at 3×10 5The secondary function converges to the optimal solution within the evaluation, while the traditional method requires 1.5×10 6 times, and the calculation efficiency is increased by 80%. In addition, the present invention supports real-time feedback and dynamic adjustment. The traditional method has a high calculation cost. Each optimization requires iterative calls to the numerical model thousands to millions of times, and each calculation cycle takes more than a dozen hours to dozens of hours. The present invention adopts a knowledge transfer mechanism and uses explicit knowledge transfer to exchange data between the main task and the auxiliary task. High-quality solutions can be transmitted from the auxiliary task to the main task, and the calculation cycle is within a few hours. If a well location plan cannot effectively reduce the pollution concentration, the optimization process can support manual input of information and parameters for adjustment, and update the remediation plan once, greatly improving the real-time performance and adaptability of the remediation.

[0050] (3)Optimization of remediation cost and pollutant removal efficiency

[0051] Through a two-layer optimization framework, the present invention screens low-well location plans (referring to the optimized plans with relatively few injection and production wells screened out during the upper-layer global optimization) in the upper-layer global optimization, and accurately controls the injection and production water volumes in the lower-layer local optimization, achieving dual optimization of the remediation cost and the pollutant removal efficiency. In a case of a chlorinated hydrocarbon contaminated site in Jiangsu, the present invention reduces the number of production wells from 23 to 13, and the total production water volume from 1150 m³ / d to 392 m³ / d. The remediation cost is reduced by 70%, and at the same time, the proportion of residual pollutants is reduced from 12.5% to 0.25%, significantly superior to the traditional method.

[0052] (4)Innovation of dynamic penalty coefficient and constraint handling

[0053] The present invention proposes an adaptive penalty coefficient mechanism, allowing mild constraint violations in the initial stage to explore diversity and strictly restricting in the later stage to ensure the feasibility of the solution. This innovation effectively solves the deficiencies of the traditional method in constraint handling and significantly increases the proportion of feasible solutions. In the test problems, the proportion of feasible solutions of the present invention is increased from 65% of the traditional method to 92%, further verifying its technical advantages.

[0054] In summary, through a two-layer coupled optimization architecture, a collaborative mechanism of multi-form search, and a dynamic knowledge transfer strategy, the present invention has achieved a breakthrough improvement in core indicators such as remediation cost, pollutant removal efficiency, and calculation efficiency. Compared with the traditional method, the present invention has significant advantages in terms of solution diversity, calculation efficiency, real-time performance, and remediation effect, and the technical solution is irreplaceable, meeting the requirements of invention creativity. Description of the Drawings

[0055] Figure 1 It is a schematic flow chart of a simulation-optimization method for a groundwater pumping and treatment pollution remediation system in a chlorinated hydrocarbon contaminated site based on a multi-form evolutionary search paradigm in the specific implementation manner.

[0056] Figure 2 For Figure 1 The solution set diagram when the groundwater pumping-injection-treatment pollution remediation system simulation-optimization method for the chlorinated hydrocarbon contaminated site shown in the figure runs for 100 generations. Specific implementation manners

[0057] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention.

[0058] The specific solution process of the BLMFO algorithm used in the present invention is as follows:

[0059] (a) Problem definition and initialization

[0060] The upper-level problem is defined as the global objective, that is, minimizing the pumping cost and maximizing the pollution treatment efficiency; the lower-level problem is positioned as the specific objective, that is, the layout of pumping and injection wells, the pumping and injection water volumes, etc.; the upper-level decision sets boundary conditions for the lower-level, but the optimization results of the lower-level will be fed back to the upper-level. The initialization of BLMFO is divided into two parts: the upper-level and the lower-level. Randomly generate N L upper-level solutions and N F lower-level solutions. To ensure the optimization efficiency and the diversity of the solution set, quality control is required for the initial population. By means of random generation and fitness screening, solutions that significantly violate the constraint conditions are eliminated, and it is ensured that the solutions are evenly distributed to cover the search space. The fitness value of the upper-level solution depends on the lower-level optimization result.

