Method for regulating activity and distribution of complex stratum microbial agent for repairing underground pollution

Abstract

This application provides a method for regulating the activity and distribution of microbial agents used to remediate underground contamination in complex formations. The method includes: collecting microbial agent samples and monitoring parameters such as temperature, humidity, and oxygen levels in real time; using a genetic algorithm to optimize the combination of environmental variables to obtain an activity maintenance threshold; employing a microbial agent survival model to construct a three-dimensional mesh representation based on formation scanning data; using finite element analysis to analyze permeability differences and determine complex formation zones; based on a uniform distribution scheme, using real-time sensor feedback of injection process data to determine distribution path deviation; if the deviation exceeds a preset value, dynamically adjusting the injection point position to obtain a precise penetration trajectory; using the precise penetration trajectory to map the coverage area to the pollution source location; verifying the coverage integrity through Monte Carlo simulation to determine the pollution coverage mesh.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for regulating the activity and distribution of microbial agents used in the remediation of underground contamination in complex formations. Background Technology

[0002] In the field of environmental governance, underground pollution remediation is a crucial task, directly related to the protection of soil and groundwater resources and the safety of human health.

[0003] With the acceleration of industrialization, underground pollution has become increasingly serious, especially heavy metal pollution, such as chromium pollution, which poses a huge threat to ecosystems and public safety.

[0004] How to efficiently and accurately remediate underground pollution has become a major issue that urgently needs to be addressed in the field of environmental science and engineering.

[0005] Currently, although many remediation methods can address underground contamination to some extent, they generally suffer from insufficient adaptability and operability.

[0006] Many technologies often struggle to cope with the differences between different regions when faced with complex geological environments, resulting in uneven repair effects or even complete failure in some key areas.

[0007] In addition, the active substances used in the repair process, such as microbial agents, often lose their activity due to limitations in transportation or storage conditions, making it difficult to play their due role in practical applications.

[0008] This limitation makes restoration efforts inadequate when faced with large-scale, complex environments.

[0009] Against this backdrop, the core technical challenges facing underground pollution remediation are gradually becoming apparent.

[0010] A key factor is maintaining the vitality of the active microbial agent. From production to application, the microbial agent is easily affected by changes in the external environment, which can lead to a significant decrease in its activity and thus make it unable to effectively degrade pollutants.

[0011] Another related factor is the complexity of the geological environment. Differences in permeability in different areas can hinder the uniform distribution of microbial agents, especially in low-permeability areas where microbial agents have difficulty reaching the core of the contamination, resulting in incomplete remediation coverage.

[0012] For example, in practice, in some contaminated areas, the presence of obstruction layers in the strata may prevent the microbial agent from flowing only on the surface and from penetrating into the most severely contaminated deep areas, thus greatly reducing the remediation effect.

[0013] Therefore, how to ensure the efficient survival of active bacterial agents from production to application under complex geological conditions, and how to ensure that they can accurately penetrate different permeability areas and fully cover the contaminated area, has become a key issue in the field of underground pollution remediation.

[0014] Instructions for Bacterial Agents and Packaging Materials

[0015] The microbial agents used in this invention are selected according to the type of underground pollution. Dechlorinating bacteria are used for chlorinated hydrocarbon pollution, chromium-reducing bacteria are used for heavy metal pollution such as chromium, and degrading bacteria are used for polycyclic aromatic hydrocarbon pollution. The suitable survival environment for the microbial agents is a temperature of 20-30℃ and a humidity of 60-70%. The survival density of the microbial agents is not less than 10^6 cells / mL under anaerobic conditions and not less than 10^7 cells / mL under aerobic conditions. The bacterial agent encapsulation material is a compound of sodium alginate-calcium chloride composite system and polyvinyl alcohol-based composite material. The basic ratio is sodium alginate:calcium chloride:polyvinyl alcohol = 4:1:5 (mass ratio). The basic thickness of the sodium alginate curing layer is 1mm. When the soil pH value is <4.5 or >8, the neutralizing agent component is increased (mass ratio 5%), and the curing layer thickness is increased by 2mm. When the heavy metal ion content is severely exceeded, the antioxidant component is increased (mass ratio 3%), and the polyvinyl alcohol content is increased to 60%. The oxygen permeability of the encapsulation material is ≥0.2, and the water permeability is adapted to the soil moisture requirements. Summary of the Invention

[0016] This invention provides a method for regulating the activity and distribution of microbial agents used in the remediation of underground contamination in complex formations, mainly comprising:

[0017] By collecting microbial agent samples and monitoring parameters such as temperature, humidity, and oxygen levels in real time, a genetic algorithm is used to optimize the combination of environmental variables to obtain an activity maintenance threshold. Based on the activity maintenance threshold, underground stratum scanning data, including soil type and porosity distribution, is obtained to determine the degree of environmental impact. If the environmental impact exceeds the preset threshold, the parameters of the microbial agent encapsulation material are adjusted to obtain a microbial agent survival model. Using the microbial agent survival model, a three-dimensional mesh representation is constructed based on the stratum scanning data. Finite element analysis is used to analyze permeability differences and determine complex stratum zones. For these complex stratum zones, a neural network is used to predict the microbial agent flow path and determine whether permeability differences cause flow stagnation. If stagnation areas exist, the injection pressure gradient is optimized to obtain a uniform distribution scheme. Based on the uniform distribution scheme, real-time sensor feedback of injection process data is used to determine the distribution path deviation. If the deviation is greater than a preset value, the injection point position is dynamically adjusted to obtain a precise penetration trajectory. Using the precise penetration trajectory, the coverage area is mapped to the pollution source location. Monte Carlo simulation is used to verify the coverage integrity and determine the pollution coverage mesh. The pollution coverage grid is obtained and fed back to the initial activity maintenance threshold for iterative iteration. The overall remediation parameters are then determined to converge. If they do not converge, the combination of environmental variables is reset to obtain the final distribution optimization result.

[0018] The method for regulating the activity and distribution of microbial agents for remediation of underground contamination in complex formations, as described in this invention, is applicable to various types of organic / inorganic underground contamination, including chlorinated hydrocarbons, chromium, lead, and polycyclic aromatic hydrocarbons. It is adaptable to various complex formations such as sand-pebble-clay interbedded formations, karst formations, and silty formations, covering a contamination depth range of 5-50 meters underground. The equipment requirements for engineering implementation are: a servo-controlled injection system, a ground-penetrating radar MALA ProEx, a 3D seismic survey instrument, an electrochemical / thermometer / capacitive sensor array, and a conventional data processing server. There is no need for special high-end equipment. The implementation cost can be reduced by optimizing the sensor layout (arranged according to the gradient of the contamination area). Compared with traditional remediation methods, the equipment cost is reduced by approximately 20%.

[0019] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0020] This invention discloses a method for regulating the activity and distribution of microbial agents in complex formations for the remediation of underground contamination. Addressing the core issues of maintaining microbial agent activity, ensuring uniform distribution, and maintaining the integrity of the contamination cover in underground contamination remediation, this method achieves precise regulation through multi-dimensional environmental parameter monitoring and optimization. The invention collects parameters such as temperature, humidity, and oxygen in real time, combines this with a genetic algorithm to optimize environmental variables, determines the threshold for maintaining microbial agent activity, and adjusts the encapsulation material parameters based on formation scanning data to construct a microbial agent survival model. Simultaneously, it uses 3D mesh and finite element simulation to analyze formation permeability differences, employs neural networks to predict flow paths, and optimizes the injection pressure gradient to ensure uniform distribution. Finally, Monte Carlo simulation is used to verify the integrity of the contamination cover, and remediation parameters are dynamically iterated to achieve distribution optimization. The most critical innovation of this invention lies in combining microbial agent activity regulation with formation complexity distribution optimization. Through multi-model collaboration and a real-time feedback mechanism, it significantly improves the survival rate of microbial agents and the efficiency of contamination remediation in complex formations, providing systematic and efficient technical support for underground contamination control. Attached Figure Description

[0021] Figure 1 This is a flowchart of the activity and distribution regulation method of the present invention.

[0022] Figure 2 This is a schematic diagram of the activity and distribution regulation method of the present invention.

[0023] Figure 3 This is another schematic diagram of the activity and distribution regulation method of the present invention.

[0024] Figure 4 This is a topology diagram of the multi-model collaborative network of the present invention.

[0025] Figure 5 This is the state diagram of the dual optimization coupling of activity and distribution in this invention.

[0026] Figure 6 This is a diagram of the parameter convergence closed-loop iterative mechanism of the present invention.

[0027] Figure 7 This is a deployment diagram of the underground stratum microbial agent injection system of the present invention.

[0028] Figure 8 This is a schematic diagram of the vertical cross-section of the strata showing the underground infiltration zone and the flow path of the microbial agent according to the present invention.

[0029] Figure 9 This is a curve showing the iterative optimization of the survival rate of the bacterial agent of this invention.

[0030] Figure 10 This is a thermal map showing the permeability distribution of different stratigraphic zones according to the present invention.

[0031] Figure 11 This is a composite comparison chart of the repair efficiency and coverage integrity of the present invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0033] Example 1: Remediation of Chlorinated Hydrocarbon Contaminated Site at a Chemical Plant

[0034] For the chlorinated hydrocarbon contaminated area 15-25 meters underground of a chemical plant, the strata exhibit a typical heterogeneous structure of interbedded sand, gravel and clay.

