Continuous self-adaptive calibration method and system for three-phase unbalanced migration parameters of river

Abstract

The present application relates to river water quality simulation technical field, specifically disclose river three-phase non-equilibrium migration parameter continuous self-adaptive calibration method and system, method includes: obtaining water phase concentration, suspended particle phase concentration, surface sediment concentration and the like, current sediment phase measured inventory is obtained by calculation, and the model calculated sediment inventory is obtained by using three-phase migration model simultaneously; the inventory deviation loss is calculated, and the automatic differentiation update is carried out to normalized kinetic parameter, and the parameter to be screened is obtained; based on the mass conservation residual limit value, the conservation screening parameter is obtained according to the parameter to be screened; the conservation screening parameter is taken as the regression target, and the particle settling parameter and resuspension parameter in the normalized kinetic parameter are updated according to the normalized hydrodynamic state change of adjacent time window, and the evolution parameter is obtained.The present application avoids the parameter mutation caused by discrete flow state switching, and improves the continuity, stability and checkability of model output in continuous hydrological process.

Description

Technical Field

[0001] This invention relates to the field of river water quality simulation technology, and in particular to a continuous adaptive calibration method and system for river three-phase non-equilibrium migration parameters. Background Technology

[0002] Pollutants in riverine environments undergo non-equilibrium migration processes such as adsorption, desorption, sedimentation, and resuspension between the aqueous phase, suspended particulate phase, and sediment phase. The three-phase migration model is a core tool for quantitatively characterizing this process. The kinetic parameters in the model (such as aqueous-particulate exchange parameters, particle sedimentation parameters, and resuspension parameters) change continuously with the hydrodynamic state, and their accurate assignment directly affects the reliability of simulation and prediction.

[0003] Existing parameter calibration methods typically search for optimal parameter combinations based on pre-defined hydrological flow regimes (such as normal flow, rising water, receding water, or event flow). In practical applications, these methods suffer from the following drawbacks: First, hydrodynamic states evolve continuously over time. Switching parameters according to discrete flow regime labels can easily lead to non-physical abrupt changes in parameters between adjacent calculation windows, resulting in discontinuous jumps in model output near flow regime boundaries. Second, traditional parameter calibration methods often use minimizing the observation error of water phase concentration as the objective function, failing to fully utilize the crucial information of sedimentary facies inventory, which has a cumulative constraint effect on the three-phase distribution. This results in parameter combinations that, while able to fit water phase concentration well, cannot accurately reflect the long-term mass accumulation or release trend of sedimentary facies. Third, while some existing studies incorporate mass conservation into parameter selection criteria, they often lack clear definitions of mass state benchmarks and time windows, or continue to update parameters during periods without effective observational data, leading to insufficient verifiability of the model output.

[0004] Therefore, how to update the three-phase non-equilibrium migration parameters under continuously changing hydrodynamic boundary conditions while simultaneously having verifiable sediment inventory basis, executable mass conservation screening conditions, and a freezeable anomaly handling path is an urgent problem to be solved in this field. Summary of the Invention

[0005] This invention aims to solve the above-mentioned problems. To this end, this invention provides a continuous adaptive calibration method and system for river three-phase non-equilibrium migration parameters. This invention uses sediment inventory deviation to drive parameter updates, uses mass conservation residuals to screen candidate parameters, and uses hydrodynamic state to continuously drive parameter evolution, avoiding parameter abrupt changes caused by discrete flow state switching, and improving the continuity, stability, and verifiability of model output in continuous hydrological processes.