[0061] (b) Introducing auxiliary tasks to simplify the double-layer problem

[0062] The core of the optimization process of BLMFO is to introduce auxiliary tasks through Multiform Optimization (MFO) to reduce the complexity of double-layer optimization. MFO constructs multiple auxiliary tasks to simplify the original double-layer problem, and at the same time improves the algorithm efficiency through parallel optimization. First, a complete double-layer optimization is carried out in the main task. For the upper-level solution, the lower-level optimization is called to find the corresponding optimal lower-level solution. The main task needs to handle the mutual dependence between the upper and lower levels through nested optimization.

[0063] Auxiliary task A: Ignore the lower-level optimization problem and directly optimize the upper-level objective function, but ignore the optimization of the lower-level objective, and only indirectly constrain the feasibility of the lower-level through a penalty function.

[0064] Auxiliary task B: Convert the double-layer problem into a single-layer problem. By directly integrating the feedback objective of the lower-level into the upper-level objective, a new single-layer multi-objective problem is formed.

[0065] And decompose the complex double-layer multi-objective optimization problem into several sub-objectives, such as optimizing the pollutant removal amount or pumping cost separately.

[0066] Optimize the population in parallel in the main task and the auxiliary task. Use the multi-objective evolutionary algorithm (NSGA-II) for population evolution, and explore the search space by combining crossover, mutation, and selection operations. The auxiliary task quickly provides high-quality candidate solutions for sub-problems with lower computational complexity, and these solutions will be fed back to the main task through the knowledge transfer mechanism.

[0067] (c) Information transfer mechanism

[0068] Enhance the diversity of the solution set and the global search ability by dynamically exchanging solutions between the main task and the auxiliary task. Knowledge transfer mainly includes two directions: solution transfer from the auxiliary task to the main task, and solution feedback from the main task to the auxiliary task. Solution transfer from the auxiliary task to the main task: After every fixed number of generations (such as 10 generations), select the best-performing solutions from the auxiliary task and transfer them to the main task. The selection of these solutions is based on the following rules:

[0069] Non-dominance: Solutions belonging to the Pareto front in the auxiliary task solution set will be preferentially selected.

[0070] Fitness improvement: The fitness value of the auxiliary task solutions in the main task population is higher than the average fitness of the current population.

[0071] Diversity maintenance: Ensure that the transferred solutions do not cause the population to be overly concentrated in a certain local area through the Crowding Distance.

[0072] Solution feedback from the main task to the auxiliary task: Select the solution with the largest crowding distance from the Pareto front solution set of the main task and feed it back to the auxiliary task to optimize the diversity of the auxiliary task population. After receiving the main task solutions, the auxiliary task uses these solutions as a guide for population initialization or directly replaces part of the population, promoting the auxiliary task to explore the search space more comprehensively.

[0073] Formation of the Pareto solution set: After each population evolution, perform non-dominated sorting on all solutions in the main task to generate the Pareto front solution set. Retain the diversity by selecting non-dominated solutions and combining the crowding distance, and finally form the solution set. The high-quality solutions provided by the auxiliary task will also be gradually supplemented into the Pareto solution set of the main task through knowledge transfer.

[0074] The BLMFO algorithm is particularly suitable for complex problems that require hierarchical decision-making compared to traditional optimization methods. In such problems, the upper and lower layers have different objectives and constraints. By searching simultaneously in the original problem space and the auxiliary task space and leveraging knowledge transfer, the search efficiency and optimization performance can be improved. The multi-form search framework (BLMFO) accelerates and simplifies the optimization process by generating auxiliary tasks (such as ignoring the lower-layer optimization problem or transforming the problem into a single-layer multi-objective optimization). This method can generate alternative models in groundwater pollution management to simulate different remediation strategies or pollution propagation scenarios, thereby providing additional information support for the original optimization.

[0075] The core of the proposed multi-form evolutionary search paradigm (MFO) lies in parallelly exploring the search space by constructing multiple optimization tasks (the main task and auxiliary tasks), thereby enhancing the solution efficiency for complex optimization problems. Its characteristics are as follows:

[0076] 1) Construct auxiliary tasks that are related to the main task but have lower complexity;

[0077] 2) Each task occupies a different local region in the original search space;

[0078] 3) Execute the search in parallel between the main task and the auxiliary tasks;

[0079] 4) The auxiliary tasks help the main task improve the convergence speed through knowledge transfer;

[0080] 5) Transfer the high-quality solutions obtained in the auxiliary tasks to the main task;

[0081] 6) The solutions in the main task can also be fed back to the auxiliary tasks to improve search diversity.