[0035] First, real-time geological parameters of the area were obtained through sensors, and monitoring showed that the groundwater flow rate was extremely slow.

[0036] Genetic algorithms were used to optimize and determine the activity maintenance threshold of the bacterial agent under the current anaerobic environment (e.g., the survival density of dechlorinating bacteria is not less than 10^6 cells / mL).

[0037] For the formation scanning data (porosity 0.25-0.35), a three-dimensional finite element model is constructed. If the simulation finds that there is penetration resistance in the clay layer, the injection pressure is optimized by neural network, and the injection pressure is adjusted from atmospheric pressure to pulse pressure (fluctuation of 0.5MPa-0.8MPa).

[0038] Based on the feedback of the distribution path deviation, the injection well point will be finely adjusted by 2-3 meters in real time to ensure that the bacterial agent can accurately penetrate the sand and gravel layer and spread to the edge of the clay layer.

[0039] Ultimately, Monte Carlo simulations verified that the pollution plume coverage in this complex geological formation increased from the traditional 45% to 88%, and the remediation cycle was shortened by approximately 30%.

[0040] Example 2: Remediation of Karst Formations at a Chromium-Contaminated Site

[0041] The chromium contamination area located 8-15 meters underground in a metallurgical plant is a typical karst formation with a heterogeneous structure featuring caves and fissures. The groundwater flow rate is 0.05 m / d, the soil pH is 3.8, the porosity is 0.18-0.25, and the chromium ion concentration exceeds the standard by 2 times.

[0042] First, real-time geological parameters of the area were collected using a sensor array. Then, a genetic algorithm was used to optimize the data and determine the activity maintenance threshold of chromium-reducing bacteria under the current anaerobic environment as a survival density of not less than 10^6 cells / mL. The suitable environmental parameters were a temperature of 22-28℃ and a humidity of 65-70%.

[0043] Based on the formation scanning data, due to the high level of environmental impact, the parameters of the bacterial agent encapsulation material were adjusted: the neutralizing agent content was increased (5% by mass), the thickness of the sodium alginate curing layer was increased to 3mm, and a bacterial agent survival model was constructed.

[0044] A three-dimensional mesh model was generated using the Delaunay triangulation algorithm. Permeability differences were analyzed through finite element simulation to determine complex strata zones. Cave areas were marked as high-permeability zones, and fractured areas were marked as blocking zones.

[0045] By predicting the flow path of the inoculant using a convolutional neural network, the injection pressure gradient in the blocked region is optimized into a pulsed pattern (fluctuation of 0.6-0.9 MPa, with a period of 10 seconds) to obtain a uniform distribution scheme.

[0046] Based on the distribution path deviation fed back by real-time sensors, the injection well point is finely adjusted by 0.3 meters in the vertical Z-axis direction and 0.2 meters in the horizontal X-axis direction to obtain a precise penetration trajectory;

[0047] Ultimately, Monte Carlo simulations verified that the pollution plume coverage under the karst strata increased from the traditional 38% to 85%, the bacterial agent survival rate remained above 90%, the remediation cycle was shortened by about 25%, and the chromium ion degradation rate reached over 80%.

[0048] like Figure 1As shown, the activity and distribution control method of the present invention includes steps S101 to S107. First, in step S101, parameters such as temperature, humidity, and oxygen levels are monitored in real time by collecting bacterial agent samples. A genetic algorithm is used to optimize the combination of environmental variables to obtain an activity maintenance threshold. In step S102, underground stratum scanning data, including soil type and porosity distribution, is obtained based on the activity maintenance threshold. The degree of environmental influence is assessed. If the environmental influence exceeds a preset threshold, the parameters of the bacterial agent encapsulation material are adjusted to obtain a bacterial agent survival model. In step S103, using the bacterial agent survival model, a three-dimensional mesh representation is constructed based on the stratum scanning data. Permeability differences are analyzed through finite element simulation to determine complex stratum zones. In step S104, the complex stratum zones are obtained, and a neural network is used to predict the bacterial agent flow path. It is determined whether permeability differences cause flow stagnation. If stagnation areas exist, the injection pressure gradient is optimized to obtain a uniform distribution scheme. In step S105, based on the uniform distribution scheme, real-time sensor feedback of injection process data is used to determine the distribution path deviation. If the deviation is greater than a preset value, the injection point position is dynamically adjusted to obtain a precise penetration trajectory. Step S106 employs a precise penetration trajectory to map the coverage area based on the pollution source location, verifies the coverage integrity through Monte Carlo simulation, and determines the pollution coverage grid. Step S107 acquires the pollution coverage grid, feeds it back to the initial activity maintenance threshold for iterative iteration, and determines whether the overall remediation parameters converge. If converged, the final distribution optimization result is output; if not, the environmental variable combination is reset and the process returns to step S101 for re-iteration.

[0049] like Figure 2 As shown, the system architecture of the activity and distribution regulation method of the present invention includes four main functional areas. The environmental monitoring and activity optimization area includes an environmental parameter monitoring module and a genetic algorithm optimization module. The environmental parameter monitoring module collects data through temperature, humidity, and oxygen sensor arrays, while the genetic algorithm optimization module optimizes the combination of environmental variables and outputs the activity maintenance threshold. The stratigraphic analysis and survival modeling area includes a stratigraphic scanning data acquisition module, a fungal agent encapsulation material adjustment module, and a fungal agent survival model construction module. The stratigraphic scanning data acquisition module acquires soil type and porosity distribution data, while the fungal agent survival model construction module uses the SVM algorithm to establish a survival model. The flow prediction and injection optimization area includes a 3D mesh construction and finite element simulation module, a stratigraphic complex zoning module, a CNN flow path prediction module, and an injection pressure gradient optimization module. Each module sequentially processes permeability analysis, zoning data, and flow path information. The coverage verification and iterative convergence area includes a real-time sensor feedback module, a precise penetration trajectory generation module, a Monte Carlo coverage verification module, and an iterative convergence judgment module. If convergence fails, the iterative convergence judgment module transmits the feedback signal to the environmental parameter monitoring module for re-iteration.

[0050] like Figure 3As shown, the detailed sub-module interactions of the activity and distribution regulation method of this invention are divided into three levels. The upper level is the activity regulation layer, which includes five sub-modules: environmental parameter acquisition, deviation sequence calculation, sliding window feature extraction, genetic algorithm optimization, and activity maintenance threshold determination. These sub-modules are sequentially connected to complete the entire activity regulation process. The middle level is the distribution optimization layer, which includes six sub-modules: formation scanning, 3D mesh construction, finite element permeability analysis, CNN flow prediction, injection pressure optimization, and sensor feedback adjustment. The activity threshold output by the activity regulation layer serves as the input parameter for the distribution optimization layer. The lower level is the verification iteration layer, which includes six sub-modules: accurate penetration trajectory, pollution source coverage mapping, Monte Carlo simulation, coverage integrity determination, iterative convergence, and final optimization results. The penetration data from the distribution optimization layer is passed to the verification iteration layer for coverage verification. There are cross-data flow connections between the three levels. The threshold data from the activity regulation layer is passed down to the distribution optimization layer, and the flow prediction data and penetration data from the distribution optimization layer are passed down to the verification iteration layer. If the verification iteration layer fails to converge, it sends a feedback signal back to the environmental parameter acquisition sub-module of the activity regulation layer to initiate a new round of iteration.

[0051] like Figure 4 As shown, this invention employs a multi-model collaborative network architecture to achieve the regulation of activity and distribution of microbial agents for remediation in complex formations. This network uses the Finite Element Method (FEM) as its core hub node, forming a complete data transfer topology with a Genetic Algorithm (GA), Support Vector Machine (SVM), Convolutional Neural Network (CNN), Linear Regression (LR), and Monte Carlo Simulation (MC). Specifically, the Genetic Algorithm transmits the optimized activity maintenance threshold to the SVM; the SVM outputs the microbial agent survival model parameters to the Finite Element Method for formation mechanical analysis; the Finite Element Method transmits complex formation partition data to the CNN for feature extraction and partition identification; the CNN outputs the flow path prediction results to the CNN for accurate fitting; and the CNN transmits the accurate penetration trajectory to the Monte Carlo Simulation for probabilistic verification. After the Monte Carlo Simulation is completed, the iterative feedback information is fed back to the Genetic Algorithm, forming a convergence loop until the preset convergence threshold condition is met. This multi-model collaborative mechanism ensures the ability to search for the globally optimal solution for activity regulation and distribution optimization.

[0052] like Figure 6As shown, this invention employs a closed-loop iterative convergence mechanism to achieve global optimization of remediation parameters. Seven processing nodes are arranged sequentially in a ring: environmental variable optimization, encapsulation material adjustment, formation zoning analysis, flow path prediction, penetration trajectory correction, cover integrity verification, and remediation parameter evaluation. Each node transmits processing results clockwise. The outer ring of each node is labeled with its corresponding key algorithm, including Genetic Algorithm (GA), Support Vector Machine (SVM), Finite Element Method (FEM), Convolutional Neural Network (CNN), Linear Regression (LR), and Monte Carlo Simulation (MC). After the remediation parameter evaluation is completed, the system compares the current iteration parameter difference with a preset convergence threshold: if the parameter difference is less than the convergence threshold, the system is considered to have converged, and the final distribution optimization result is output; if the parameter difference is greater than or equal to the convergence threshold, the system returns to the environmental variable optimization node via a loop path to start a new round of iteration. This closed-loop mechanism ensures that the remediation scheme gradually approaches the global optimum in the multi-dimensional parameter space.