[0006] This invention provides a continuous adaptive calibration method for non-equilibrium migration parameters of river three phases, the technical solution of which is as follows: (The method includes...) S1: Obtain historical data on the concentration of water phase, suspended particulate phase, surface sediment concentration, surface sediment thickness, dry bulk density, sampling area, hydrodynamic boundary, discharge load and upstream boundary concentration of the receiving river, and calibrate to obtain reference inventory, reference mass, hydrodynamic normalized reference, mass conservation residual limit, normalized kinetic parameters and their physical ranges. S2: Obtain the concentration of water phase, concentration of suspended particles phase, concentration of surface sediment, thickness of surface sediment, dry bulk density, sampling area, hydrodynamic boundary, discharge load and upstream boundary concentration, calculate the current sediment phase inventory, and use the three-phase migration model to calculate the model-calculated sediment inventory. S3: Calculate the inventory deviation loss based on the reference inventory, the model-calculated sediment inventory, and the current measured sedimentary facies inventory, and perform automatic differential updates on the normalized kinetic parameters to obtain the parameters to be screened; S4: Based on the parameters to be screened, the three-phase migration model is executed again, and the three-phase mass conservation residual is calculated in conjunction with the reference mass. If it is less than or equal to the mass conservation residual limit, the parameter to be screened is used as the conservation screening parameter; otherwise, multiple candidate parameter groups are generated based on the parameters to be screened, and the corresponding three-phase mass conservation residuals are calculated for each group. Within the physical range of the parameters, the candidate parameter group whose mass conservation residual does not exceed the mass conservation residual limit and whose distance from the parameter to be screened is the smallest is selected as the conservation screening parameter. S5: The normalized hydrodynamic state is calculated based on the hydrodynamic boundary and the normalized hydrodynamic reference. The particle settling parameter and resuspension parameter in the normalized dynamic parameters are updated according to the change of the normalized hydrodynamic state in adjacent time windows to obtain the evolution parameters. The evolution parameters are used to back-substitute to the three-phase migration model of the next time window.

[0007] Furthermore, in step S3, the formula for calculating inventory deviation loss is: in, For the inventory deviation loss of the i-th segment of the receiving river; This represents the measured sedimentary facies inventory for the i-th river segment. Calculate the sediment stock for the model of the i-th river segment. This is the reference inventory for the i-th river segment.

[0008] Furthermore, in step S3, the formula for calculating the updated value of the normalized kinetic parameter of type k to be screened is: in, This is the k-th parameter to be filtered for the i-th river segment; For the i-th river segment, the k-th type of current normalized dynamic parameters; Let be the update step size for the k-th type of parameter; for right The gradient value.

[0009] Furthermore, the normalized kinetic parameters include: water-particle phase exchange parameters, particle settling parameters, resuspension parameters, and sediment-water phase desorption parameters.

[0010] Furthermore, in step S4, the three-phase mass conservation residual is the difference between the total three-phase mass change and the algebraic sum of the input mass, output mass, degradation loss mass, and gas-water exchange loss mass, and then divided by the reference mass.

[0011] Furthermore, in step S4, if the three-phase mass conservation residual is greater than the mass conservation residual limit, multiple candidate parameter groups are generated within the physical range of each parameter by enumeration or iteration based on the parameter to be screened, and each candidate parameter group is re-substituted into the three-phase migration model to calculate the corresponding three-phase mass conservation residual; among the candidate parameter groups whose three-phase mass conservation residual is less than or equal to the mass conservation residual limit, the candidate parameter group with the smallest distance from the parameter to be screened is selected as the conservation screening parameter.

[0012] Furthermore, in step S5, the normalized hydrodynamic state is the weighted sum of the two quotients obtained by dividing the flow velocity and water depth by the normalized reference value of the flow velocity and the normalized reference value of the water depth, respectively. The hydrodynamic boundary includes flow velocity and water depth; the hydrodynamic normalization reference includes normalized reference values ​​for flow velocity and water depth.

[0013] Furthermore, the update relationship between particle settling parameters and resuspension parameters is expressed as follows: in, The parameters for the next time window of the k-th class in the i-th river segment; For the i-th river segment, the k-th type of current normalized dynamic parameters; For the k-th type of conservation screening parameters in the i-th river segment; The first response coefficient of the k-th parameter in the i-th river segment; The second response coefficient for the k-th parameter of the i-th river segment; It represents the normalized hydrodynamic state increment for the adjacent time window of the i-th river segment.