[0082] MFO is not directly equivalent to the surrogate model and does not completely replace the original problem. Instead, it optimizes through multiple problem perspectives simultaneously to achieve the effects of reducing complexity and improving efficiency.

[0083] A simulation-optimization method for a groundwater pumping and treatment pollution remediation system in a contaminated site based on the multi-form evolutionary search paradigm, including:

[0084] Construct a groundwater flow model for the contaminated site. On this basis, set the initial pumping and injection well scheme and establish a solute transport model for groundwater pollutants in the contaminated site;

[0085] Determine the management objective function and constraint conditions for the hydraulic control of the groundwater pollution range in the contaminated site, establish a groundwater pollutant treatment model with pumping and injection water treatment as the core, optimize the groundwater pumping and injection treatment in the contaminated site using a two-layer multi-objective optimization framework, solve the interdependence problem between the upper-level global objective and the lower-level local objective through nested optimization, and construct an auxiliary task using a multi-form evolutionary search paradigm; the upper-level global objective, i.e., the management objective function, includes minimizing the remediation cost and maximizing the pollutant removal efficiency; the lower-level local objectives include the pumping and injection water volume and the operation time of each well.

[0086] Update the state variables through the groundwater flow model and the solute transport model of groundwater pollutants in the contaminated site, calculate the value of the management objective function of the groundwater pollutant treatment model, and determine whether the constraint conditions are satisfied; at the same time, select the decision variables by optimizing the groundwater pollutant treatment model and return to the groundwater flow model and the solute transport model of groundwater pollutants in the contaminated site to update the state variables.

[0087] See Figure 1 , a simulation-optimization method for a groundwater pumping-injection treatment pollution remediation system in a chlorinated hydrocarbon contaminated site based on a multi-form evolutionary search paradigm, which is based on a groundwater solute transport simulation model and aims to design the most efficient groundwater pollutant extraction and treatment system. Taking the groundwater chlorobenzene contaminated site in a chemical industrial park as an example, the specific implementation steps of this method are described as follows:

[0088] Step (1), statistically analyze the borehole stratigraphic data and the distribution data of pollutants in the groundwater. Using the MODFLOW module of GMS software, determine the aquifer structure, source-sink terms, and boundary conditions in the study area, construct the groundwater flow model of the study area, and adjust the groundwater flow model parameters to the optimal through groundwater observation data. The permeability coefficient values of each stratum range from 1×10 -5 -1.1×10 -3 cm / s, and the porosity is 0.3. On this basis, set an initial pumping and injection well scheme, with 23 pumping wells and 10 injection wells. The maximum pumping and injection volume of the injection wells and pumping wells is taken. The pumping well is 50 m 3 / d, and the injection well is 100 m 3 / d, and use the MT3DMS module to establish the solute transport model of pollutants in the study area. Set the dispersion coefficient in the vertical direction to 0.076, while the dispersion coefficient in the horizontal direction is determined to be 0.76, and the simulation remediation period is 6 months.

[0089] Step (2): Determine the management objective function and constraint conditions for hydraulic control of the groundwater pollution range at the site. In the pumping and injection water treatment system, decision variables usually involve the following aspects: the pumping and injection rates of pumping and injection wells, the location and number of wells, the operation time and scheduling, and determine the specific time sequence of pumping, injecting, and treating water, etc. In this case, the optimized decision variable is the pumping and injection volume of the pumping and injection wells, the state variables are the groundwater level and pollutant concentration values in the site, and a mathematical model of the groundwater pollutant treatment model with pumping and injection water treatment as the core is established.