[0053] More specifically, an embodiment of the present invention provides a method for regulating the activity and distribution of microbial agents used in the remediation of underground contamination in complex formations, which may include:

[0054] Step S101 involves collecting bacterial agent samples and monitoring parameters such as temperature, humidity, and oxygen levels in real time. A genetic algorithm is then used to optimize the combination of environmental variables to obtain the activity maintenance threshold.

[0055] Real-time data collection of sample temperature, humidity, and oxygen levels using a sensor array yields a sequence of the three environmental parameters. The deviation between the current value and the historical mean of each parameter is calculated, resulting in a deviation sequence. A sliding window is used to extract local variation features from the deviation sequence, resulting in a local variation feature sequence. Temperature, humidity, and oxygen level are selected from this local variation feature sequence to form a feature vector for the current moment. This feature vector is then input into a genetic algorithm for optimization, iteratively searching for combinations of environmental parameters to obtain a candidate combination set. The activity maintenance level for each candidate combination is simulated and calculated at the corresponding moment, resulting in an activity maintenance level sequence for each combination. The activity maintenance level sequences for each combination are compared, and the combination with the highest level is selected to determine the activity maintenance threshold.

[0056] The initial population size of the genetic algorithm is 100 combinations of environmental parameters, with a crossover rate of 0.8, a mutation rate of 0.05, and 50 iterations. The fitness function is calculated with the bacterial activity maintenance rate as the core evaluation index, taking into account the environmental adaptability of temperature, humidity, and oxygen levels.

[0057] In one possible implementation, sample temperature, humidity, and oxygen levels are collected in real time via a sensor array. This can be understood as deploying multiple sensor nodes in a biological laboratory environment. For example, the temperature sensor uses a thermistor type, the humidity sensor uses a capacitive principle, and the oxygen sensor is based on an electrochemical method. These sensors collect data once per second, forming a sequence containing three types of parameters. For example, the current sequence is temperature 25.5°C, humidity 60%, and oxygen 21%. This yields the current environmental three-parameter sequence, which can be used for subsequent environmental optimization.

[0058] For example, in cell culture, the process of calculating the deviation of the current value of each parameter from the historical mean based on the current environmental three-parameter sequence first requires maintaining a historical data buffer. Assuming the historical mean is 24°C temperature, 55% humidity, and 20% oxygen, the current deviation is calculated as a temperature deviation of 1.5°C, a humidity deviation of 5%, and an oxygen deviation of 1%. The deviation sequence is obtained by subtraction, which helps to identify the impact of environmental fluctuations on cell viability.

[0059] In one possible implementation, a sliding window is used to extract local variation features for the deviation sequence. The sliding window is a time series analysis technique.

[0060] Specifically, it involves setting a fixed-size window, such as a window width of 10 time points, sliding it over the deviation sequence, and calculating statistical characteristics within the window each time, such as mean, variance, or slope. For the temperature deviation sequence, the local mean deviation of 0.8°C and variance of 0.2 are calculated within the window. This extracts the local variation feature sequence, reflecting the short-term trend changes of environmental parameters.

[0061] For example, three types of features—temperature, humidity, and oxygen level—can be selected from the local change feature sequence to form the feature vector at the current moment. The feature selection here can be based on correlation analysis, choosing the most representative ones such as the local mean and slope of temperature, the variance of humidity, and the peak value of oxygen, forming a multi-dimensional vector, such as [0.8,0.2,5.0,1.5,1.0,0.5]. This vector captures the environmental dynamics and is used as input for the optimization algorithm.

[0062] In one possible implementation, the feature vector at the current moment is input into the genetic algorithm optimization process. The genetic algorithm is a biomimetic optimization method that simulates natural selection. It iteratively searches through initialization of the population, selection, crossover, and mutation operations. For example, the initial population may have 100 combinations of environmental parameters, such as temperature range of 22-28°C, humidity of 50-70%, and oxygen of 18-22%. Based on the feature vector as the fitness function input, after 50 generations, a set of candidate combinations is obtained, such as 10 preferred combinations. This can efficiently search the parameter space for maintaining cell viability in cell culture.

[0063] For example, the activity maintenance level at each time point is simulated and calculated for each candidate combination set. The simulation calculation can be performed using a mathematical model, such as an activity decay model simplified based on the Arrhenius equation. For each combination input parameter, the activity level score is calculated. For example, combination one has a temperature of 26°C, humidity of 62%, and oxygen of 20.5%, corresponding to a level of 85%, thus obtaining the activity maintenance level sequence for each combination. This step ensures quantitative evaluation.

[0064] In one possible implementation, the activity maintenance level sequences of each combination are compared, and the combination with the highest level is selected to determine the activity maintenance threshold. For example, the combination with the highest level of 90% in the sequence corresponds to a temperature of 25.8°C, a humidity of 61%, and an oxygen level of 21.2%. The level of this combination is then set as the threshold of 80%. This can bring a stable cell activity maintenance effect in business and improve culture efficiency.

[0065] Step S102: Based on the activity maintenance threshold, obtain underground stratum scanning data including soil type and porosity distribution, determine the degree of environmental impact, and if the environmental impact exceeds the preset threshold, adjust the parameters of the bacterial agent encapsulation material to obtain the bacterial agent survival model.

[0066] Subsurface strata data, including soil type and pore distribution information, is acquired using scanning equipment and stored as an initial dataset. Based on this initial dataset, soil type and pore distribution characteristics are analyzed, and a classification method is used to conduct a preliminary assessment of the environmental impact, resulting in an environmental impact level. If the environmental impact level exceeds a preset threshold, specific data on soil type and pore distribution are extracted, and combined with activity maintenance requirements, it is determined whether the survival conditions of the microbial agent are met. Based on the assessment results, for cases where the microbial agent survival conditions are insufficient, the material parameters for the microbial agent encapsulation are adjusted to generate an optimized material configuration scheme. The optimized material configuration scheme is obtained, and combined with the environmental impact level, a microbial agent survival model is constructed. A support vector machine algorithm is used to predict the survival probability and determine the survival model parameters. Based on the survival model parameters, the activity maintenance state of the microbial agent in the subsurface strata is simulated to obtain the final activity distribution results.

[0067] The environmental impact level is divided into three levels: low, medium, and high, based on the degree of interference of the soil strata with the survival of the microbial agent. The low level is characterized by soil pH of 5-7, porosity ≥0.3, and no heavy metal ion exceedances. The medium level is characterized by soil pH of 4.5-5 or 7-8, porosity of 0.2-0.3, and slight exceedances of heavy metal ions (≤1 time of the national limit). The high level is characterized by soil pH <4.5 or >8, porosity <0.2, and severe exceedances of heavy metal ions (>1 time of the national limit). The preset threshold for environmental impact is medium level, meaning that when the environmental impact level reaches medium or above, it is judged as exceeding the preset threshold.

[0068] The support vector machine algorithm for constructing the microbial agent survival model uses the radial basis function as the kernel function, with a kernel parameter gamma of 0.1 and a penalty function coefficient C of 1.0. The training set samples are microbial agent survival rate data under different material configurations and different environmental influence levels. The model training termination condition is that the prediction accuracy of the training set is ≥90%.

[0069] Specifically, adjusting the parameters of the bacterial agent encapsulation material refers to dynamically adjusting the composition of the bacterial agent encapsulation carrier based on the degree of influence of the underground environment (such as pH value, heavy metal ion concentration, and salinity). For example, when the degree of environmental influence exceeds a preset threshold (such as pH value below 4.5), the system automatically increases the proportion of neutralizing agents or antioxidants in the encapsulation material, or increases the thickness of the sodium alginate solidification layer, to physically shield the penetration of harmful environmental factors and ensure that the bacterial agent maintains high biological activity before reaching the target contaminated area.

[0070] In one possible implementation, when acquiring underground strata data through scanning equipment, a ground-penetrating radar scanner can be used to detect underground structures.

[0071] For example, in an agricultural soil improvement project, scanning equipment emits electromagnetic wave signals that penetrate the soil layer and reflect echoes. Based on the intensity and time delay of the echoes, the soil type, such as sandy soil or clay, as well as the density and size of the pore distribution, are identified. This data is converted into digital format in real time and stored as an initial dataset containing information such as soil particle size distribution and pore volume ratio, thus providing a basis for subsequent analysis.

[0072] For example, when analyzing soil type and pore distribution characteristics based on an initial dataset, the classification method can be a decision tree-based algorithm. First, the soil type is classified into categories such as acidic soil, neutral soil, and alkaline soil. Then, the statistical characteristics of pore distribution, such as average pore diameter and connectivity index, are calculated. These characteristics are then input into the classification model to assess the environmental impact.

[0073] For example, uneven pore distribution may lead to uneven water penetration, thus the environmental impact level is initially assessed as low, medium and high, with the high level indicating the potential risk of loss of microbial agent activity.

