High-salinity wastewater purification method that couples advanced oxidation and membrane filtration

Abstract

This invention discloses a method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration, belonging to the field of high-salinity wastewater treatment and purification technology. The method includes collecting multi-source heterogeneous data streams of high-salinity wastewater conductivity fluctuations, UV absorbance, and suspended solids concentration. These data streams are then reconstructed into a standardized time-series matrix through adaptive filtering and outlier removal. The matrix is ​​analyzed to generate a pollutant feature map. The spatial topology of the oxidation reaction zone and membrane separation zone is dynamically divided, generating treatment domain configuration data containing zone boundary coordinates. A non-uniform mesh generation engine is used to construct a gradient-resolution discretized computational space, conducting cross-scale mass transfer simulations and outputting a snapshot of the full flow field concentration distribution. The membrane module operation logs are combined to calculate the fouling deposition rate on the membrane surface. The zone boundary coordinates are then corrected in reverse and iteratively converged to output a purification path plan. This scheme can dynamically match pollutant characteristics, optimize the treatment zone layout, and adapt to the high-salinity wastewater purification process.

Description

Technical Field

[0001] This invention belongs to the field of high-salinity wastewater treatment and purification technology, specifically a high-salinity wastewater purification method that couples advanced oxidation and membrane filtration. Background Technology

[0002] Current high-salinity wastewater purification methods often employ a coupled advanced oxidation and membrane filtration process. During implementation, only water quality parameters such as conductivity, UV absorbance, and suspended solids concentration are collected. However, adaptive filtering and outlier removal are not performed after parameter acquisition, making it impossible to reconstruct a standardized time-series matrix from multi-source heterogeneous data. Furthermore, matrix analysis cannot identify the dominant pollutant type or quantify the initial load, and corresponding pollutant characteristic maps cannot be generated. The oxidation reaction zone and membrane separation zone use a fixed spatial layout, with partition boundaries manually defined. This prevents dynamic spatial topology division based on pollutant characteristics and the generation of treatment domain configuration data containing partition boundary coordinates.

[0003] Existing treatment processes lack a non-uniform mesh generation engine, making it impossible to construct a discretized computational space with gradient resolution. This hinders cross-scale mass transfer simulations within the computational space, prevents tracking of the spatiotemporal evolution of pollutant concentration and oxidant diffusion fields, and fails to output a snapshot of the entire flow field concentration distribution. Furthermore, it cannot rely on extreme regions of the entire flow field concentration distribution and historical operating logs of the membrane module to calculate the fouling deposition rate on the membrane surface. It also lacks a technical path to reverse-correct the boundary coordinates of the zones based on the fouling deposition rate, and cannot achieve dynamic optimization of the treatment domain configuration through iterative convergence. The purification path planning cannot adapt to the real-time treatment status of the wastewater. In particular, the pretreatment-aeration-sedimentation process widely used in coal mine water treatment also suffers from the same technical shortcomings. Its fixed layout of aeration and sedimentation zones prevents dynamic optimization of spatial topology and operating parameters based on influent water quality, making it difficult to achieve efficient pretreatment, efficient aeration, and efficient sedimentation in synergy. This results in poor system resistance to shock loads and low operating efficiency. This invention addresses these shortcomings by providing a dynamic control and optimization method applicable to high-salinity wastewater and coal mine water scenarios. Summary of the Invention

[0004] This invention aims to solve at least one of the technical problems existing in the prior art;

[0005] Therefore, this invention proposes a method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration, comprising:

[0006] Collect multi-source heterogeneous data streams of high-salt wastewater to be treated, the multi-source heterogeneous data streams covering conductivity fluctuation values, ultraviolet absorbance and suspended solids concentration;

[0007] The multi-source heterogeneous data stream is input into the preprocessing pipeline, and after adaptive filtering and outlier removal, it is reconstructed into a normalized time series matrix.

[0008] The standardized time series matrix is ​​analyzed to identify the dominant pollutant types and quantify their initial loads, generating a pollutant feature map.

[0009] Based on the pollutant characteristic map, the spatial topology of the oxidation reaction zone and the membrane separation zone is dynamically divided, and processing domain configuration data containing the partition boundary coordinates is generated.

[0010] Read the processing domain configuration data, call the non-uniform mesh generation engine, and construct a discretized computational space with gradient resolution;

[0011] Perform cross-scale mass transfer simulations within the discretized computational space, track the spatiotemporal evolution of the pollutant concentration field and the oxidant diffusion field, and output a snapshot of the concentration distribution of the entire flow field.

[0012] The extreme value regions in the full flow field concentration distribution snapshot are extracted, and the fouling deposition rate on the membrane surface is estimated by combining the historical operation logs of the membrane module.

[0013] Based on the membrane surface fouling deposition rate, the partition boundary coordinates in the processing domain configuration data are corrected in reverse, iterating until convergence, and finally outputting the purification path plan for the high-salinity wastewater to be treated.

[0014] Furthermore, the reconstruction into a normalized time series matrix includes:

[0015] The multi-source heterogeneous data streams are timestamped to eliminate timing misalignment caused by differences in sensor sampling frequencies;

[0016] Perform sliding window integration on the aligned data stream to extract the mean, variance, and peak-to-valley difference within the window, forming a multidimensional statistical feature vector.

[0017] The multidimensional statistical feature vector is mapped to a preset standard range space, and the missing sampling points are filled in by linear interpolation to generate the standardized time series matrix.

[0018] Furthermore, the process of identifying the dominant pollutant type and quantifying its initial load to generate a pollutant characteristic map includes:

[0019] Principal component decomposition was performed on the standardized time series matrix to separate independent component signals representing organic matter, inorganic salts, and colloidal particles;

[0020] Calculate the energy entropy value of each independent component signal, and determine the component with the highest energy entropy value as the dominant pollutant type;

[0021] Based on the spectral absorption characteristic curves of the dominant pollutant type, its mass concentration in the original water sample is calculated by inversion, and the mass concentration is bound to the corresponding conductivity fluctuation value to draw the pollutant characteristic spectrum.

[0022] Furthermore, the dynamic division of the spatial topology of the oxidation reaction zone and the membrane separation zone, and the generation of processing domain configuration data including the coordinates of the partition boundaries, includes:

[0023] Read the mass concentration gradient in the pollutant characteristic spectrum, and set the high concentration area as the core region of the oxidation reaction and the low concentration area as the membrane separation buffer zone;

[0024] The partition weighting coefficient is calculated based on the area ratio of the oxidation reaction core region to the membrane separation buffer region.

[0025] Based on the partition weight coefficient, virtual dividing lines are arranged on the preset reactor geometric model, and the curvature of the virtual dividing lines is dynamically adjusted by the fluid shear force prediction model.

[0026] Record the coordinates of the intersection point between the virtual dividing line and the inner wall of the reactor geometric model, and encapsulate them as the processing domain configuration data.

[0027] Furthermore, the invocation of the non-uniform mesh generation engine to construct a discretized computational space with gradient resolution includes:

[0028] Parse the processing domain configuration data and extract the grid encryption requirement levels on both sides of the virtual separator line;

[0029] A high-density mesh generation mode is activated on the side of the oxidation reaction core region to generate tetrahedral units with sub-millimeter side lengths;

[0030] A sparse mesh generation mode is activated on the membrane separation buffer side to generate hexahedral units with side lengths on the order of centimeters.