[0090] ;

[0091] ;

[0092] ;

[0093] ;

[0094] Where: is the remediation cost; represents minimizing the remediation cost; is the pumping and injection cost coefficient of the pumping and injection well, set to 0.15 yuan / (m 3 ·m); is the th pumping and injection volume of the 3 th pumping and injection well (unit: m / d); is the continuous pumping and injection time of the th pumping and injection well (unit: d); is the number of pumping and injection wells controlled by the pumping and injection - treatment pollution remediation system; is the number of iterative optimizations, is the percentage of the remaining pollutant mass in groundwater (unit: kg) after the th iterative optimization to the initial pollutant mass in groundwater (unit: kg); represents iterative optimization to maximize the pollutant removal efficiency; is the pollutant concentration in the groundwater of the polluted site after the th iterative optimization; is the maximum allowable concentration value of pollutants in the groundwater of the polluted site; is the area of pollutants exceeding the standard in the groundwater of the polluted site after pumping and injection treatment and remediation;

[0095] Step (3): To optimize the treatment of the pumped and injected groundwater in the chlorinated hydrocarbon contaminated site, a bilayer multi-objective optimization framework (BLMFO) is used for the optimization calculation. The mutual dependence problem between the upper layer (global objectives: minimizing the remediation cost and maximizing the pollutant removal efficiency) and the lower layer (local objective: pumping and injection volume) is solved through nested optimization, and an auxiliary task is constructed using the multi-form evolutionary search paradigm (MFO) to simplify the complexity and improve the optimization efficiency.

[0096] The upper layer objectives are to minimize the remediation cost and maximize the pollutant removal efficiency. The decision variables include the positions, numbers, and operation times of the pumping and injection wells. After fixing the upper layer variables (such as the well positions), the pumping and injection volume is optimized to meet the constraints of the groundwater flow field and pollutant concentration (such as the pollutant concentration being lower than the safety threshold). The upper layer generates initial solutions (such as randomly generating 10 sets of pumping and injection well layout schemes), and each solution corresponds to an independent optimization in the lower layer (invoking the MT3DMS module to simulate the pollutant migration). The NSGA-II algorithm is used in the lower layer to optimize the pumping and injection volume, generating the optimal lower layer solution set (Pareto front) corresponding to the upper layer solution and feeding it back to the upper layer to calculate the global fitness value.

[0097] The NSGA-II algorithm initializes the population randomly, where each individual represents the pumping and injection volume vector group of all pumping and injection wells:

[0098] ;

[0099] ;

[0100] where, represents the total number of individuals in the population generated by NSGA-II, and the individual represents a pumping and injection scheme, .

[0101] BLMFO uses explicit knowledge transfer (KT) to perform data interaction between the two auxiliary tasks and the original problem to enhance the optimization effect.

[0102] Meanwhile, an auxiliary task of the multi-form evolutionary search paradigm (MFO) is constructed:

[0103] Construct auxiliary task 1: First, ignore the lower layer optimization constraints, simplify the multi-objective problem, and regard the lower layer optimization as a penalty term added to the upper layer objective function.

[0104] ;

[0105] ;

[0106] where: represents the objective function of the auxiliary task 1; represents the upper-level objective function, that is, minimizing the remediation cost and maximizing the pollutant removal efficiency; is the remediation cost; is the percentage of the remaining pollutant mass in the groundwater after the pumping and injection treatment to the initial pollutant mass in the groundwater; represents the decision variable vector; is the penalty coefficient, used to control the penalty degree for violating the constraints. It can be set to 0.1 in the initial stage, increased by 0.1 every 50 generations, and finally stabilized at 1.0; represents the lower-level constraint violation degree, , is the single-well pumping and injection water volume constraint, is the hydraulic gradient constraint.

[0107] Directly optimize the upper-level objective, ignoring the lower-level independence, and quickly screen out low-cost pumping and injection well layout schemes through this task (set fewer wells while meeting the removal efficiency, but the feasibility needs to be verified later).

[0108] Construct auxiliary task 2: Integrate the lower-level objectives (such as the relationship between pumping and injection water volume and pollutant concentration) into the upper level to form a single-layer multi-objective optimization problem. Through this task, discover the trade-off relationship between pumping and injection water volume and pollutant removal efficiency, and provide diversified choices for decision-makers.

[0109] The lower-level optimization problem is used to simulate pollutant migration to calculate the final residual amount of pollutants:

[0110] ;

[0111] where: is the pollutant degradation rate, is the hydrodynamic dispersion coefficient, is the groundwater flow velocity, and are respectively the second-order gradient term (diffusion term) and the first-order gradient term (convection term) of the pollutant concentration , represents the total number of grid cells after the spatial discretization of the polluted site, represents the label of the grid cell, , represents the optimal lower-level decision variable set under the given upper-level decision variable .