[0074] In one possible implementation, when the environmental impact level exceeds a preset threshold, such as intermediate or above, specific data needs to be extracted. For example, the porosity data of sandy soil (0.35) and the pore connectivity index of clay (0.6) are extracted from the initial dataset. Combined with the activity maintenance requirements, such as the bacterial agent needing at least 0.2 oxygen permeability and a suitable pH range of 5-7, it is determined whether the survival conditions meet the standards. This judgment process involves comparing the actual data with the required threshold. If the porosity is lower than the requirement, it is considered insufficient, thereby triggering the adjustment mechanism.

[0075] For example, in cases where the conditions for the survival of the microbial agent are insufficient, adjusting the material parameters of the microbial agent encapsulation can include changing the permeability of the encapsulation film.

[0076] For example, the original polymer film thickness was adjusted from 0.1 mm to 0.05 mm, and hydrophilic additives were added to improve water permeability, resulting in an optimized material configuration scheme, such as using polyvinyl alcohol-based composite materials, to ensure that the bacterial agent can maintain its activity under high environmental influences. This scheme was verified through simulation tests to check its effect on oxygen and humidity control.

[0077] In one possible implementation, after obtaining the optimized material configuration scheme, when constructing the microbial agent survival model in conjunction with the environmental impact level, the support vector machine algorithm is used to predict the survival probability. First, historical data is collected as a training set, including survival rate samples under different material configurations. Then, the features are mapped to a high-dimensional space through kernel functions such as radial basis functions, and the hyperplane is optimized to classify whether the agent survives or not. For example, if the input feature vector contains a material permeability of 0.4 and a high environmental impact level, the algorithm calculates the support vector and predicts a survival probability of 0.75, thereby determining the model parameters such as the penalty function coefficient C of 1.0 and the kernel parameter gamma of 0.1. These parameters reflect the robustness of the model to noise.

[0078] For example, when simulating the activity maintenance state of microbial agents in underground strata based on survival model parameters, finite element simulation software can be used to input parameters, distribute microbial agent points in a virtual underground model, and calculate the activity distribution. For example, the activity level at a depth of 2 meters is 85%. The final activity distribution results show that the maintenance rate is high in the central area and low at the edge. This helps to optimize the microbial agent application strategy and improve soil remediation efficiency in actual agricultural applications.

[0079] Step S103: Using a microbial agent survival model, a three-dimensional mesh representation is constructed based on the formation scanning data. The permeability difference is analyzed through finite element simulation to determine the complex zoning of the formation.

[0080] Acquire formation scanning data. Generate a 3D mesh model using a mesh generation algorithm. Calculate the survival probability of each mesh cell using a fungal agent survival model. Simulate the permeability distribution of each mesh cell using the finite element method. Obtain the permeability value of each mesh cell based on the permeability simulation results. Determine the difference in permeability values ​​between adjacent mesh cells; if the difference exceeds a preset threshold, it is marked as a partition boundary. Merge mesh cells with similar permeability through connectivity analysis to determine complex formation partitions.

[0081] The preset threshold for the difference in permeability values ​​between adjacent grid cells is 0.05 m / d. That is, when the difference is greater than 0.05 m / d, the location is marked as the partition boundary.

[0082] The three-dimensional stratigraphic mesh model uses the Delaunay triangulation algorithm to generate tetrahedral meshes with a mesh cell size of 1m×1m×1m, constrained by stratigraphic fault planes and stratigraphic interfaces. The connectivity analysis uses a seed-based region merging algorithm, which uses randomly selected unassigned mesh cells as seeds to merge adjacent mesh cells with a permeability difference ≤0.05m / d to form connected partitions.

[0083] like Figure 7 As shown, the underground stratum microbial agent injection system of the present invention adopts a vertical deployment method from the surface to the deep underground layers. A control center 701 is located at the surface layer for overall control, and a microbial agent storage tank 702 stores the microbial agent solution to be injected. An injection pump 703 is connected to the microbial agent storage tank 702 via a pipeline and pressurizes and delivers the microbial agent to the injection well 704. The injection well 704 extends vertically from the surface to the deep underground layers. In the shallow underground layers, i.e., at a depth of 0 to 5 meters, temperature sensors 705, humidity sensors 706, and oxygen sensors 707 are arranged around the injection well 704 to monitor key environmental parameters such as temperature, humidity, and oxygen concentration in real time. In the deep underground layers, i.e., at a depth of 5 to 15 meters, the strata are distributed from top to bottom as a sandy layer, a clay layer, and a gravel layer, separated by natural transition interfaces in the form of wavy lines. The microbial agent diffuses outwards from the injection point through the injection well 704, and the diffusion path is indicated by dashed arrows. Contaminated area 708 is located at the junction of deep underground sandy and clay layers, and is marked with a shaded area. Pressure sensor 709 and flow meter 710 are distributed in the gravel layer to monitor pressure changes and flow data during the bacterial agent injection process, providing real-time feedback for dynamic adjustment of injection parameters.

[0084] like Figure 10 The figure shows the spatial distribution characteristics of permeability in the horizontal formation profile after optimization using this method. The main plot on the left is a 10x10 grid heatmap, using grayscale color mapping to represent the permeability values ​​of each zone, ranging from 0.1 to 0.9. High-permeability areas (permeability greater than 0.7) are marked with white dashed boxes; these areas are located in the upper right of the grid and represent favorable channels for preferential diffusion of the fungal agent. Restricted areas (permeability less than 0.2) are marked with black dashed boxes; these areas are located in the lower left of the grid, where the fungal agent flow is obstructed. Black arrows indicate the main flow path of the fungal agent after optimization using the CNN flow prediction model and injection pressure. This path effectively bypasses the restricted areas and fully utilizes the high-permeability channels, achieving uniform distribution of the fungal agent in the formation. The subplot on the right is a one-dimensional permeability profile curve along the grid's central axis, visually reflecting the spatial variation characteristics of permeability along the vertical direction. Through a sensor feedback dynamic adjustment mechanism, this method can respond in real-time to the heterogeneity of formation permeability and continuously optimize the injection strategy to improve the uniformity of fungal agent distribution.

[0085] In one possible implementation, obtaining stratigraphic scanning data typically relies on geophysical logging or 3D seismic exploration techniques.

[0086] For example, in oil and gas field development, sonic logging tools and resistivity logging tools deployed in wellbores can continuously acquire sonic transit time and resistivity curves of formations. These data reflect the lithology and pore fluid properties of the formations. Simultaneously, by combining surface-excited and subsurface-received 3D seismic wave reflection data, large-scale stratigraphic structures and property inversion models can be constructed. These multi-source heterogeneous scanning data, after coordinate alignment and standardization, collectively form the foundation dataset for subsequent analyses.

[0087] Specifically, mesh generation algorithms are the process of discretizing a continuous geological space into a finite number of cells. A commonly used method is tetrahedral mesh generation, which can adapt well to complex geological structural boundaries.

[0088] For example, using known fault planes and stratigraphic interfaces as constraints, the algorithm first generates triangular meshes on these constraint surfaces, then advances into the three-dimensional space to generate a set of non-overlapping tetrahedral elements. Each element is assigned initial attributes interpolated from the scan data, such as its lithofacies type, which provides a spatial basis for subsequent differential analysis. The fungal agent survival model is a prediction function based on the response to multiple environmental factors.

[0089] For example, in the application scenario of microbial enhanced oil recovery, this model considers multiple survival limiting factors such as temperature, salinity, pH, nutrient concentration, and formation pressure. For each grid cell, the model calls its corresponding environmental parameter vector and calculates the fitness score of the microbial agent in that specific environment, i.e., the survival probability, through a pre-trained machine learning model or empirical formula. This probability value quantifies the likelihood that the microbial agent will colonize within the cell and maintain its metabolic activity. The finite element method is used to simulate permeability distribution, the core of which is to discretize and solve the partial differential equations describing fluid flow in porous media.

[0090] For example, local flow equations can be established on each grid cell based on Darcy's law and the law of conservation of mass. By giving boundary conditions, such as the pressure of the injection well, and considering the inherent permeability parameters of each cell determined by lithology and pore structure, a large set of linear equations on the entire grid model can be systematically assembled and solved to obtain the distribution of the pressure and velocity fields, and finally, the equivalent permeability value of each cell under dynamic flow conditions can be derived. After obtaining the permeability values ​​of each grid cell, the partition boundaries need to be identified.

[0091] For example, a permeability variation threshold can be set. When the difference between the calculated values ​​of two adjacent cells exceeds this threshold, a significant physical property barrier is considered to exist between them, and the common surface between these two cells is marked as a potential partition boundary. These boundaries initially divide the homogeneous grid model into multiple blocks. Finally, connectivity analysis is used to merge blocks with similar properties. One embodiment employs a seed-based region merging algorithm. First, an unassigned grid cell is randomly selected as a seed, and all its adjacent cells are checked. If the difference between the permeability value of an adjacent cell and the seed cell is within an acceptable range, they are merged into the same region. Then, using the newly merged cell as the new frontier, this process is repeated until it cannot be expanded further, thus forming a connected partition. After traversing all cells, the entire stratigraphic model can be divided into several complex partitions with internal homogeneity and clear boundaries, providing accurate geological basis for subsequent differentiated engineering scheme design.

[0092] Step S104: Obtain complex formation partitions, use neural networks to predict the flow path of the fungal agent, determine whether the permeability difference causes flow stagnation, and if stagnation areas exist, optimize the injection pressure gradient to obtain a uniform distribution scheme.