[0031] A gradual mesh size fusion is performed in the transition zone of the virtual separator line to ensure that the nodes of the high-density mesh and the sparse mesh coincide at the interface, thus completing the construction of the discretized computation space.

[0032] Furthermore, a cross-scale mass transfer simulation is performed within the discretized computational space to track the spatiotemporal evolution of the pollutant concentration field and the oxidant diffusion field, outputting a snapshot of the full flow field concentration distribution, including:

[0033] Import the three-dimensional topological information of the discretized computation space and initialize the flow field velocity boundary conditions and wall slip conditions;

[0034] Simulated oxidant particles are injected, and the coupled equations of the Navier-Stokes equations and the convection-diffusion equations are solved to simulate the migration trajectory of the simulated oxidant particles in the flow field.

[0035] After each preset time step, the amount of pollutants and oxidant residue in each grid cell within the discretized computation space is statistically analyzed.

[0036] The statistical results are rendered as a pseudo-color cloud map and stored in time series to form a snapshot of the concentration distribution of the entire flow field.

[0037] Furthermore, extreme value regions are extracted from the full flow field concentration distribution snapshot, and combined with the historical operating logs of the membrane module, the fouling deposition rate on the membrane surface is estimated, including:

[0038] Traverse the entire flow field concentration distribution snapshot, lock the set of grid cells where the pollutant concentration exceeds the threshold, and mark it as the extreme value region;

[0039] Retrieve the historical operation logs of the membrane module and extract records of the membrane pressure differential rise slope and chemical cleaning cycle under the same influent water quality conditions;

[0040] Establish a correlation mapping table between pollutant flux and membrane pressure gradient rise slope within the extreme value region;

[0041] Based on the aforementioned association mapping table, the slope of the membrane pressure difference increase under the current operating conditions is obtained by looking up the table and converted into the pollutant deposition thickness per unit time, which is defined as the pollutant deposition rate on the membrane surface.

[0042] Further, based on the contamination deposition rate on the membrane surface, the partition boundary coordinates in the processing domain configuration data are reverse-corrected, iterated until convergence, including:

[0043] Determine whether the contaminant deposition rate on the membrane surface exceeds a preset warning threshold;

[0044] If the limit is exceeded, the virtual dividing line in the processing domain configuration data is shifted along the fluid flow direction to expand the coverage of the oxidation reaction core area, thereby reducing the pollutant load entering the membrane separation zone;

[0045] If it does not exceed the limit, maintain the current virtual separator position;

[0046] Based on this, the position of the virtual dividing line is adjusted, and the operations of dynamically dividing the spatial topology, constructing the discretized computational space, performing cross-scale mass transfer simulation, and estimating the membrane surface contaminant deposition rate are re-executed until the change in the membrane surface contaminant deposition rate in three consecutive iterations is less than the set error band, at which point convergence is determined.

[0047] Furthermore, the step of performing a mesh size gradient fusion in the transition zone of the virtual separator line to ensure that the nodes of the high-density mesh and the sparse mesh coincide at the interface includes:

[0048] A grid size scaling factor is defined in the transition zone of the virtual dividing line, and the grid size scaling factor varies exponentially with the distance from the virtual dividing line.

[0049] The base mesh template is stretched or compressed using the mesh size scaling factor to generate a transition mesh layer;

[0050] Boolean operations are performed on the boundary edges between the transition mesh layer and the high-density mesh and sparse mesh to forcibly delete dangling nodes and merge duplicate nodes, thereby achieving node overlap.

[0051] Further, the step of rendering the statistical results into a pseudo-color cloud map and storing it in time series to form the snapshot of the full flow field concentration distribution includes:

[0052] Read the remaining amount of pollutants and oxidant residue in each grid cell and normalize them to a value range of zero to one.

[0053] According to the value range, a preset color lookup table is matched, and each grid cell is assigned a corresponding color code;

[0054] The color-coded grid cells are reorganized according to their spatial coordinates in the discretized computation space to generate a two-dimensional projection view and a three-dimensional volume drawing view.

[0055] The two-dimensional projection view and the three-dimensional volume drawing view are bound to a timestamp and packaged and stored as a snapshot of the full flow field concentration distribution.

[0056] Compared with the prior art, the beneficial effects of the present invention are:

[0057] Multi-source heterogeneous data streams of conductivity fluctuations, UV absorbance, and suspended solids concentration of high-salinity wastewater to be treated are collected. The data streams are then imported into a pretreatment pipeline to perform adaptive filtering and outlier removal, reconstructing a standardized time-series matrix. This process removes abnormal interference information from the data, allowing the water quality data to present regular time-series characteristics. By analyzing the standardized time-series matrix, the dominant pollutant types are identified and the initial load is quantified. The generated pollutant feature map can intuitively reflect the pollutant composition and load status of the wastewater. Based on the pollutant feature map, the spatial topology of the oxidation reaction zone and membrane separation zone is dynamically divided, generating treatment domain configuration data containing the boundary coordinates of the zones. This allows the division of treatment areas to directly match real-time pollutant characteristics, ensuring that the spatial layout of the oxidation reaction and membrane separation zones meets the real-time treatment requirements of the wastewater.

[0058] After reading the processing domain configuration data, the non-uniform mesh generation engine is invoked to construct a gradient resolution discretized computational space that can adapt to the simulation accuracy requirements of different regions. Cross-scale mass transfer simulations are performed within this space, which can completely track the spatiotemporal evolution of pollutant concentration fields and oxidant diffusion fields. The output full flow field concentration distribution snapshot can accurately locate the extreme value regions of pollutant accumulation. By extracting the extreme value regions and combining them with the historical operation logs of the membrane module, the fouling deposition rate on the membrane surface can be accurately calculated. The partition boundary coordinates in the processing domain configuration data are corrected in reverse using this rate. Through iterative convergence, the partition boundaries can be continuously adapted to the water quality and membrane operation status during the treatment process. The final output purification path plan can fit the actual working conditions of real-time treatment of high-salinity wastewater. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating the steps of the high-salt wastewater purification method that couples advanced oxidation and membrane filtration according to the present invention.

[0060] Figure 2 To generate a flowchart for processing domain configuration data;

[0061] Figure 3 This is a time-series evolution analysis diagram of pollutant concentrations;

[0062] Figure 4 Draw views for a 3D volume;

[0063] Figure 5 This is an iterative convergence graph of the deposition rate of contaminants on the membrane surface. Detailed Implementation

[0064] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] See Figure 1This invention provides a method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration. The method includes: starting with the acquisition of a multi-source heterogeneous data stream of the high-salinity wastewater to be treated. This data stream encompasses real-time monitored conductivity fluctuations, ultraviolet absorbance at a specific wavelength, and suspended solids concentration. The acquired multi-source heterogeneous data stream is then input into a pre-processing pipeline. In this pipeline, an adaptive filtering algorithm smooths random noise in the data, and outlier detection and removal rules based on statistical distribution are applied to remove abnormal sampling points. The purified data stream is reconstructed into a standardized time-series matrix for subsequent unified processing and analysis. The system analyzes this standardized time-series matrix, identifies the dominant pollutant types in the wastewater using signal processing and pattern recognition techniques, quantifies their initial load according to relevant models, and finally generates a pollutant feature map that comprehensively reflects the types and spatial distribution characteristics of pollutants. Based on the generated pollutant feature map, the system dynamically divides the spatial topology of the oxidation reaction zone and the membrane separation zone. This process analyzes the concentration gradient information in the spectrum to dynamically determine the optimal boundaries of the two functional zones within the reactor's virtual space, generating processing domain configuration data containing detailed zone boundary coordinates. The system then reads this processing domain configuration data and invokes the embedded non-uniform mesh generation engine. Based on the functional zone division requirements, this engine generates a high-density mesh near the oxidation reaction zone, a sparse mesh in the membrane separation zone, and achieves a smooth gradient in mesh size in the transition region, thereby constructing a discretized computational space with gradient resolution.