[0112] It is also necessary to set the cross-task knowledge transfer (KT) to be executed once after a certain number of iterations (10 times) to ensure the co-evolution of the main task and the auxiliary task. When selecting solutions, priority should be given to the solutions belonging to the Pareto front in the auxiliary task, and the fitness of the auxiliary solutions in the main task should be higher than the current population mean. The solutions are screened by the Crowding Distance to avoid population aggregation (such as ensuring that the layout of injection and production wells covers different sub-regions of the polluted area). When the solutions of the auxiliary task are transferred to the main task, the lower-layer variables (such as injection and production water volumes) need to be supplemented, and the extension is completed by randomly extracting the lower-layer parameters of the historical optimal solutions. The feasible solutions of the main task are fed back to the auxiliary task to guide its search direction (such as initializing the auxiliary population with the verified well location scheme).

[0113] The algorithm terminates when the following conditions are met:

[0114] 1. The number of generations of iteration reaches the set upper limit (such as 100 generations).

[0115] 2. The non-dominated solutions converge to a stable Pareto front.

[0116] The main purpose of constructing the auxiliary task 1 is to reduce the computational complexity by expanding the search space, but it may not fully meet the lower-layer optimal solutions. Constructing the auxiliary task 2 directly converts the original two-layer multi-objective optimization problem into a single-layer multi-objective optimization problem, reducing the computational complexity but increasing the number of objectives and the optimization difficulty. During the optimization solution process, the auxiliary task 1, the auxiliary task 2, and the knowledge transfer mechanism are parallel, ultimately improving the quality of optimization and the solution efficiency.

[0117] Step (4): Couple the simulation model of the reaction and transport of groundwater flow and pollution components in the polluted area with the optimization management model of hydraulic control: The groundwater level and the pollutant concentration of each grid are used as state variables, and the pumping flow rates of each well are used as decision variables. The state variables are continuously updated through the simulation model constructed in step (1), the objective function value of the optimization management model constructed in step (2) is calculated, and it is judged whether the constraint conditions are met; at the same time, the decision variables are selected through the optimization management model and returned to the simulation model to update the state variables.

[0118] Step (5): Select the extraction and treatment plan corresponding to the optimal solution obtained in step (4), substitute it into the numerical simulation model, and predict the pollutant control effect in the polluted area. Combining the simulation results, verify the reliability of the optimized solution, and adjust the layout of production wells and the injection and production strategy according to the actual situation to form a more scientific and effective remediation plan.

[0119] See Figure 2, select a remediation plan among the Pareto frontiers. In the figure, from left to right and top to bottom, they are marked as Plan 1, Plan 2, Plan 3, Plan 4, Plan 5, and Plan 6 in sequence. Through analysis, it can be seen that Plan 1 and Plan 2 have low pumping volume and high remaining pollution volume. The advantage is that the cost may be relatively low, which is suitable for situations with limited budgets or occasions with lower requirements for environmental interference. The disadvantages are relatively obvious, with a large amount of remaining pollutants, unsatisfactory remediation effects, and potentially greater long-term environmental and health risks. Plan 3 and Plan 4 represent medium pumping volume and medium remaining pollution volume, achieving a balance between cost and remediation effect, and are suitable for scenarios seeking a medium level of treatment. However, it may take a longer remediation time to achieve the desired pollutant removal effect. Plan 5 and Plan 6 have the best remediation effect and the least remaining pollution volume, and are most suitable for occasions where pollution needs to be reduced urgently to meet safety standards. At the same time, the cost is very high, and the requirements for facilities are also higher, which is not suitable for projects with limited budgets. These six plans represent different trade-off points, and which one to choose depends on which goal is more important and the degree of compromise that one is willing to accept between the two.

[0120] A computer device includes a memory and a processor. The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory. When the computer program runs, it causes the processor to execute the above-mentioned simulation-optimization method for the groundwater pumping-injection-treatment pollution remediation system of chlorinated hydrocarbon contaminated sites based on the multi-form evolutionary search paradigm.