[0093] Acquire formation scan data. Process the formation scan data using a convolutional neural network to predict the flow path of the fungal agent at various formation locations. Calculate the probability of fungal agent passage at each location based on the flow path to obtain the distribution of flow hindrance. Determine the degree of flow hindrance; if the flow hindrance in a certain area exceeds a preset threshold, it is marked as a hindrance area. For the locations of hindrance areas, simulate the flow state of the fungal agent under different injection pressure gradients using the finite difference method. Compare the uniformity of fungal agent coverage corresponding to each injection pressure gradient, and select the pressure gradient value with the highest uniformity. Use the selected pressure gradient value to generate the final injection parameters, obtaining a uniform distribution scheme for the fungal agent.

[0094] The degree of flow hindrance is characterized by the ratio of the actual probability of the agent passing through to the theoretical probability of passing through. The preset threshold is 0.5. That is, when the ratio is ≤0.5, the region is marked as a hindrance region.

[0095] The convolutional neural network used to predict the flow path of fungal agents consists of two convolutional layers, one pooling layer, and two fully connected layers. The kernel size of the convolutional layers is 3×3, the activation function is ReLU, the pooling layer uses max pooling, the loss function is mean squared error (MSE), the training iterations are 200, the input layer features are three-dimensional formation mesh porosity, pressure gradient vector, and fluid viscosity coefficient, and the training dataset is historical formation injection flow path matching data with formation parameters.

[0096] For the blocked areas, the pulsed pressure gradient control cycle is 10 seconds, and the pressure fluctuates regularly within the preset range; the pressure gradient range for sand and gravel strata is 0.3-0.5 MPa, for clay strata it is 0.5-0.8 MPa, and for karst strata it is 0.6-0.9 MPa; for non-blocked areas, atmospheric pressure injection is used, with a pressure gradient of 0.2-0.3 MPa.

[0097] In the neural network prediction, the input layer's feature parameters include: three-dimensional formation mesh porosity, pressure gradient vector, and fluid viscosity coefficient extracted from finite element simulation. The neural network learns from historical formation injection data and outputs the predicted range of the fungal agent concentration. If the prediction results indicate flow hindrance in certain low-permeability zones, the pressure amplitude or pulse frequency of the injection pump is adjusted to disrupt the local formation stress balance, thereby achieving uniform distribution of the fungal agent in complex, heterogeneous formations.

[0098] like Figure 8 As shown, this invention demonstrates the permeability zone structure and bacterial agent flow characteristics at different depths underground through a vertical stratigraphic profile. From the surface downwards, the zones are: topsoil (0-2m), high-permeability sandy zone 801 (2-6m), medium-permeability clay zone 802 (6-9m), low-permeability gravelly zone 803 (9-12m), and dense rock layer impeding zone 804 (12-14m). Each zone is separated by a wavy boundary 805, reflecting the transitional characteristics of the natural strata. Multiple injection points 806 are set at the surface, and the injection well casing extends from the surface to each stratum depth. Sensors 808 are distributed on the well casing for real-time monitoring of the bacterial agent concentration and permeability status of each layer. After injection at the injection points 806, the bacterial agent forms a flow path 807 and diffuses within each stratum. In the high-permeability zone 801, the flow path exhibits a smooth, straight diffusion pattern; in the medium-permeability zone 802, the flow path is slightly tortuous but still maintains continuity; in the low-permeability zone 803, the flow path becomes tortuous and the flow rate is limited; while in the stagnant zone 804, the flow of the microbial agent is blocked by the dense rock layer and needs to detour to the upper permeable zone. This profile visually reflects the impact of different formation permeability characteristics on the delivery efficiency of the microbial remediation agent, providing a basis for formulating differentiated injection strategies for different zones.

[0099] In one possible implementation, acquiring formation scanning data can be accomplished using seismic exploration equipment or drilling log tools, which are capable of capturing acoustic reflections or resistivity changes in underground rock formations, thereby generating high-resolution image data.

[0100] For example, in oil extraction operations, operators use 3D seismic scanners to scan target oil fields, collecting data including properties such as rock density, porosity, and fracture distribution. This data is stored in digital format for easy subsequent processing. This acquisition process emphasizes data accuracy and completeness, ensuring coverage of the entire stratigraphic area and avoiding the omission of key geological features.

[0101] For example, when processing these formation scan data through a convolutional neural network, the data can first be input into a pre-trained CNN model, which consists of multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layers extract features, such as edge detection and texture recognition, through kernel filtering, thereby predicting the flow path of the fungicide.

[0102] Specifically, during processing, the model analyzes the pore connectivity in the scanned images, simulating possible routes for the fungal agent to diffuse from the injection point outwards. For example, in sandstone layers, the path might flow along high-permeability channels, while in mudstone layers it would be blocked. By training on a dataset, the model learns historical flow patterns and ultimately outputs a path vector for each formation location, which helps visualize potential flow directions.

[0103] In one possible implementation, when calculating the probability of bacterial agent passage at each location based on these flow paths, a Monte Carlo simulation method can be used to estimate the probability value by randomly sampling resistance factors along the path, such as viscosity and pressure difference.

[0104] For example, in a reservoir restoration project, if there are multiple branches on the path at a specific location, the probability calculation will consider the pass rate of each branch and summarize to obtain an overall pass rate distribution map, thereby deriving the flow hindrance distribution. This distribution is represented by a heat map, with high hindrance areas shown in red, reflecting the parts where the microbial agent has difficulty penetrating.

[0105] For example, when determining the degree of flow obstruction and marking obstruction areas, a threshold such as 0.5 is set. When the degree of obstruction in a certain area exceeds this value, it is marked as an obstruction area.

[0106] Specifically, this involves creating a gridded distribution map, comparing the resistance value of each grid cell, and marking it with a boundary line if it is higher than a threshold. For example, in complex carbonate rock formations, high resistance areas often correspond to low porosity areas. This step ensures that subsequent simulations target the problem area.

[0107] In one possible implementation, when simulating the flow state of the microbial agent under different injection pressure gradients using the finite difference method for these stagnant regions, the method discretizes the formation into a grid and solves the difference form of the flow equation.

[0108] For example, in the simulation, the pressure gradient is set from 10 kPa / m to 50 kPa / m, the change in the concentration of the microbial agent at each time step is calculated, and the flow velocity and diffusion range are observed to evaluate the state change, which can reveal how pressure overcomes the blockage in the business of improving oil recovery.

[0109] For example, when comparing the uniformity of the coverage of the microbial agent corresponding to each injection pressure gradient, the standard deviation index can be used to quantify the uniformity, and the gradient value with the lowest standard deviation is selected.

[0110] Specifically, in one example, at a gradient of 30 kPa / m, the coverage uniformity reaches 85%, which is higher than other values, based on the statistical analysis of the simulation data.

[0111] In one possible implementation, when generating the final injection parameters using the selected pressure gradient value, it is incorporated into the injection plan, such as adjusting the pumping rate and the concentration of the microbial agent, to obtain a uniform distribution plan, which can improve the coverage efficiency of the microbial agent in the target area and achieve better resource utilization in actual oilfield operations.

[0112] Step S105: According to the uniform distribution plan, use real-time sensor feedback to inject process data, judge the distribution path deviation. If the deviation is greater than the preset value, dynamically adjust the injection point position to obtain an accurate penetration trajectory.

[0113] Obtain the uniform distribution path planning result as the initial trajectory benchmark. Real-time collect the injection process data through the sensor array to obtain the current distribution trajectory sequence. Compare the current distribution trajectory sequence with the initial trajectory benchmark point by point to obtain the deviation sequence. If the deviation at any position in the deviation sequence is greater than the preset threshold, it is determined that the trajectory deviates. Calculate the deviation direction and amplitude according to the deviation sequence to obtain the adjustment vector. Use the linear regression method to fit the trend of the deviation sequence to determine the spatial displacement of the injection point. Superimpose the spatial displacement on the current injection point coordinates to obtain the updated injection point position.

[0114] The preset threshold for the distribution path deviation is 0.5 meters. If the formation permeability is lower than 0.1 m / d, the deviation threshold is adjusted to 0.3 meters; when the actual deviation value at any position exceeds the corresponding threshold, it is determined that the trajectory deviates.

[0115] The path deviation vector is calculated according to the three-dimensional space coordinates (horizontal X-axis, horizontal Y-axis, vertical Z-axis). The compensation displacement is the product of the modulus of the deviation vector and the adjustment coefficient of 0.8. The adjustment accuracy of the servo injection system is ±0.1 meters, and the response time ≤ 1 second. It can offset the position of the injection pipe nozzle in one or more dimensions of the horizontal X / Y axes and the vertical Z-axis, and keep the injection pressure gradient stable after adjustment.

[0116] Such as Figure 5As shown, the core mechanism of this invention lies in the dual coupling of two state chains: activity regulation and distribution optimization. The left-hand activity regulation state chain sequentially goes through seven states: environmental parameter acquisition, deviation calculation, feature extraction, GA optimization, threshold determination, encapsulation adjustment, and survival model establishment. The right-hand distribution optimization state chain sequentially goes through eight states: formation scanning, grid construction, permeability analysis, flow prediction, pressure optimization, sensor feedback, trajectory correction, and coverage verification. Data exchange between the two chains is achieved through three coupling points: the first coupling point transmits the activity maintenance threshold as a parameter constraint condition for formation scanning; the second coupling point uses the output of the survival model as an input parameter for the agent survival probability in grid construction; and the third coupling point feeds back the coverage verification results to the environmental parameter re-acquisition stage, forming a closed-loop regulation. The above coupling mechanism ensures the dynamic consistency between the activity regulation parameters and the distribution optimization parameters, avoiding the local optimum problem that may be caused by a single optimization chain.