[0066] Within this discretized computational space, the system performs cross-scale numerical simulations of mass transfer. This simulation solves the coupled governing equations of fluid dynamics and mass transfer, tracks the evolution of the pollutant concentration field and the diffusion field of the added oxidant over time, and outputs the simulation results at each moment as a snapshot of the overall flow field concentration distribution. From these snapshots, the system extracts extreme regions where the pollutant concentration exceeds a set threshold and, combined with the historical operating logs of the membrane module stored in the database, calculates the fouling deposition rate on the membrane surface under the current simulation conditions through correlation analysis. The system compares the calculated membrane surface fouling deposition rate with a preset warning threshold. If the threshold is exceeded, the system reverses the adjustment of the partition boundary coordinates in the processing domain configuration data, expands the range of the oxidation reaction zone, and restarts the entire process from mesh generation to deposition rate calculation with this new configuration. This process is iterated until the change in the membrane surface fouling deposition rate tends to stabilize after several consecutive iterations, reaching a convergence state. At this point, the system outputs the final determined purification process path plan that can effectively control membrane fouling.

[0067] In one embodiment of the present invention, the collected multi-source heterogeneous data streams of high-salinity wastewater are first time-stamp aligned to unify the time reference of data from sensors with different sampling frequencies and eliminate timing misalignments. In coal mine water treatment scenarios, the collected conductivity fluctuation values ​​can reflect the concentration changes of dissolved inorganic salts such as sulfates and chlorides in the water; ultraviolet absorbance can be used to characterize the load of organic pollutants containing aromatic rings or conjugated double bonds, such as phenols and polycyclic aromatic hydrocarbons produced by coal chemical processes; and suspended solids concentration directly reflects the content of solid particles such as coal powder and rock powder in the water. Through real-time acquisition and standardized processing of these multi-source heterogeneous data streams, the system can accurately sense fluctuations in influent water quality, providing a data foundation for subsequent efficient pretreatment. For example, the system can dynamically adjust the dosage of pretreatment units such as coagulation and flocculation or the sludge discharge cycle based on sudden changes in suspended solids concentration. For the continuous data streams after time-stamp alignment, the system performs sliding window integration operations, calculating the arithmetic mean, statistical variance, and the difference between peak and trough values ​​within each sliding window. These calculation results together constitute a multi-dimensional statistical feature vector. The system maps the multidimensional statistical feature vector to a predefined standard range space. This process normalizes the data through linear scaling and fills in missing sampling points caused by data transmission interruptions using linear interpolation of adjacent data, ultimately generating a complete and well-organized standardized time series matrix. Principal component decomposition is then performed on the standardized time series matrix to separate the original mixed signal into several independent signal components. These components correspond to the main components of the wastewater, such as organic pollutants, inorganic salt ions, and colloidal particles. The energy entropy value of each independent component signal is calculated, and the signal component with the highest energy entropy value is identified as the dominant pollutant type in the current wastewater. Based on the known spectral absorption characteristic curve of this dominant pollutant type, its mass concentration in the original water sample is calculated using an inversion algorithm. This mass concentration value is then correlated and bound with the synchronously acquired conductivity fluctuation value, ultimately creating a pollutant feature map containing information on pollutant type, concentration, and conductivity correlation.

[0068] In practical implementation, preprocessing of multi-source heterogeneous data streams and generation of pollutant feature maps are performed. In one example scenario, the high-salt wastewater to be treated comes from saline organic wastewater discharged from a chemical plant. The multi-source heterogeneous data stream specifically includes conductivity fluctuation values ​​collected once per second by an online conductivity meter, ultraviolet absorbance at a wavelength of 254 nm collected once every two seconds by an online ultraviolet spectrometer, and suspended solids concentration data collected once every five seconds by a laser turbidimeter. In practical implementation, due to the differences in sampling frequencies of the three sensors, the directly acquired raw data are not synchronized in time series, so a timestamp alignment operation must be performed. The timestamp alignment operation uses the time point with the highest sampling frequency, i.e., once per second, as the reference timestamp. A linear interpolation method is used for the ultraviolet absorbance data and suspended solids concentration data to calculate their corresponding values ​​at each reference timestamp, thereby generating a set of aligned data streams with a unified time reference and an equal number of data points. In some embodiments, for the continuous data stream after timestamp alignment, the system performs a sliding window integration operation to extract features. The window width is set to five seconds, covering five consecutive baseline timestamp data points. Within each sliding window, the system calculates the arithmetic mean of all conductivity fluctuations, the statistical variance of all UV absorbance values, and identifies the maximum and minimum suspended solids concentration values, calculating their difference. These three calculations—mean, variance, and peak-to-valley difference—together constitute a multidimensional statistical feature vector representing the water quality state of that window. As the window slides in one-second increments, the system generates a series of multidimensional statistical feature vectors arranged chronologically.

[0069] In its implementation, the system maps multidimensional statistical feature vectors to a predefined standard range space. The standard range space defines the target range for each feature vector; for example, conductivity features are mapped to the [0, 100] interval, ultraviolet features to the [0, 10] interval, and suspended matter features to the [0, 5] interval. The mapping process is accomplished using a linear scaling function. If the original mean conductivity is between 4000 and 5000 microsieverts per centimeter, it is projected to the range of 0 to 100 through a linear transformation. During data acquisition, signal transmission interruptions may lead to missing suspended matter concentration data at a certain reference timestamp. In this case, the system uses linear interpolation to calculate the arithmetic mean of the two most recent valid suspended matter concentration data before and after the missing time point and fills in the missing time point. Through the above linear interpolation and linear scaling, a complete and well-organized standardized time series matrix is ​​finally generated, where rows represent time points and columns represent normalized feature values ​​of different dimensions. The method for generating standardized time series matrices can be specifically applied to the purification and treatment of coal mine water or coal chemical wastewater. In this application scenario, the oxidation reaction zone corresponds to the aeration zone in the coal mine water treatment process, and its main function is to achieve the oxidative degradation of pollutants through aeration. The membrane separation zone corresponds to the sedimentation zone, and its main function is to achieve mud-water separation through gravity sedimentation. Through dynamic control, the synergistic optimization of efficient pretreatment, efficient aeration, and efficient sedimentation in coal mine water treatment can be achieved.

[0070] In some embodiments, principal component decomposition (PCD) is performed on the standardized time series matrix to separate independent component signals. Taking a standardized time series matrix containing 1000 time points and 3 feature dimensions as an example, the PCD algorithm decomposes it into three independent signal components. After decomposition, component one exhibits low-frequency slow fluctuations and is highly correlated with changes in conductivity; component two exhibits mid-frequency fluctuations and is highly correlated with changes in ultraviolet absorbance; and component three exhibits high-frequency random fluctuations and is correlated with changes in suspended matter concentration. In a specific example, these three independent component signals are respectively interpreted as representing changes in inorganic salt ion concentration, changes in organic pollutant concentration, and changes in colloidal particle concentration.