[0014] Furthermore, the particle settling parameters and resuspension parameters continuously regress to the conservation screening parameters based on the normalized hydrodynamic state changes in adjacent time windows, without performing discrete parameter switching according to normal flow, rising water, receding water, or event flow.

[0015] This invention also provides a continuous adaptive calibration system for river three-phase nonequilibrium migration parameters, the technical solution of which includes: The parameter calibration unit is used to acquire historical data on the concentration of water phase, suspended particulate phase, surface sediment concentration, surface sediment thickness, dry bulk density, sampling area, hydrodynamic boundary, discharge load and upstream boundary concentration of the receiving river, and to calibrate the reference inventory, reference mass, hydrodynamic normalized reference, mass conservation residual limit, normalized kinetic parameters and their physical ranges. The three-phase migration calculation unit is used to obtain the concentration of water phase, suspended particle phase, surface sediment concentration, surface sediment thickness, dry bulk density, sampling area, hydrodynamic boundary, discharge load and upstream boundary concentration, and calculate the current sediment phase measured stock. At the same time, the model-calculated sediment stock is obtained by using the three-phase migration model. The automatic differential update unit is used to calculate the inventory deviation loss based on the reference inventory, the sediment inventory calculated by the model and the current measured sedimentary facies inventory, and to perform automatic differential update on the normalized kinetic parameters to obtain the parameters to be screened. The conservation screening unit is used to re-execute the three-phase migration model based on the parameters to be screened, and calculate the three-phase mass conservation residuals in conjunction with the reference mass. If the residuals are less than or equal to the mass conservation residual limit, the parameters to be screened are used as conservation screening parameters; otherwise, multiple candidate parameter groups are generated based on the parameters to be screened, and the corresponding three-phase mass conservation residuals are calculated for each group. Within the physical range of the parameters, the candidate parameter group whose mass conservation residuals do not exceed the mass conservation residual limit and whose distance from the parameters to be screened is the smallest is selected as the conservation screening parameters. The parameter evolution unit is used to calculate the normalized hydrodynamic state based on the hydrodynamic boundary and the hydrodynamic normalized reference. With the conservation screening parameters as the regression target, the particle settling parameters and resuspension parameters in the normalized dynamic parameters are updated according to the changes in the normalized hydrodynamic state in adjacent time windows to obtain the evolution parameters. The evolution parameters are used to substitute back into the three-phase migration model of the next time window.

[0016] The above-described one or more technical solutions in the embodiments of the present invention have at least one of the following technical effects: 1. This invention constructs a loss function by calculating the normalized deviation between the sediment stock and the measured stock of the current sediment facies under the same river segment and time window. The parameter gradient direction is obtained through automatic differentiation, allowing parameter updates to directly respond to changes in the mass state of the sediment facies. Compared to calibration methods that rely solely on water concentration errors, the sediment stock has a cumulative constraint effect on the three-phase non-equilibrium exchange process, effectively reflecting the long-term drift of sedimentation and resuspension parameters, thus improving the reliability of parameter updates.

[0017] 2. This invention re-substitutes the parameters to be screened obtained by automatic differentiation into the three-phase migration model, calculates the residual of the total mass change and the algebraic sum of external source and sink terms within the same time window of the three phases relative to the reference mass, and retains only the parameter group whose residual does not exceed the preset limit as the conservation screening parameter; it screens out parameter combinations that cause mass non-conservation due to parameter out-of-bounds or numerical integration cumulative error, ensuring that the mass state output by the model can be measured and closed within the same river section and the same time window, and significantly improves the verifiability of the simulation results.

[0018] 3. This invention uses the conservation screening parameters as the regression target and continuously updates the particle settling and resuspension parameters based on the normalized hydrodynamic state increment of adjacent time windows. It does not perform discrete parameter switching according to the traditional labels of normal flow, rising water or event flow. This allows the parameters to transition smoothly with the hydrodynamic state, eliminates the parameter jump phenomenon at the flow regime boundary, and improves the continuity and stability of the model output in continuous hydrological processes.