[0117] For example, in the business scenario of injecting microbial agents into oil formations, it is first necessary to obtain the uniformly distributed path planning results as the initial trajectory benchmark, which involves using a pre-simulated model to define the ideal injection path.

[0118] Specifically, such a benchmark can be generated using geological simulation software, such as a three-dimensional model built based on formation permeability data, to simulate the ideal trajectory of the fungicide diffusion from the injection well to the reservoir, thereby ensuring that the initial plan covers the entire target area and avoids local accumulation.

[0119] In one embodiment, injection process data is collected in real time by a sensor array to obtain the current distribution trajectory sequence. Here, the sensor array can be a combination of pressure sensors and flow meters arranged in the wellbore and formation. They collect data once per second, for example, to monitor the flow rate and position of the microbial agent during the injection process, forming a sequence composed of timestamps and coordinate points to reflect the actual distribution.

[0120] For example, by comparing the current distribution trajectory sequence with the initial trajectory reference point by point, a deviation sequence can be obtained. This can be achieved through a coordinate matching algorithm, such as comparing each point in the sequence with the reference point and calculating the Euclidean distance. If the sequence point is (x1, y1, z1) and the reference is (x0, y0, z0), then the deviation value is the distance difference, forming a list of deviation values ​​for subsequent analysis.

[0121] In one embodiment, if the deviation at any position in the deviation sequence is greater than a preset threshold, the trajectory is determined to have deviated. For example, if the threshold is set to 0.5 meters, when the deviation at a certain point reaches 0.6 meters, the system will mark it as a deviation. This helps to identify problems in a timely manner and avoid uneven injection.

[0122] For example, the direction and magnitude of the deviation can be calculated from the deviation sequence to obtain an adjustment vector. This can be processed through vector operations, such as extracting the direction component from the deviation sequence, where a positive x-axis deviation indicates a rightward deviation and the magnitude is the average deviation value, thereby synthesizing a three-dimensional adjustment vector to guide corrections.

[0123] In one embodiment, a linear regression method is used to fit the trend of the deviation sequence to determine the spatial displacement of the injection point. Here, linear regression can fit the deviation value to a scatter plot of time or location. For example, the least squares method can be used to estimate the slope of the trend line to predict the displacement, such as calculating the amount that needs to be moved 0.3 meters towards the z-axis. This step connects the deviation analysis with the actual adjustment to ensure logical continuity.

[0124] For example, the spatial displacement is superimposed on the current injection point coordinates to obtain the updated injection point position. For instance, if the current coordinates are (100, 200, 50) and the displacement is (0, 0.2, 0.3), then the new coordinates are (100, 200.2, 50.3). This optimizes the injection strategy and enables more precise control of fungal agent distribution in business operations.

[0125] Step S106: Using a precise penetration trajectory, the coverage area is mapped for the location of the pollution source. The coverage integrity is verified through Monte Carlo simulation to determine the pollution coverage grid.

[0126] Obtain the coordinates of the pollution source location. Generate a precise penetration trajectory based on the pollution source location coordinates. Calculate the coverage area mapping using the precise penetration trajectory. Perform random sampling within the coverage area mapping using Monte Carlo simulation. Statistically analyze the distribution density of sampling points obtained from the Monte Carlo simulation. If the sampling point distribution density meets a preset uniformity condition, the coverage integrity is deemed passed. Perform mesh generation using the coverage area mapping that has passed the coverage integrity test to obtain the pollution coverage mesh.

[0127] The uniformity condition for the distribution density of sampling points is: the variance of the sampling point density of all sub-regions within the coverage area is ≤0.1, and the deviation rate between the density of each sub-region and the average density is ≤20%.

[0128] The Monte Carlo simulation involved 10,000 random samplings. The coverage area was mapped to sub-regions of 1m×1m×1m. Sampling points were randomly distributed within each sub-region, and the total number of sampling points was allocated according to the volume ratio of the sub-regions.

[0129] like Figure 11As shown in the figure, the quantitative comparison results of our method with three comparative methods in terms of two key indicators: repair efficiency and cover integrity are presented. The X-axis represents the four different methods, the left Y-axis is a bar chart representing the percentage of repair efficiency, and the right Y-axis is a line chart representing the percentage of cover integrity. Our method achieves a repair efficiency of 93% and a cover integrity of 96%, both significantly better than the other three methods. The traditional uniform injection method has a repair efficiency of only 62% and a cover integrity of 58%, indicating that the uniform injection strategy cannot effectively cope with the heterogeneous characteristics of complex formations. The feedback-free adjustment method has a repair efficiency of 71% and a cover integrity of 68%, indicating that the lack of a real-time sensor feedback mechanism prevents the injection strategy from dynamically adapting to changes in formation conditions. The single-model optimization method has a repair efficiency of 78% and a cover integrity of 75%, showing some improvement but still inferior to the multi-model collaborative optimization strategy used in our method. Our method achieves optimal performance in both repair efficiency and cover integrity through a combination of Monte Carlo simulation verification and iterative convergence optimization.

[0130] For example, in the field of environmental monitoring, the process of obtaining the location coordinates of pollution sources typically involves using a global positioning system in conjunction with ground sensors to determine the specific location.

[0131] Specifically, the latitude and longitude data of the pollution source are first captured by a satellite signal receiver. For example, the emission outlet of a chemical plant is located at 30 degrees north latitude and 120 degrees east longitude. Then, the altitude information is obtained by combining it with lidar scanning to form three-dimensional coordinates such as (120, 30, 50). This can accurately locate the pollution source and provide basic data for subsequent trajectory generation.

[0132] In one embodiment, when generating a precise penetration trajectory based on the location coordinates of the pollution source, a path optimization algorithm can be used to calculate a straight or curved path from the starting point to the pollution source.

[0133] For example, assuming the pollution source coordinates are (120,30,50), the system will consider terrain obstacles such as hills or buildings, and fit a trajectory curve that avoids the obstacles using the least squares method to ensure that the trajectory can directly penetrate to the core area of ​​the pollution source, thereby forming a precise trajectory sequence with a length of 5 kilometers starting from the monitoring station.

[0134] For example, the process of accurately penetrating the trajectory to calculate the coverage mapping requires expanding the trajectory into a projection of an area of ​​influence.

[0135] Specifically, the system will be based on a trajectory point diffusion model. For example, it will draw a circular area with a radius of 2 kilometers centered on each trajectory point, and then overlay all areas to form an overall mapping map. This will result in an elliptical range mapping that covers the pollution diffusion path, with an area of ​​about 10 square kilometers, which will be used for subsequent sampling and analysis.

[0136] In one embodiment, the method of random sampling within the coverage map using Monte Carlo simulation simulates the uncertainty of pollution distribution by generating a large number of random points.

[0137] For example, 10,000 sampling points are set up within the mapping area, and the system randomly selects locations such as (120.1, 30.2, 45), and repeats this process multiple times to assess the possible distribution paths of pollution particles. This helps to capture the influence of random factors such as wind direction changes.

[0138] For example, the steps for statistically analyzing the distribution density of sampling points obtained from Monte Carlo simulations include counting and density calculations for all sampling points.

[0139] Specifically, the mapping area is divided into small blocks, each with an area of ​​1 square kilometer. Then, the number of points in each block is calculated. For example, if a block has 500 points, the density is 500 points / square kilometer. The overall distribution is quantified by the average density and variance, providing data support for the uniformity judgment.

[0140] In one embodiment, if the sampling point distribution density meets a preset uniformity condition, the logic for determining that the coverage integrity has passed is based on threshold comparison.

[0141] For example, if the preset condition is that the density variance is less than 0.1, and the statistical result shows that the variance is 0.05, then it is considered to pass. This means that the sampling points uniformly cover the mapped area, avoiding the problem of local density or sparseness, thus confirming the integrity of the coverage.

[0142] For example, the process of obtaining a contaminated coverage grid by performing grid division through coverage integrity mapping can use the equal-spacing grid method to subdivide the area.

[0143] Specifically, the mapping is divided into a 10x10 grid, with each grid cell measuring 1 km x 1 km. For example, starting from coordinates (120, 30) and expanding eastward and northward, a grid map containing 100 cells is formed. Each cell is labeled with an estimated pollution concentration, thus obtaining a detailed pollution cover grid for further environmental governance planning.

[0144] Step S107: Obtain the pollution coverage grid, feed it back to the initial activity maintenance threshold for iterative cycling, determine whether the overall remediation parameters have converged, and if not converged, reset the combination of environmental variables to obtain the final distribution optimization result.