[0071] In practical implementation, the system calculates the energy entropy value of each independent component signal. The calculation of the energy entropy value depends on the energy distribution of the signal. The organic pollutant signal represented by component two has a more dispersed energy distribution across multiple frequency bands, and its calculated energy entropy value is higher than that of components one and three. The system identifies component two, which has the highest energy entropy value, as the dominant pollutant type in the current wastewater. It can be understood that if component one has the highest energy entropy value in another set of data, the system will identify inorganic salt ions as the dominant pollutant type. Based on the spectral absorption characteristic curve of the dominant pollutant type, its mass concentration in the original water sample is calculated. Assuming the dominant pollutant type is identified as phenol, the system calls the pre-stored phenol spectral absorption characteristic curve, which describes the quantitative relationship between absorbance and mass concentration of phenol solution at different wavelengths. At the characteristic wavelength of 254 nm, the mass concentration of phenol is positively correlated with the ultraviolet absorbance value, and this relationship can be calculated using the following formula:

[0072]

[0073] in: This represents the mass concentration of phenol, expressed in milligrams per liter. This represents the ultraviolet absorbance value measured or calculated at a wavelength of 254 nanometers. The molar absorptivity of phenol at a wavelength of 254 nanometers is expressed in liters per mole-cm. The optical path length, in centimeters, represents the optical path length for spectral measurements. The system substitutes the restored UV absorbance data from the standardized time-series matrix into the formula to invert the phenol mass concentration at each time point. It can be understood that the inversion calculation process strictly relies on pre-stored spectral absorption characteristic curves, which differ for different pollutants. In practice, the system binds the calculated phenol mass concentration with the conductivity fluctuation values ​​collected at the corresponding time points. This binding operation establishes a one-to-one correspondence between the two parameters along the time dimension, forming a data pair. The system plots the phenol mass concentration change curve and the conductivity fluctuation value change curve on the same coordinate system with time as the horizontal axis, thereby generating a pollutant characteristic map. This pollutant characteristic map visually displays the concentration trend of the dominant organic pollutant phenol over time and simultaneously reflects the changes in water conductivity, revealing the potential correlation between pollutant load and salinity fluctuations.

[0074] In one embodiment of the present invention, see [reference] Figure 2The system reads the pollutant characteristic spectrum and analyzes the mass concentration gradient distribution presented in the spectrum. It designates areas with concentrations above a set threshold as the oxidation reaction core zone, and surrounding areas with lower concentrations as membrane separation buffer zones. Based on the area ratio of the oxidation reaction core zone to the membrane separation buffer zone within the virtual reaction space, a partitioning weight coefficient is calculated. Based on this partitioning weight coefficient, the system dynamically arranges a virtual dividing line within a preset geometric model representing the actual reactor size. The curvature of this dividing line is not fixed but dynamically adjusted by a fluid shear force prediction model based on the simulated flow field conditions to ensure that the dividing line shape conforms to fluid dynamics principles. Finally, the system records the coordinates of a series of intersection points between this virtual dividing line and the inner wall of the reactor geometric model and encapsulates this coordinate data into structured processing domain configuration data.

[0075] In practical implementation, functional zones are dynamically divided based on pollutant characteristic maps, and treatment domain configuration data is generated. The pollutant characteristic maps read by the system show that the phenol concentration is higher in the left half of the virtual reactor model, fluctuating between 80 and 100 mg / L, while the concentration is lower in the right half, fluctuating between 10 and 30 mg / L. In coal mine water treatment applications, the system dynamically designates areas with high concentrations of organic matter such as phenols as aeration core zones, and areas with high suspended solids concentrations or requiring deep sedimentation as sedimentation buffer zones, based on pollutant characteristic maps. By dynamically adjusting the spatial topology of these two functional zones, the layout of aerators and the distribution of aeration intensity can be optimized, concentrating aeration energy in areas with the highest pollutant load, avoiding ineffective aeration, and thus achieving efficient aeration. The dynamic adjustment of the virtual dividing line is essentially an optimization process for the optimal boundary between the aeration zone and the sedimentation zone. In practical implementation, the system sets the high-concentration area of ​​the mass concentration gradient as the oxidation reaction core zone and the low-concentration area as the membrane separation buffer zone. Specifically, the system sets a concentration threshold of 50 mg / L, classifying all spatial locations in the pollutant characteristic map with phenol concentrations higher than 50 mg / L as the oxidation reaction core region, and all spatial locations with phenol concentrations lower than 50 mg / L as the membrane separation buffer zone. In some embodiments, the system calculates a partitioning weight coefficient based on the area ratio of the oxidation reaction core region to the membrane separation buffer zone. Assuming that the oxidation reaction core region occupies an area of ​​2.5 square meters and the membrane separation buffer zone occupies an area of ​​7.5 square meters based on spatial point statistics, the partitioning weight coefficient is calculated to quantify the impact of the spatial proportion of different functional zones on the subsequent arrangement of dividing lines. Partitioning Weight Coefficient The calculation can be expressed as:

[0076]

[0077] in: The partition weight coefficient is a dimensionless number. The calculated area representing the core region of the oxidation reaction, in square meters; This represents the calculated area of ​​the membrane separation buffer zone, in square meters. Substituting the example data, the area of ​​the oxidation reaction core region... The membrane separation buffer area is 2.5. The value is 7.5, and the partition weight coefficient is calculated. It is 0.25. This is understandable; it represents the partition weighting coefficient. The larger the value, the larger the processing space required for the core region of the oxidation reaction.

[0078] In practical implementation, based on the partition weighting coefficient, the system arranges virtual dividing lines on a preset reactor geometric model. The preset reactor geometric model is a three-dimensional cylindrical model with a length of 5 meters and a diameter of 2 meters. The system uses a calculated partition weighting coefficient of 0.25. Initially, a virtual dividing line was placed on a cross-section 1.25 meters from the reactor inlet. The curvature of the virtual dividing line was dynamically adjusted by a fluid shear force prediction model. Based on preliminary flow field simulation, the fluid shear stress prediction model calculated the distribution of fluid shear stress at different locations within the reactor. It can be understood that the virtual dividing line is not a plane, but a potentially curved surface, and its shape needs to adapt to the flow field to reduce fluid dead zones.

[0079] In some embodiments, the dynamic adjustment process of the curvature of the virtual separator line specifically involves: the fluid shear force prediction model outputs a distribution map of high shear stress regions; the system adjusts the shape of the virtual separator line to fit as closely as possible to the edge of the high shear stress regions, ensuring that the core area of ​​the oxidation reaction is located in a region of intense fluid mixing. For example, in an example scenario, the simulation shows that the shear stress in the central region at the bottom of the reactor is high; therefore, the virtual separator line bulges downstream, forming a parabolic surface rather than a plane. In a specific implementation, the system records the coordinates of the intersection points between the virtual separator line and the inner wall of the reactor geometric model. For a cylindrical reactor model, the virtual separator line, as a curved surface, intersects with the inner wall of the cylinder to form a three-dimensional closed curve. The system samples and records the three-dimensional coordinates (x, y, z) of hundreds of points on this closed curve with millimeter-level precision. The system encapsulates the sequence of all coordinate points of this spatial closed curve, along with the basic dimensional parameters of the reactor model, into a structured processing domain configuration data file. Optionally, the processing domain configuration data file is stored in JSON or XML format to facilitate reading and parsing by subsequent modules.