[0019] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the method provided by the present invention.

[0022] Figure 2 This is a continuous evolution trajectory diagram of the sedimentation rate constant of suspended particles provided by the present invention.

[0023] Figure 3 This is a comparison chart of the convergence of the three-phase mass conservation residuals provided by the present invention.

[0024] Figure 4 This is a comparison chart of inventory deviation losses provided by the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. The following embodiments are used to illustrate this invention but should not be used to limit the scope of this invention.

[0026] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0027] The following is combined with Figures 1 to 4 The present invention will be further described in detail below, including a continuous adaptive calibration method and system for river three-phase nonequilibrium migration parameters: In this embodiment, as Figure 1 As shown, a continuous adaptive calibration method for the three-phase nonequilibrium migration parameters of a river is provided, comprising the following steps: S1: Obtain historical data on the concentration of water phase, suspended particulate phase, surface sediment concentration, surface sediment thickness, dry bulk density, sampling area, hydrodynamic boundary, discharge load, and upstream boundary concentration of the receiving river, and calibrate to obtain reference inventory, reference mass, hydrodynamic normalized reference, mass conservation residual limit, normalized kinetic parameters, and their physical ranges.

[0028] This embodiment is applied to a receiving river monitoring and calculation system with river segment location information. Multiple calculation river segments are divided along the river's length. Each record includes the river segment identifier, start and end times of the time window, sampling depth, quality control identifier, water phase concentration, suspended particulate phase concentration, surface sediment concentration, surface sediment thickness, dry bulk density, sampling area, flow velocity, water depth, flow rate, discharge load, and upstream boundary concentration.

[0029] The hydrodynamic boundary includes velocity, depth, and flow rate. Velocity and depth are used to construct the normalized hydrodynamic state, while flow rate is used to calculate the output rate constant and upstream mass input.

[0030] The mass of pollutants in the aqueous phase can be calculated based on the concentration and volume of the water phase. Similarly, the mass of pollutants in the suspended particulate phase can be calculated based on the concentration and volume of the suspended particulate phase. The measured sediment inventory is derived from the surface sediment concentration, surface sediment thickness, dry bulk density, and sampling area. Only records with consistent river section identification, time window, and quality control identification can be used as the three-phase state input for the same calculation period.

[0031] Discharge load refers to the mass rate of pollutants discharged from external pollution sources within the calculated river segment, including point and non-point sources. It is physically divided into aqueous discharge load and suspended particulate discharge load, which are respectively incorporated into the aqueous and particulate phase inputs of the three-phase migration model. Upstream boundary concentration refers to the pollutant concentration in the inflow from the adjacent upstream segment. It is also divided into aqueous upstream boundary concentration and particulate upstream boundary concentration, which are multiplied by the flow rate to obtain the upstream input mass rate. Based on the discharge load and upstream boundary concentration, the aqueous pollutant input mass rate and the suspended particulate pollutant input mass rate can be calculated.

[0032] In this embodiment, the upstream boundary concentration and emission load are both taken from the measured records corresponding to the time window, which are external input data and do not depend on the model calculation results of the upstream river segment. Therefore, each calculation river segment independently performs inventory deviation loss calculation, gradient update, conservation screening, and parameter evolution with the measured values ​​of the upstream boundary concentration and emission load as input within the same time window, and the parameter updates between river segments are uncoupled.