[0121] A computer-readable storage medium stores a program or instructions. When the program or instructions are executed by a computer device, it causes the computer device to execute the above-mentioned simulation-optimization method for the groundwater pumping-injection-treatment pollution remediation system of chlorinated hydrocarbon contaminated sites based on the multi-form evolutionary search paradigm.

[0122] In addition, it should be understood that after reading the above description of the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the appended claims of this application.

Claims

1. A simulation-optimization method for groundwater pumping-treatment pollution remediation system in contaminated sites based on a multi-form evolutionary search paradigm, characterized in that: include: Construct a groundwater flow model for the contaminated site, set up an initial pumping and injection well plan based on it, and establish a solute transport model for groundwater pollutants in the contaminated site; Determine the management objective function and constraints of hydraulic control of groundwater contamination range at contaminated sites, establish a groundwater pollutant treatment model with pumping and injection water treatment as the core, use a two-layer multi-objective optimization framework to optimize groundwater pumping and injection treatment at contaminated sites, solve the interdependence problem between the upper global objective and the lower local objective through nested optimization, and use a multi-form evolutionary search paradigm to construct auxiliary tasks; The upper-level global objectives, i.e., the management objective function, include minimizing the cost of remediation and maximizing the efficiency of pollutant removal; Lower local objectives include the amount of water pumped and injected and the operating time of each well; The state variables are updated through the groundwater flow model of the contaminated site and the solute transport model of groundwater pollutants in the contaminated site, the management objective function value of the groundwater pollutant treatment model is calculated, and it is determined whether the constraints are met; at the same time, the decision variables are selected by optimizing the groundwater pollutant treatment model, and the state variables are returned to the groundwater flow model of the contaminated site and the solute transport model of groundwater pollutants in the contaminated site to update.

2. The method for simulating and optimizing the contaminated site groundwater pumping-treatment pollution remediation system based on a multi-form evolutionary search paradigm according to claim 1 is characterized in that: The construction of the groundwater flow model of the contaminated site includes: collecting data on the borehole formation of the contaminated site and the distribution data of pollutants in the groundwater, using the MODFLOW module of the GMS software to determine the aquifer structure, source and sink items and boundary conditions of the contaminated site, constructing the groundwater flow model of the contaminated site, and adjusting the parameters of the groundwater flow model to the optimum through groundwater observation data; The MT3DMS module is used to establish a solute transport model of groundwater pollutants in contaminated sites.

3. The method for simulating and optimizing the groundwater pumping and treatment pollution remediation system of a contaminated site based on a multi-form evolutionary search paradigm according to claim 1 is characterized in that: The state variables in the contaminated site groundwater pumping and injection-treatment pollution remediation system are the groundwater level and the pollutant concentration in the contaminated site. The decision variables include any one or more of the following: the amount of water pumped and injected from the pumping and injection wells, the location and number of the pumping and injection wells, the operating time and scheduling, which determine the specific timing of pumping and injection.

4. The method for simulating and optimizing the groundwater pumping and treatment pollution remediation system of a contaminated site based on a multi-form evolutionary search paradigm according to claim 3 is characterized in that: The decision variables include the water injection volume of the pumping and injection wells, and the following groundwater pollutant treatment model is established: ； ； ； ； in: is the cost of repair; represents minimizing the cost of repair; is the pumping and injection cost coefficient of the pumping and injection well; It is The water extraction and injection volume of the pumping and injection wells; It is The continuous pumping and injection time of the pumping and injection wells; It is the pump-injection-treatment pollution remediation system that controls the number of pump-injection wells; is the number of iterative optimizations, It is The quality of residual pollutants in groundwater after the first iteration optimization Initial contaminant quality in groundwater percentage of It represents iterative optimization to maximize pollutant removal efficiency; It is The concentration of pollutants in groundwater at the contaminated site after the first iteration of optimization; is the maximum permissible concentration of pollutants in groundwater at a contaminated site; It is the area of ​​pollutants in groundwater where the concentration exceeds the standard after pumping and injection treatment and remediation; It is the initial area of ​​the polluted site where the groundwater concentration exceeds the standard pollutant concentration; Optimize pumping and injection volumes to meet groundwater flow field and pollutant concentration constraints.