[0080] In one embodiment of the invention, the system analyzes and processes domain configuration data to extract the differentiated grid resolution requirements of different regions on both sides of the virtual separator line. On the oxidation reaction core side, the system activates a high-density grid generation mode, generating tetrahedral elements with sub-millimeter side lengths to accurately capture the intense chemical reactions and mass transfer processes within this region. On the membrane separation buffer side, the system activates a sparse grid generation mode, generating hexahedral elements with centimeter side lengths, improving overall computational efficiency while maintaining computational accuracy. In the transition zone near the virtual separator line, the system performs a grid size gradient fusion operation to achieve a smooth transition from a high-density grid to a sparse grid. A grid size scaling factor is defined in the transition zone of the virtual separator line, which varies exponentially with the distance from the center point of the computational cell to the virtual separator line. This grid size scaling factor is used to stretch or compress a set of basic grid templates to generate a series of transition grid layers with gradually varying sizes. Subsequently, Boolean operations are performed on the boundary edges between the transition mesh layer and the high-density mesh layer and the sparse mesh layer to forcibly delete the dangling nodes caused by size mismatch and merge the duplicate nodes with overlapping positions, ensuring that the mesh nodes at the interface of different density mesh regions are completely overlapped, thus achieving continuity and consistency in the computational space.

[0081] In practical implementation, a non-uniform mesh generation engine is invoked to construct a discretized computational space, and mesh fusion is performed in the transition zone. In one example scenario, the processing domain configuration data defines a cuboid virtual reactor with a curved virtual dividing line inside, which divides the reactor space into an oxidation reaction core region and a membrane separation buffer region. In practical implementation, the system parses the processing domain configuration data and extracts the mesh refinement requirement level on both sides of the virtual dividing line. The oxidation reaction core region side, due to the expected rapid chemical reaction and mass transfer, is marked as requiring a high-resolution mesh, and the mesh refinement requirement level is set to "high"; the membrane separation buffer region side, mainly for physical separation, is marked as suitable for a lower-resolution mesh, and the mesh refinement requirement level is set to "low". In some embodiments, a high-density mesh generation mode is activated on the oxidation reaction core region side, and the system generates tetrahedral elements with sub-millimeter side lengths. For example, specifying a target side length of 0.5 mm for the tetrahedral elements, the mesh generation algorithm fills the space occupied by the oxidation reaction core region with millions of tiny tetrahedra of uniform shape and size, thereby finely characterizing the geometric features of the region. Activating the sparse mesh generation mode on the membrane separation buffer side generates hexahedral elements with side lengths in the centimeter range. For example, specifying a target side length of 2.0 cm for the hexahedral elements, the mesh generation algorithm will generate a relatively small number of large-sized structured hexahedral meshes within the membrane separation buffer space. A comparison of the key parameters for the two meshing modes is shown in Table 1.

[0082] Table 1: Comparison of Grid Parameters for Different Functional Zones

[0083] ;

[0084] In practice, a gradual mesh size blending is performed within the transition zone of the virtual separator line. The transition zone is defined as the area within a certain distance on either side of the virtual separator line, for example, a 5 cm area on each side of the separator line. A mesh size scaling factor is defined within the transition zone. Grid size scaling factor The distance from the center point of the computing unit to the virtual dividing line It changes exponentially, and the relationship is as follows:

[0085]

[0086] in: This represents the grid size scaling factor, which is a dimensionless scaling factor. This represents the base scaling factor at the location of the virtual separator line, and is set to 1. This represents the scaling gradient coefficient, controlling the drasticness of size changes. Its negative value ensures that the size changes with distance. Increase, Decrease; This represents the vertical distance from the center point of the calculation unit to the virtual dividing line, in meters. hour, The grid size is consistent with the high-density grid reference size; when When the membrane separation buffer is increased, The value decreases exponentially, and the target mesh size increases accordingly.

[0087] In some embodiments, a mesh size scaling factor is used to stretch or compress the base mesh template. The base mesh template is a series of regular tetrahedral elements with a size of 0.5 mm. The system calculates the corresponding mesh size scaling factor based on the position of the center point of each base mesh template element. And multiply the length of each side of the unit by the corresponding... Value, thereby achieving the stretching of the unit ( ) or compression ( This generates a series of transition mesh layers with continuously varying sizes. It can be understood that on the side closer to the core region of the oxidation reaction, the transition mesh size is approximately 0.5 mm; on the side closer to the membrane separation buffer zone, the transition mesh size is approximately 2.0 cm.

[0088] In the specific implementation, Boolean operations are performed on the boundaries between the transitional mesh layer and the high-density and sparse meshes. Since different mesh layers are generated independently at the boundaries, their node positions are not geometrically perfectly aligned, exhibiting slight misalignment and overlap. The system uses Boolean operations to forcibly delete dangling nodes that belong to only one mesh layer and are not connected to nodes in adjacent mesh layers. Simultaneously, the system detects node pairs with identical or extremely close spatial coordinates within the tolerance range and merges these duplicate nodes into a shared node. Optionally, the coordinate tolerance for node merging is set to one-tenth of the minimum mesh size. By performing the operations of deleting dangling nodes and merging duplicate nodes, the nodes at the interfaces of the high-density mesh, transitional mesh layer, and sparse mesh achieve complete overlap, thereby constructing a seamless, node-continuous, and gradient-resolution single discretized computational space.

[0089] See Figure 3 This is a time-series evolution chart of pollutant concentration, primarily used to demonstrate the results of cross-scale mass transfer simulations. The oxidation reaction core zone, as the main site of pollutant degradation, achieves rapid and deep removal of pollutants through advanced oxidation reactions. The initial high concentration decreased by 73% within 4 seconds, demonstrating the high efficiency of the oxidation process. The membrane separation buffer zone, serving as a deep purification and protection stage, undertakes the physical separation of remaining pollutants, resulting in a stable concentration decrease and preventing damage to the membrane module from pollutant surges. The two curves intersect at approximately 4 seconds, indicating that the rapid degradation in the oxidation core zone effectively reduces the pollutant load entering the membrane separation zone. This perfectly validates the design rationality of the "oxidation pretreatment + membrane separation" coupled process. The oxidation process reduces the load on the membrane module, while the membrane process ensures the quality of the effluent. The chart visually demonstrates the pollutant removal effect of the coupled process and quantifies its performance.