[0033] Before the first parameter update, calibration was performed on each river segment to establish the following benchmarks: (a) Reference stock: The median of the effective values ​​of the measured stock of all sedimentary phases in the river segment during the calibration period, or a stock benchmark value not less than the preset measurement lower limit, was taken to avoid outliers from skewing the benchmark during extreme hydrological events. (b) Reference mass: The total mass of the three phases (water phase, suspended particulate phase, and sedimentary phase) at the beginning of the time window was taken. The reference mass was taken independently for each time window, and the total mass of the three phases at the beginning of each effective time window was used as the reference mass benchmark for that time window. (c) Hydrodynamic normalized reference: This included normalized reference values ​​for velocity and water depth, taken as the median of velocity and water depth for all effective time windows during the calibration period, to approximate the hydrodynamic conditions corresponding to the average annual water level. (d) Normalized kinetic parameters include water-particle phase exchange parameters (water-to-suspended particle phase transfer rate constant and suspended particle-to-water phase transfer rate constant), particle settling parameters (suspended particle settling rate constant), resuspension parameters (sediment resuspension rate constant), and sediment-water phase desorption parameters (sediment-to-water phase desorption rate constant). These normalized kinetic parameters are obtained by dividing each parameter by its corresponding baseline value determined during the calibration phase, and then performing dimensionless normalization. Recalibration is performed when there are substantial changes in mineral composition, pollutant types, river boundary, or monitoring section layout. In this embodiment, the quantitative criteria for substantial changes include: a change rate in the dominant mineral composition of the riverbed exceeding 20%; a replacement of pollutant types or a change in pollutant concentration exceeding one order of magnitude; a modification of the river boundary topography exceeding one average river width; or a shift in the monitoring section location exceeding one average river width. The physical parameter range of the normalized kinetic parameters: the lower limit is zero or a very small positive number to satisfy the non-negativity requirement of the rate constant; the upper limit is determined according to the theoretical limiting rate of the corresponding physical process, by dividing each limiting rate by the reference value of the corresponding calibration stage. (e) Mass conservation residual limit: the mass conservation residual is calculated one by one over all effective time windows using the calibration reference parameter set (i.e., all normalized parameters = 1.0), and the 95th percentile of the absolute value of the residual is taken. If this percentile is less than 0.01 (i.e., the mass conservation closure accuracy is better than 1%), then 0.01 (lower limit threshold) is taken.

[0034] Among the above benchmarks, the benchmark values ​​of the normalized kinetic parameters during the calibration phase are determined as follows and fixed after calibration. These values ​​do not change with updates to the normalized kinetic parameters during operation, thus ensuring that the normalization calculation does not depend on subsequent iteration results: Using the measured water phase concentration, suspended particle phase concentration, and measured sediment phase inventory of the river section during the calibration period as observations, and aiming to minimize the sum of inventory deviation losses across all effective time windows during the calibration period, offline calibration is performed on the water phase to suspended particle phase transfer rate constant, suspended particle phase to water phase transfer rate constant, suspended particle settling rate constant, sediment resuspension rate constant, and sediment phase to water phase desorption rate constant. The calibration results are taken as benchmark values. In this embodiment, the benchmark values ​​are: water phase to suspended particle phase transfer rate constant 0.20 / d, suspended particle phase to water phase transfer rate constant 0.15 / d, suspended particle settling rate constant 0.5 / d, sediment resuspension rate constant 0.4 / d, and sediment phase to water phase desorption rate constant 0.1 / d.

[0035] In this embodiment, the three-phase migration model can be expressed as: Changes in water phase quality: Changes in the mass of suspended particulate phases: Sedimentary facies quality changes: in, Let t represent the mass of pollutants in the aqueous phase and t represent time. The rate constant for the transfer of water phase to suspended particulate phase. The rate constant for the transfer of suspended particles to the aqueous phase is... The mass of pollutants in the suspended particulate phase. Let be the rate constant of sediment desorption from the water phase. The mass of pollutants in the sedimentary phase, Let $\frac{ ... For the mass rate of input of aqueous pollutants, The pollutant degradation rate constant, The gas-water exchange rate constant, Let be the settling rate constant of suspended particles. Let be the sediment resuspension rate constant. The output rate constant of the suspended particulate phase. The mass rate of suspended particulate pollutants is input. The three-phase migration model employs a fourth-order Runge-Kutta method for integration, with an integration step size no greater than 1 / 100 of the time window length, determined through line search during the calibration phase. Automatic differentiation constructs a differentiable computational graph based on the above equations, and backpropagation is performed on the inventory deviation loss relative to each normalized kinetic parameter to obtain the gradient.