[0145] The current contaminated grid is acquired through the grid data acquisition module. This grid is then input into the threshold calculation unit to obtain the initial activity maintenance threshold. Using this initial threshold as a starting point, a loop iterative calculation is initiated to update the activity distribution state. The current overall remediation parameters are extracted from the loop iterative calculation. It is determined whether the difference between the overall remediation parameters and the parameters from the previous round is less than a preset convergence threshold. If so, the loop terminates, and the distribution optimization result is obtained; otherwise, subsequent operations continue. For overall remediation parameters that do not meet the convergence condition, the current combination of environmental variables is obtained. A reset operation is performed on the environmental variable combination to generate a new combination. This new combination of environmental variables is fed back to the initial activity maintenance threshold calculation unit, and the loop iterative calculation is restarted. The final activity distribution state is extracted from the converged overall remediation parameters to obtain the final distribution optimization result.

[0146] The overall repair parameter convergence threshold is 5%, which means that when the relative difference between the current overall repair parameter and the previous round parameter is ≤5%, the parameter is considered to have converged and the loop is terminated.

[0147] The feedback mechanism here establishes a physical coupling between the "contamination coverage grid" and the "activity threshold": if the Monte Carlo simulation shows insufficient contamination coverage integrity (i.e., remediation dead zones exist), the system will determine that the current activity maintenance threshold is too high, causing premature inactivation of the microbial agent, and thus lower the initial threshold requirement or increase the injection pressure, initiating a new round of iteration. This feedback ensures that the remediation scheme not only theoretically meets the bioactivity standard, but also physically covers the entire contamination plume area.

[0148] For example, in environmental monitoring systems, grid data acquisition modules are typically responsible for acquiring data in real time from sensor networks. This data includes the specific coordinates and concentration distribution of pollution-covered grids. In a pollution control project in an industrial park, this module would collect pollutant indicators in the soil and air, such as heavy metal concentrations or volatile organic compound levels, using ground- and airborne sensor arrays, thus forming a two-dimensional or three-dimensional grid model. This process first involves data preprocessing, such as filtering noise and calibrating sensor readings to ensure the accuracy of the grid. Subsequently, this grid data is transmitted to a processing unit, providing the basis for subsequent calculations.

[0149] Specifically, after the contaminated cover grid is input into the threshold calculation unit, the unit calculates the initial activity maintenance threshold based on the pollutant density and diffusion model in the grid. Here, the threshold calculation unit can be understood as an algorithmic framework that integrates the pollutant decay function and environmental carrying capacity parameters.

[0150] For example, in a river pollution remediation scenario, if the grid shows that phosphorus concentrations in certain areas exceed the standard, the cell will determine an initial threshold, such as 0.5 mg / L, by assessing the diffusion rate and biodegradation rate of the pollutant in the water flow. This threshold represents the minimum pollutant control level required to maintain ecological activity. The calculation process involves first mapping grid points into a mathematical model, and then applying weighting factors to adjust the threshold to reflect the sensitivity of different areas.

[0151] In one embodiment, when initiating a loop iteration using an initial activity maintenance threshold as a starting point, the activity distribution state is updated. This involves iterative algorithms such as gradient descent variants, simulating changes in pollutant distribution in each round.

[0152] For example, in soil remediation operations, an initial threshold is input into a simulation engine that iteratively calculates the activity state, i.e. how pollutants migrate and degrade in the grid. A new distribution map is generated after each update, thereby gradually optimizing the remediation strategy.

[0153] For example, the process of extracting the current overall remediation parameters from iterative calculations can be achieved by summarizing the iteration results. In a case of urban pollution control, these parameters might include remediation coverage and resource allocation efficiency. When extracting from the iterations, the average improvement value of all grid points is calculated to form overall parameters such as a purification efficiency of 85%.

[0154] Specifically, when determining whether the difference between the overall repair parameters and the parameters of the previous round is less than the preset convergence threshold, if the difference, such as 0.01, is less than the threshold of 0.05, the loop is terminated and the distribution optimization result is obtained; otherwise, it continues.

[0155] For example, in a wastewater treatment project, if the parameter difference is large, it indicates that the optimization has not converged and further adjustments are needed.

[0156] In one embodiment, for overall repair parameters that do not meet convergence criteria, obtaining the current combination of environmental variables involves extracting variables such as temperature and humidity. In marine pollution monitoring, these variables, such as water temperature (25 degrees Celsius) and salinity levels, are collected as key factors influencing pollutant dispersion.

[0157] For example, resetting a combination of environmental variables to generate a new combination can be achieved by adjusting the variable values ​​through random perturbation or optimization algorithms. In an agricultural soil pollution remediation project, resetting might adjust the temperature from 20 degrees Celsius to 22 degrees Celsius, generating a new combination to simulate different scenarios.

[0158] Specifically, feeding back new combinations of environmental variables to the initial activity maintenance threshold calculation unit when re-entering the loop ensures the continuity of iteration.

[0159] For example, in air pollution control, after a new combination of inputs, the threshold unit will recalculate the threshold, driving the loop toward convergence.

[0160] In one embodiment, the process of extracting the final activity distribution state from the converged overall remediation parameters to obtain the final distribution optimization result includes visualization and verification steps. In a plant emission control operation, the final state displays an optimized pollutant distribution map, achieving more uniform remediation coverage, thereby improving overall environmental restoration efficiency.

[0161] like Figure 9 The figure shows the trend of the bacterial agent survival rate during 20 iterations of optimization and its comparison with other methods. The X-axis represents the iteration number, the left Y-axis represents the percentage of bacterial agent survival rate, and the right Y-axis represents the environmental parameter deviation value. This method uses a genetic algorithm to optimize environmental variables, combined with an SVM survival prediction model and a dynamic adjustment strategy for encapsulation materials. The bacterial agent survival rate gradually increased from approximately 55% in the first iteration to approximately 92% in the 20th iteration, indicated by a light gray area below the curve. In the 8th iteration, the survival rate exceeded the critical threshold of 80%, indicating that this optimization strategy can achieve significant performance improvement in a relatively small number of iterations. In contrast, the bacterial agent survival rate of traditional methods fluctuated between 40% and 50%, failing to achieve effective optimization; the method without encapsulation adjustment, although slowly increasing from 50% to approximately 65%, showed limited improvement. The right-axis environmental parameter deviation curve shows a continuously decreasing trend, indicating that this method can effectively converge the environmental parameters to the optimal range, thus providing stable and suitable formation environmental conditions for bacterial agent survival.

[0162] The interface specifications of the grid data acquisition module, threshold calculation unit, and simulation unit involved in this invention are as follows: 1. Data transmission format: All adopt JSON format, including parameter name, parameter value, acquisition / calculation timestamp, and grid unit coordinates; 2. Input / output interface: The output of the grid data acquisition module is the three-dimensional coordinates of the contaminated grid and the agent coverage density, which serve as the input of the threshold calculation unit; the output of the threshold calculation unit is the activity maintenance threshold, which serves as the input of the simulation unit; the output of the simulation unit is the overall remediation parameters, which are fed back to the threshold calculation unit to form a closed loop; 3. Data transmission rate: ≥10MB / s, meeting the requirements of real-time control.

[0163] Notes on the engineering implementation of this invention:

[0164] Sensor calibration: The sensor array must be calibrated on-site before installation. The calibration error of the temperature sensor should be ≤ ±0.1℃, the calibration error of the humidity sensor should be ≤ ±1%, and the calibration error of the oxygen sensor should be ≤ ±0.5%. On-site re-inspection should be carried out every 7 days to ensure data accuracy.

[0165] Drug injection system maintenance: The injection port of the servo drug injection system needs to be cleaned regularly to prevent blockage by formation sediment. The cleaning cycle is 3 days. The pressure sensor needs to be calibrated regularly. The calibration cycle is 15 days to ensure accurate pressure gradient control.

[0166] Formation parameter update: The underground formation parameters will change slightly during the injection process. Every 24 hours, the formation scanning data is re-acquired by the scanning equipment to update the three-dimensional mesh model and ensure the accuracy of subsequent simulations and predictions.

[0167] Storage and transportation of microbial agents: During the transportation of microbial agents to the site, the temperature should be maintained at 20-30℃ and the humidity at 60-70%, and the agents should be sealed and packaged. The transportation time should not exceed 24 hours. On-site storage should be carried out in a low-temperature refrigerated box, and the storage time should not exceed 7 days to ensure the initial activity of the microbial agents.

[0168] Injection operation specifications: During the injection of microbial agent, maintain a stable injection rate of 0.5-1 m³ / h to avoid sudden changes in the injection rate that could damage the formation structure and affect the distribution of the microbial agent.

[0169] Compared with traditional microbial agent remediation methods, the core technical effects of this invention are quantitatively compared as follows: 1. Pollution plume coverage: increased from 40%-45% in traditional methods to 85%-88%; 2. Microbial agent survival rate: increased from around 60% in traditional methods to over 90%; 3. Remediation cycle: shortened by 25%-30% compared to traditional methods; 4. Distribution deviation rate: reduced from over 10% in traditional methods to below 5%; 5. Pollutant degradation rate: degradation rate of pollutants such as chlorinated hydrocarbons / chromium reaches over 80%, an improvement of 30% compared to traditional methods.