[0090] In one embodiment of the present invention, within the constructed discretized computational space, the system imports its three-dimensional topological information and sets velocity boundary conditions for the flow field inlet and outlet, as well as slip boundary conditions for the reactor wall. Simulated particles representing oxidants are injected into the flow field, and the migration and diffusion trajectories of these oxidant particles in the flowing carrier are simulated by numerically solving the coupled equations of the Navier-Stokes equations and the convection-diffusion equations. After calculation at each preset physical time step, the residual pollutant mass and the remaining amount of oxidant in each grid cell within the discretized computational space are statistically analyzed. The statistical results of each time step are rendered as pseudo-color cloud maps and stored in a time series, forming a complete snapshot of the full flow field concentration distribution capable of displaying the spatiotemporal evolution of pollutants and oxidants. The calculated residual pollutant and oxidant values ​​within each grid cell are read and normalized to a value range of zero to one. Based on the normalized values, a preset color lookup table is matched, and each grid cell is assigned a corresponding red, green, and blue color code. All color-coded mesh cells are reassembled and projected according to their coordinates in the actual three-dimensional discretized computation space to generate a two-dimensional cross-sectional projection view and a three-dimensional solid drawing view. These two-dimensional and three-dimensional views are bound to their corresponding simulation timestamps and packaged and stored as a data set, which constitutes the snapshot of the entire flow field concentration distribution.

[0091] In practical implementation, mass transfer simulation is performed within the discretized computational space, and a snapshot of the concentration distribution across the entire flow field is output. When simulating the aeration process in coal mine water, the injected simulated oxidant particles can be characterized as the mass transfer process of oxygen from the air into the wastewater. By solving the flow field and mass transfer equations, the system can simulate the complete process of oxygen being released from the aeration head, diffusing in the wastewater, and reacting with pollutants. The output snapshot of the concentration distribution across the entire flow field can show both the decrease in pollutant concentration and the spatial distribution of dissolved oxygen concentration, thus providing a visual assessment of aeration efficiency and pollutant removal efficiency. In one example scenario, the discretized computational space is constructed based on a cuboid reactor model measuring 2 meters long, 1 meter wide, and 1 meter high, which is divided into approximately one million grid cells by a non-uniform grid generation engine. In practical implementation, the three-dimensional topological information of the discretized computational space is imported. This information includes the geometric coordinates, volume, and connectivity of each grid cell with its adjacent cells. The system initializes the flow field velocity boundary conditions and wall slip conditions. For example, the velocity boundary condition at the reactor inlet is set to uniform inflow with a velocity value of 0.1 m / s; the pressure boundary condition at the reactor outlet is set to 0 Pascal; and the velocity boundary conditions at all reactor walls are set to no slip condition, i.e., the fluid velocity at the wall is zero.

[0092] In some embodiments, the system injects simulated oxidant particles to simulate mass transfer. These simulated oxidant particles are not discrete particle entities but are defined as a concentration field, with an initial simulated oxidant concentration, such as 50 mg / L, set at the reactor inlet. The system solves a coupled set of Navier-Stokes equations and convection-diffusion equations to simulate the migration trajectory. The Navier-Stokes equations are used to solve for the velocity and pressure distributions in the flow field, while the convection-diffusion equations are used to solve for the transport process of the simulated oxidant concentration within the solved flow field. The solution process is performed using the finite volume method on each grid cell, with each preset physical time step set to 0.1 seconds. It is understood that the choice of time step must satisfy the stability conditions of the numerical computation.

[0093] In practical implementation, after each 0.1-second time step, the system statistically analyzes the remaining pollutant and oxidant amounts in each grid cell within the discretized computational space. The remaining pollutant amount is obtained by solving the convection-diffusion reaction equation for the pollutant, while the oxidant amount is obtained by solving the convection-diffusion consumption equation for the oxidant. For a computational model containing approximately one million grid cells, the system records and updates the scalar data in these one million cells at the end of each time step. The statistical workload at different time steps is shown in Table 2.

[0094] Table 2: Statistical Table of Calculation for Each Time Step

[0095] ;

[0096] In practice, rendering the statistical results into a pseudo-color cloud map involves data normalization and color mapping. The system reads the residual pollutant value within each grid cell. This value is then normalized to a range of zero to one. The normalization calculation is performed using the following formula:

[0097]

[0098] in: This represents the normalized value, which is between 0 and 1; This represents the amount of pollutants remaining within a grid cell, expressed in milligrams per liter. This represents the maximum amount of pollutants remaining in the current entire flow field; This represents the minimum amount of pollutants remaining in the current entire flow field. Based on the calculated... The system matches the value to a preset color lookup table. For example, when... When it is close to 0, it matches dark blue. When it is close to 0.5, it matches green. When the value is close to 1.0, it matches dark red and assigns a corresponding red, green, and blue color code to each grid cell.

[0099] In some embodiments, color-coded mesh cells are reassembled according to their spatial coordinates in the discretized computation space to generate a two-dimensional projection view and a three-dimensional volume rendering view. For the two-dimensional projection view, the system projects the three-dimensional space along a coordinate axis, maximizing the size of all mesh cells along the projection path. The values ​​are used as projection points and colored to generate a 2D pseudo-color cross-section. For the 3D volume rendering view, the system uses a ray casting algorithm, treating the entire 3D discretized computational space as a semi-transparent medium. Based on the color encoding and opacity settings of each mesh unit, a 3D volume rendering view is synthesized. It can be understood that a 2D projection view is convenient for observing cross-sectional distribution, while a 3D volume rendering view can display the overall distribution trend within 3D space.

[0100] In practical implementation, the two-dimensional projection view and the three-dimensional volume rendering view are bound to timestamps and packaged and stored as a snapshot of the entire flow field concentration distribution. The two-dimensional projection image and three-dimensional volume rendering image files generated by the system for each simulation time step all use the corresponding simulation time as part of their filenames. Optionally, all time-series image files belonging to the same simulation task are stored in a separate folder, accompanied by an index file recording the correspondence between timestamps and filenames. This folder containing the time-series images and the index file constitutes a complete snapshot of the entire flow field concentration distribution.

[0101] See Figure 4 This is a 3D volumetric view that visually presents the spatial mass transfer simulation results of the oxidation-membrane coupling process. The image clearly demonstrates the gradient resolution design of "high-density grid in the oxidation zone + sparse grid in the membrane zone + gradient grid in the transition zone." The oxidation zone uses sub-millimeter-level grids to accurately capture intense chemical reactions, while the membrane zone uses centimeter-level grids to improve computational efficiency. The transition zone achieves seamless integration, fully complying with the patented technical solution. The concentration cloud map visually presents the spatiotemporal evolution of pollutants from the high-concentration zone in the influent to the low-concentration zone in the membrane zone after oxidation and degradation, providing data support for optimizing the reactor flow field and zoning boundaries. The low-concentration distribution in the membrane separation zone (right side) proves that the oxidation pretreatment effectively reduces the pollutant load entering the membrane module, which can be directly used to calculate the fouling deposition rate on the membrane surface, providing a basis for iterative optimization of the zoning design.