5. The method for simulating and optimizing the groundwater pumping and treatment pollution remediation system of a contaminated site based on a multi-form evolutionary search paradigm according to claim 1 is characterized in that: The upper layer generates multiple groups of initial pumping and injection well schemes as initial solutions. Each solution of the upper layer, i.e., the lower layer corresponding to each group of pumping and injection well schemes, is optimized independently, and the MT3DMS module is called to simulate the migration of groundwater pollutants. The lower layer optimizes the water pumping and injection volume, generates the optimal lower layer solution set corresponding to the upper layer solution, and feeds it back to the upper layer to calculate the global fitness value.

6. The method for simulating and optimizing the groundwater pumping and treatment pollution remediation system of a contaminated site based on a multi-form evolutionary search paradigm according to claim 5 is characterized in that: The lower layer uses the NSGA-II algorithm to optimize the water pumping and injection volume. The NSGA-II algorithm randomly initializes the population Individual, each of which Represents the water extraction and injection volume vector group of all water extraction and injection wells: ； ； in, Represents the total number of individuals in the population generated by NSGA-II. represents a pumping scheme, .

7. The method for simulating and optimizing the groundwater pumping and treatment pollution remediation system of a contaminated site based on a multi-form evolutionary search paradigm according to claim 1, characterized in that: The auxiliary tasks constructed by using the multi-form evolutionary search paradigm include: Construct auxiliary task 1: Ignore the lower-level optimization constraints, simplify the multi-objective problem, and treat the lower-level optimization as a penalty term and add it to the upper-level objective function: ； ； in: represents the objective function of auxiliary task 1; represents the upper objective function, i.e., minimizing the remediation cost and maximizing the pollutant removal efficiency; is the cost of repair; It is the percentage of the mass of residual pollutants in groundwater after pumping and injection treatment and remediation to the mass of initial pollutants in groundwater; represents the decision variable vector; is the penalty coefficient; represents the violation degree of the lower constraint, , is the single well pumping and injection volume constraint, is the hydraulic gradient constraint; Construct auxiliary task 2: Integrate the lower-level local objectives into the upper-level to form a single-level multi-objective optimization problem. Through this task, the trade-off between the amount of pumping and injection water and the efficiency of groundwater pollutant removal is discovered; the lower-level objectives include the relationship between the amount of pumping and injection water and the concentration of groundwater pollutants; The lower optimization problem is used to simulate the migration of groundwater pollutants to calculate the final residual amount of groundwater pollutants: ； in: is the pollutant degradation rate, is the hydrodynamic dispersion coefficient, is the groundwater flow velocity, and The pollutant concentrations are The second-order gradient term and the first-order gradient term of represents the total number of grid cells after the contaminated site is discretized. represents the grid cell number, , Indicates that given the upper decision variable The optimal set of lower-level decision variables under ; The two-layer multi-objective optimization framework uses explicit knowledge transfer to perform data interaction between the two auxiliary tasks and the original problem to enhance the optimization effect. That is, after a certain number of iterations, an explicit knowledge transfer across tasks is performed to ensure the co-evolution of the main task and the auxiliary task.

8. The method for simulating and optimizing the groundwater pumping and treatment pollution remediation system of a contaminated site based on a multi-form evolutionary search paradigm according to claim 1 or 7, characterized in that: Prioritize solutions that belong to the Pareto front in the auxiliary task. The fitness of the solution of the auxiliary task in the main task must be higher than the current population mean. The solution is screened by crowding distance to avoid population aggregation and ensure that the layout of the pumping and injection wells covers different sub-areas of the contaminated site. When the solution of the auxiliary task is transferred to the main task, the lower-level decision variables including the amount of water pumped and injected are supplemented, and the expansion is completed by randomly extracting the lower-level parameters of the historical optimal solution; The feasible solution of the main task is fed back to the auxiliary task to guide its search direction, including initializing the auxiliary population using the verified well location scheme.

9. A computer device comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, wherein: When the computer program is running, the processor executes the simulation-optimization method for groundwater pumping-treatment pollution remediation system of contaminated sites based on a multi-form evolutionary search paradigm as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores programs or instructions, and when the programs or instructions are executed by a computer device, the computer device executes the simulation-optimization method for groundwater pumping-treatment pollution remediation system of contaminated sites based on a multi-form evolutionary search paradigm as described in any one of claims 1-8.

Images