[0170] The correlation between each technical feature and its corresponding beneficial effect is as follows: 1. Genetic algorithm optimizes the combination of environmental variables and accurately determines the activity maintenance threshold, increasing the activity maintenance rate of the microbial agent in complex formations by more than 30%; 2. Support vector machine constructs a microbial agent survival model and adjusts the encapsulation material parameters, effectively shielding harmful environmental factors in the formation, increasing the microbial agent survival rate by more than 25%; 3. Convolutional neural network predicts the microbial agent flow path and combines it with pressure gradient optimization to solve the problem of blockage in low-permeability formations, increasing the uniformity of microbial agent distribution by more than 40%; 4. Dynamically adjusting the injection point position to obtain a precise penetration trajectory, reducing the distribution deviation rate by more than 5%; 5. Monte Carlo simulation verifies the coverage integrity and combines it with closed-loop iteration, increasing the pollution plume coverage rate by more than 40%; 6. The overall parameter cyclic iteration convergence mechanism shortens the remediation cycle by 25%-30%.

[0171] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to specific implementations. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for regulating the activity and distribution of microbial agents used in the remediation of underground contamination in complex formations, characterized in that, The method includes: By collecting microbial agent samples and monitoring parameters such as temperature, humidity, and oxygen levels in real time, a genetic algorithm is used to optimize the combination of environmental variables to obtain an activity maintenance threshold. Based on the activity maintenance threshold, underground stratum scanning data, including soil type and porosity distribution, is obtained to determine the degree of environmental impact. If the environmental impact exceeds the preset threshold, the parameters of the microbial agent encapsulation material are adjusted to obtain a microbial agent survival model. Using the microbial agent survival model, a three-dimensional mesh representation is constructed based on the stratum scanning data. Permeability differences are analyzed through finite element simulation to determine complex stratum zones. After obtaining the complex stratum zones, a neural network is used to predict the microbial agent flow path and determine whether permeability differences cause flow stagnation. If stagnation areas exist, the injection pressure gradient is optimized to obtain a uniform distribution scheme. Based on the uniform distribution scheme, real-time sensor feedback of injection process data is used to determine the distribution path deviation. If the deviation is greater than a preset value, the injection point position is dynamically adjusted to obtain a precise penetration trajectory. The dynamic adjustment of the injection point position refers to: calculating the compensation displacement based on the magnitude of the path deviation vector and driving the servo injection system to offset the injection nozzle position in the horizontal or vertical dimension. Using precise penetration trajectories, the coverage area is mapped to the location of pollution sources. The integrity of the coverage is verified through Monte Carlo simulation to determine the pollution coverage grid. The pollution coverage grid is obtained and fed back to the initial activity maintenance threshold for iterative iteration. It is then determined whether the overall remediation parameters have converged. If they have not converged, the combination of environmental variables is reset to obtain the final distribution optimization result.

2. The method according to claim 1, characterized in that, The process involves collecting bacterial agent samples and monitoring parameters such as temperature, humidity, and oxygen levels in real time. A genetic algorithm is then used to optimize the combination of environmental variables to obtain the activity maintenance threshold, including: The current environmental three-parameter sequence is obtained by collecting sample temperature, sample humidity and oxygen level data in real time through a sensor array; The deviation sequence is obtained by calculating the deviation between the current value and the historical mean of each parameter based on the current environmental three-parameter sequence. A sliding window is used to extract local variation features from the deviation sequence, resulting in a local variation feature sequence. Three types of features—temperature, humidity, and oxygen level—are selected from the local change feature sequence to form the feature vector for the current moment. The current feature vector is input into the genetic algorithm optimization process, and the combination of environmental parameters is iteratively searched to obtain a set of candidate combinations; The activity maintenance level at corresponding time points is simulated and calculated for each candidate combination set to obtain the activity maintenance level sequence for each combination. By comparing the activity maintenance level sequences of each combination, the combination corresponding to the highest level is selected to determine the activity maintenance threshold.

3. The method according to claim 1, characterized in that, The process involves acquiring underground strata scanning data, including soil type and porosity distribution, based on an activity maintenance threshold, to determine the degree of environmental impact. If the environmental impact exceeds a preset threshold, the parameters of the bacterial agent encapsulation material are adjusted to obtain a bacterial agent survival model, including: Underground strata data, including soil type and porosity distribution information, are acquired using scanning equipment and stored as an initial dataset. Based on the initial dataset, soil type and pore distribution characteristics were analyzed, and a classification method was used to conduct a preliminary assessment of the environmental impact to obtain the environmental impact level. If the environmental impact level exceeds the preset threshold, specific data on soil type and pore distribution are extracted, and combined with the activity maintenance requirements, it is determined whether the survival conditions of the microbial agent meet the standards. Based on the judgment results, the material parameters for encapsulating the microbial agent were adjusted to address the insufficient survival conditions of the microbial agent, thereby generating an optimized material configuration scheme. The optimized material configuration scheme was obtained, and a microbial agent survival model was constructed based on the environmental impact level. The survival probability was predicted by the support vector machine algorithm, and the survival model parameters were determined. Based on the survival model parameters, the activity maintenance state of the microbial agent in the underground strata was simulated to obtain the final activity distribution results.

4. The method according to claim 1, characterized in that, The aforementioned method employs a microbial agent survival model, constructs a three-dimensional mesh representation based on formation scanning data, and uses finite element simulation analysis to determine complex formation zones, including: Acquire stratigraphic scan data; A 3D mesh model is generated using a mesh generation algorithm; The survival probability of each grid cell was calculated using a microbial agent survival model. The permeability distribution of each grid cell was simulated using the finite element method. The permeability values ​​of each grid cell are obtained based on the permeability simulation results; Determine the difference in permeability values ​​between adjacent grid cells. If the difference exceeds a preset threshold, mark it as a partition boundary. By merging grid cells with similar permeability through connectivity analysis, complex stratigraphic partitions can be identified.

5. The method according to claim 1, characterized in that, The process of obtaining complex formation zones involves using neural networks to predict the flow path of the inoculant, determining whether permeability differences cause flow stagnation, and optimizing the injection pressure gradient to obtain a uniform distribution scheme if stagnation exists. This includes: Acquire stratigraphic scan data; By processing formation scan data using convolutional neural networks, the flow path of fungal agents at various formation locations can be predicted. The probability of the inoculant passing through each location is calculated based on the flow path, thus obtaining the distribution of the degree of flow hindrance. Determine the degree of flow obstruction. If the degree of flow obstruction in a certain area is higher than a preset threshold, it is marked as an obstructed area. For the location of the blockage region, the flow state of the bacterial agent under different injection pressure gradients was simulated using the finite difference method. Compare the uniformity of bacterial agent coverage corresponding to each injection pressure gradient, and select the pressure gradient value with the highest uniformity of coverage. The final injection parameters are generated using the selected pressure gradient value to obtain a uniform distribution scheme for the bacterial agent.

6. The method according to claim 1, characterized in that, The process involves using real-time sensor feedback of injection process data based on a uniform distribution scheme to determine the distribution path deviation. If the deviation exceeds a preset value, the injection point position is dynamically adjusted to obtain a precise penetration trajectory. This includes: Obtain the uniformly distributed path planning results as the initial trajectory reference; The current distribution trajectory sequence is obtained by collecting injection process data in real time through a sensor array; By comparing the current trajectory sequence with the initial trajectory baseline point by point, a deviation sequence is obtained; If the deviation at any position in the deviation sequence is greater than a preset threshold, the trajectory is determined to have deviated. The direction and magnitude of the deviation are calculated based on the deviation sequence to obtain the adjustment vector; The spatial displacement of the injection point is determined by fitting the trend of the deviation sequence using linear regression. The spatial displacement is superimposed on the current injection point coordinates to obtain the updated injection point position.

7. The method according to claim 1, characterized in that, The method employs precise penetration trajectories to map coverage areas based on pollution source locations, verifies coverage integrity through Monte Carlo simulation, and determines the pollution coverage grid, including: Obtain the coordinates of the pollution source location; A precise penetration trajectory is generated based on the location coordinates of the pollution source; Coverage mapping is calculated through precise penetration trajectory calculation; Monte Carlo simulation is used to perform random sampling within the coverage map; Statistical analysis of sampling point distribution density obtained from Monte Carlo simulation; If the sampling point distribution density meets the preset uniformity condition, the coverage integrity is deemed to have passed. Mesh partitioning is performed by mapping the coverage area through coverage integrity to obtain a contaminated coverage mesh.

8. The method according to claim 1, characterized in that, The process involves acquiring a contaminated cover grid, feeding it back to the initial activity maintenance threshold, iterating repeatedly, and determining whether the overall remediation parameters have converged. If they have not converged, the environmental variable combination is reset to obtain the final distribution optimization result, including: The current polluted grid is obtained through the grid data acquisition module; The contaminated cover grid is input into the threshold calculation unit to obtain the initial activity maintenance threshold; Starting from the initial activity maintenance threshold, a loop iterative operation is initiated to update the activity distribution state; Extract the current overall repair parameters from the iterative loop calculation; Determine whether the difference between the overall repair parameters and the parameters of the previous round is less than the preset convergence threshold. If it is satisfied, terminate the loop and obtain the distribution optimization result. If it is not satisfied, continue to execute the subsequent operations. For overall repair parameters that do not meet the convergence criteria, obtain the current combination of environment variables; Perform a reset operation on the combination of environment variables to generate a new combination of environment variables; The new combination of environmental variables is fed back to the initial activity maintenance threshold calculation unit, and the loop iteration calculation is restarted. The final activity distribution state is extracted from the converged global repair parameters to obtain the final distribution optimization result.

9. An electronic device, comprising: A processor, a memory, and a program stored in the memory and executable on the processor, characterized in that, when executed by the processor, the program implements the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 8.

Images