[0102] In one embodiment of the present invention, the system traverses snapshots of the concentration distribution across the entire flow field, identifies grid cell sets where all pollutant concentrations exceed a preset safety threshold, and marks these cells as extreme value regions. Simultaneously, the system retrieves the historical operating log database of the membrane module, extracting the slope of the membrane filtration differential pressure increase over time and the corresponding chemical cleaning cycle from historical records with influent water quality conditions similar to the current simulated operating conditions. A correlation mapping table is established to associate the predicted pollutant flux values ​​within the extreme value regions of the historical data with the actually observed slope of the membrane differential pressure increase. Based on this correlation mapping table, the predicted slope of the membrane differential pressure increase under the current simulated operating conditions is obtained through a lookup operation, and this slope value is converted into the expected deposition thickness of pollutants on the membrane surface per unit time according to an empirical model. This thickness value is defined as the membrane surface fouling deposition rate under the current operating conditions. The system determines whether the calculated membrane surface fouling deposition rate exceeds a preset warning threshold. If the rate exceeds the warning threshold, the virtual separator line defined in the processing domain configuration data is shifted along the main fluid flow direction to expand the spatial coverage of the oxidation reaction core area, thereby increasing the pre-oxidation removal of pollutants in the simulation and reducing the pollutant load entering the membrane separation zone. If the rate does not exceed the warning threshold, the current virtual separator line position remains unchanged. Based on the adjusted virtual separator line position, the system re-executes the complete process from dynamically dividing the spatial topology, to constructing the discretized computational space, performing cross-scale mass transfer simulations, and finally calculating the membrane surface contaminant deposition rate. The above adjustment and simulation process is iterated until, in three consecutive iterations, the change in the membrane surface contaminant deposition rate is less than a set error band, at which point the system iterative calculation is considered converged.

[0103] In practical implementation, key area information is extracted from the full flow field concentration distribution snapshot, and historical data is combined to assess and optimize membrane fouling trends. In one example scenario, the full flow field concentration distribution snapshot records the spatial distribution of pollutant concentrations during the simulation cycle. The snapshot data shows that there is a region with persistently high pollutant concentrations downstream of the reactor, near the membrane module inlet. The system traverses the full flow field concentration distribution snapshots, identifies the set of grid cells where the pollutant concentration exceeds a threshold, and sets the pollutant concentration threshold to 85 mg / L. The system scans the concentration field data at the 8th second of the simulation, identifying all grid cells with concentration values ​​greater than 85 mg / L. These marked grid cell sets constitute extreme value regions, and their spatial locations highly overlap with the actual inlet area of ​​the membrane module. In some embodiments, the system retrieves the historical operating logs of the membrane module, extracting the membrane pressure differential rise slope and chemical cleaning cycle records under the same feed water quality conditions. The historical operating log database contains multiple historical cases. Based on the current simulated feed water characteristics, such as a phenol concentration of approximately 90 mg / L and a conductivity of approximately 20 mSiemens / cm, the system filters out historical records with similar feed water quality. From these selected records, the linear fitting slope of the membrane filtration pressure difference change with time during each operation is extracted, i.e., the slope of the membrane pressure difference increase, which is usually in kilopascals per hour. At the same time, the cumulative operating time from start-up to the need for chemical cleaning is extracted as the chemical cleaning cycle.

[0104] In practical implementation, a correlation mapping table is established between pollutant flux and the slope of membrane pressure gradient within the extreme value region. Based on the currently simulated fluid velocity field and pollutant concentration field, the system calculates the pollutant mass flux in each grid cell of the extreme value region in the direction normal to the membrane surface, and performs an area-weighted average of all flux values ​​across the entire extreme value region to obtain a flux value representing the average impact load of that region. The system retrieves data from historical operation logs that corresponds to the currently calculated... The observed membrane pressure differential rise slopes under historical operating conditions with similar values ​​were analyzed, and multiple sets of such correspondences were compiled into a correlation mapping table. Based on the correlation mapping table, the membrane pressure differential rise slope under the current operating condition was obtained by looking up the table, and then calculated using linear interpolation. The predicted membrane pressure gradient is 22.5 grams per square meter per hour. Approximately 0.55 kPa per hour. It can be understood that the correlation mapping table is a crucial data bridge connecting simulated predictions with actual operational experience. In practical implementation, the predicted membrane pressure differential rise slope is converted into the contaminant deposition thickness per unit time and the contaminant deposition rate on the membrane surface. It is obtained through a transformation relationship, the transformation formula is:

[0105]

[0106] in: This represents the rate of fouling deposition on the membrane surface, expressed in micrometers per hour. The deposition rate conversion factor is an empirical parameter obtained by fitting historical data; for example, its value is 1.8 micrometers per kilopascal. This represents the slope of the predicted membrane pressure gradient obtained from the correlation mapping table, in kilopascals per hour. The calculated... The value represents the quantitative assessment result of the membrane fouling rate under the current simulated operating conditions.

[0107] In some embodiments, the system determines whether the calculated membrane surface fouling deposition rate exceeds a preset warning threshold, which is set to 95 micrometers per hour. If the membrane surface fouling deposition rate exceeds 95 micrometers per hour, the virtual separator line in the processing domain configuration data is shifted along the fluid flow direction. In one iteration, assuming the calculated membrane surface fouling deposition rate is 108 micrometers per hour, exceeding the warning threshold, the system moves the virtual separator line downstream by 0.3 meters, thereby expanding the spatial range of the oxidation reaction core region. If the membrane surface fouling deposition rate does not exceed the warning threshold, the current position of the virtual separator line remains unchanged.

[0108] In practice, based on the adjusted virtual separator line position, the subsequent complete simulation process, starting from the dynamic partitioning of the spatial topology, is re-executed. The system uses the new processing domain configuration data as a starting point, re-grids, simulates the flow field and concentration field, and recalculates the membrane surface fouling deposition rate under the new reaction zone layout. This process iterates cyclically, comparing the latest calculated membrane surface fouling deposition rate with a warning threshold after each iteration to determine whether to continue adjusting the virtual separator line. The iterative process continues until the variation in the membrane surface fouling deposition rate is less than a set error band of 5 micrometers per hour in three consecutive iterations. Optionally, when the membrane surface fouling deposition rates obtained in the most recent three iterations are 94 micrometers per hour, 96 micrometers per hour, and 95 micrometers per hour, respectively, the maximum variation is 2 micrometers per hour, which is less than the 5 micrometers per hour error band, and the system determines that the iterative calculation has converged. Convergence means that a relatively stable processing domain spatial configuration has been found that can control the membrane fouling rate within an acceptable range.

[0109] See Figure 5This is an iterative convergence graph of the fouling deposition rate on the membrane surface. The initial fouling rate was 108 μm / h, far exceeding the warning threshold of 95 μm / h, triggering a zone boundary correction. The system shifted the virtual dividing line along the water flow direction, expanding the core oxidation reaction zone and reducing the pollutant load entering the membrane zone. After three iterations, the fouling rate dropped below the warning value for the first time, with subsequent iterations showing only minor fluctuations. The changes in the fouling rate during iterations 5-7 in the graph are all less than the set error band (approximately ±0.5 μm / h). The blue-marked "convergence region" visually represents the final stable operating condition. This quantitatively verifies the control effect of zoned adaptive optimization on membrane fouling, providing data support for extending membrane module lifespan. It can be directly used to set automatic optimization trigger conditions for process operation, enabling unattended operation of the system and providing a precise basis for determining membrane cleaning cycles, avoiding energy consumption and membrane damage caused by over-cleaning.

[0110] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration, characterized in that, include: Collect multi-source heterogeneous data streams of high-salt wastewater to be treated, the multi-source heterogeneous data streams covering conductivity fluctuation values, ultraviolet absorbance and suspended solids concentration; The multi-source heterogeneous data stream is input into the preprocessing pipeline, and after adaptive filtering and outlier removal, it is reconstructed into a normalized time series matrix. The standardized time series matrix is ​​analyzed to identify the dominant pollutant types and quantify their initial loads, generating a pollutant feature map. Based on the pollutant characteristic map, the spatial topology of the oxidation reaction zone and the membrane separation zone is dynamically divided, and processing domain configuration data containing the partition boundary coordinates is generated. Read the processing domain configuration data, call the non-uniform mesh generation engine, and construct a discretized computational space with gradient resolution; Perform cross-scale mass transfer simulations within the discretized computational space, track the spatiotemporal evolution of the pollutant concentration field and the oxidant diffusion field, and output a snapshot of the concentration distribution of the entire flow field. The extreme value regions in the full flow field concentration distribution snapshot are extracted, and the fouling deposition rate on the membrane surface is estimated by combining the historical operation logs of the membrane module. Based on the membrane surface fouling deposition rate, the partition boundary coordinates in the processing domain configuration data are corrected in reverse, iterating until convergence, and finally outputting the purification path plan for the high-salinity wastewater to be treated.

2. The method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration as described in claim 1, characterized in that, The reconstruction into a normalized time series matrix includes: The multi-source heterogeneous data streams are timestamped to eliminate timing misalignment caused by differences in sensor sampling frequencies; Perform sliding window integration on the aligned data stream to extract the mean, variance, and peak-to-valley difference within the window, forming a multidimensional statistical feature vector. The multidimensional statistical feature vector is mapped to a preset standard range space, and the missing sampling points are filled in by linear interpolation to generate the standardized time series matrix.

3. The method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration as described in claim 2, characterized in that, The process of identifying the dominant pollutant type and quantifying its initial load to generate a pollutant characteristic map includes: Principal component decomposition was performed on the standardized time series matrix to separate independent component signals representing organic matter, inorganic salts, and colloidal particles; Calculate the energy entropy value of each independent component signal, and determine the component with the highest energy entropy value as the dominant pollutant type; Based on the spectral absorption characteristic curves of the dominant pollutant type, its mass concentration in the original water sample is calculated by inversion, and the mass concentration is bound to the corresponding conductivity fluctuation value to draw the pollutant characteristic spectrum.

4. The method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration as described in claim 3, characterized in that, The dynamic division of the spatial topology of the oxidation reaction zone and the membrane separation zone, and the generation of processing domain configuration data including the coordinates of the partition boundaries, include: Read the mass concentration gradient in the pollutant characteristic spectrum, and set the high concentration area as the core region of the oxidation reaction and the low concentration area as the membrane separation buffer zone; The partition weighting coefficient is calculated based on the area ratio of the oxidation reaction core region to the membrane separation buffer region. Based on the partition weight coefficient, virtual dividing lines are arranged on the preset reactor geometric model, and the curvature of the virtual dividing lines is dynamically adjusted by the fluid shear force prediction model. Record the coordinates of the intersection point between the virtual dividing line and the inner wall of the reactor geometric model, and encapsulate them as the processing domain configuration data.

5. The method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration as described in claim 4, characterized in that, The invocation of the non-uniform mesh generation engine to construct a discretized computational space with gradient resolution includes: Parse the processing domain configuration data and extract the grid encryption requirement levels on both sides of the virtual separator line; A high-density mesh generation mode is activated on the side of the oxidation reaction core region to generate tetrahedral units with sub-millimeter side lengths; A sparse mesh generation mode is activated on the membrane separation buffer side to generate hexahedral units with side lengths on the order of centimeters. A gradual mesh size fusion is performed in the transition zone of the virtual separator line to ensure that the nodes of the high-density mesh and the sparse mesh coincide at the interface, thus completing the construction of the discretized computation space.

6. The method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration as described in claim 5, characterized in that, Perform cross-scale mass transfer simulations within the discretized computational space, track the spatiotemporal evolution of the pollutant concentration field and the oxidant diffusion field, and output a snapshot of the full flow field concentration distribution, including: Import the three-dimensional topological information of the discretized computation space and initialize the flow field velocity boundary conditions and wall slip conditions; Simulated oxidant particles are injected, and the coupled equations of the Navier-Stokes equations and the convection-diffusion equations are solved to simulate the migration trajectory of the simulated oxidant particles in the flow field. After each preset time step, the amount of pollutants and oxidant residue in each grid cell within the discretized computation space is statistically analyzed. The statistical results are rendered as a pseudo-color cloud map and stored in time series to form a snapshot of the concentration distribution of the entire flow field.

7. The method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration as described in claim 6, characterized in that, Extracting the extreme value regions from the full flow field concentration distribution snapshot and combining them with the historical operating logs of the membrane module, the fouling deposition rate on the membrane surface is estimated, including: Traverse the entire flow field concentration distribution snapshot, lock the set of grid cells where the pollutant concentration exceeds the threshold, and mark it as the extreme value region; Retrieve the historical operation logs of the membrane module and extract records of the membrane pressure differential rise slope and chemical cleaning cycle under the same influent water quality conditions; Establish a correlation mapping table between pollutant flux and membrane pressure gradient rise slope within the extreme value region; Based on the aforementioned association mapping table, the slope of the membrane pressure difference increase under the current operating conditions is obtained by looking up the table and converted into the pollutant deposition thickness per unit time, which is defined as the pollutant deposition rate on the membrane surface.

8. The method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration as described in claim 7, characterized in that, Based on the membrane surface contamination deposition rate, the partition boundary coordinates in the processing domain configuration data are reverse-corrected, iterated until convergence, including: Determine whether the contaminant deposition rate on the membrane surface exceeds a preset warning threshold; If the limit is exceeded, the virtual dividing line in the processing domain configuration data is shifted along the fluid flow direction to expand the coverage of the oxidation reaction core area, thereby reducing the pollutant load entering the membrane separation zone; If it does not exceed the limit, maintain the current virtual separator position; Based on this, the position of the virtual dividing line is adjusted, and the operations of dynamically dividing the spatial topology, constructing the discretized computational space, performing cross-scale mass transfer simulation, and estimating the membrane surface contaminant deposition rate are re-executed until the change in the membrane surface contaminant deposition rate in three consecutive iterations is less than the set error band, at which point convergence is determined.

9. The method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration as described in claim 8, characterized in that, The step of performing a gradual mesh size blending in the transition zone of the virtual separator line to ensure that the nodes of the high-density mesh and the sparse mesh coincide at the interface includes: A grid size scaling factor is defined in the transition zone of the virtual dividing line, and the grid size scaling factor varies exponentially with the distance from the virtual dividing line. The base mesh template is stretched or compressed using the mesh size scaling factor to generate a transition mesh layer; Boolean operations are performed on the boundary edges between the transition mesh layer and the high-density mesh and sparse mesh to forcibly delete dangling nodes and merge duplicate nodes, thereby achieving node overlap.

10. The method for purifying high-salinity wastewater by coupling advanced oxidation and membrane filtration as described in claim 6, characterized in that, The step of rendering the statistical results into a pseudo-color cloud map and storing it in time series to form the snapshot of the full flow field concentration distribution includes: Read the remaining amount of pollutants and oxidant residue in each grid cell and normalize them to a value range of zero to one. According to the value range, a preset color lookup table is matched, and each grid cell is assigned a corresponding color code; The color-coded grid cells are reorganized according to their spatial coordinates in the discretized computation space to generate a two-dimensional projection view and a three-dimensional volume drawing view. The two-dimensional projection view and the three-dimensional volume drawing view are bound to a timestamp and packaged and stored as a snapshot of the full flow field concentration distribution.

Images