[0036] S2: Obtain the water phase concentration, suspended particle phase concentration, surface sediment concentration, surface sediment thickness, dry bulk density, sampling area, hydrodynamic boundary, discharge load and upstream boundary concentration within the time window, calculate the current sediment phase inventory, and simultaneously use the three-phase migration model to calculate the model-calculated sediment inventory.

[0037] Within each effective time window, initial values ​​of pollutant mass in the aqueous phase, suspended particulate phase, and sediment phase are calculated based on aqueous phase concentration, suspended particulate phase concentration, surface sediment concentration, surface sediment thickness, hydrodynamic boundary, discharge load, and upstream boundary concentration. These initial values, along with the pollutant input mass rate in the aqueous phase, the pollutant input mass rate in the suspended particulate phase, the aqueous phase output rate constant, and the suspended particulate phase output rate constant, are used as input parameters for the three-phase migration model. The model calculates the mass changes in the aqueous, suspended particulate, and sediment phases, as well as the external source-sink terms and the model-calculated sediment inventory. The calculation of external source-sink terms for the river segment: input mass is determined by the discharge load and upstream boundary concentration and accumulated by the three-phase migration model during integration; output mass, degradation loss mass, and gas-water exchange loss mass are all calculated and accumulated in real-time by the three-phase migration model; external source-sink terms allow for alignment of mass status and boundary flux within the same river segment. The model-calculated sediment inventory represents the pollutant mass in the sediment phase at the end of the time window. If the river segment identifier, time window, or quality control identifier of the input record is inconsistent, the update will not be performed in this time window and the process will proceed to the frozen branch.

[0038] In this step, only the sediment stock calculated by the model is used in the calculation of subsequent steps from the parameters output by the three-phase migration model.

[0039] S3: Calculate the inventory deviation loss based on the reference inventory, the sediment inventory calculated by the model, and the measured inventory of the current sedimentary facies. Then, perform automatic differential updates on the normalized kinetic parameters to obtain the parameters to be screened.

[0040] The inventory deviation loss is determined by the square of the difference between the measured and model-calculated sediment inventory of the current sedimentary facies within the same river segment and time window, relative to a reference inventory. Normalized kinetic parameters include water-particle phase exchange parameters, particle settling parameters, resuspension parameters, and sediment-water phase desorption parameters. The parameters to be screened are updated from the normalized kinetic parameters along the gradient descent direction of the inventory deviation loss.

[0041] The formula for calculating inventory deviation loss is: in, For the inventory deviation loss of the i-th segment of the receiving river; This represents the measured sedimentary facies inventory for the i-th river segment. Calculate the sediment stock for the model of the i-th river segment. This represents the reference inventory for the i-th river segment, with units in kilograms. Take the measured value at the end of the time window and compare it with the sediment inventory calculated by the model. The samples are taken at the same time to ensure direct subtraction. If the sampling time of the actual measurement record is not at the end of the time window, it is calculated by linear interpolation along time from adjacent measured values ​​to the end of the time window, or the measured value closest to that time is taken. Calculation The surface sediment concentration, surface sediment thickness, dry bulk density, and sampling area used were obtained from records of the same sampling time point and the same sampling depth, and their river section identification, time window, and quality control identification were consistent.

[0042] Backpropagation is performed on the inventory deviation loss relative to the normalized kinetic parameters to obtain the corresponding automatic differential gradient. The formula for calculating the pending update value of the k-th class of normalized kinetic parameters is as follows: in, This is the k-th parameter to be filtered for the i-th river segment; For the i-th river segment, the k-th type of current normalized dynamic parameters; Let be the update step size for the k-th type of parameter; for right The gradient value. Also known as the learning rate, it is determined by line search during the calibration phase. Its value is such that the sum of inventory deviation losses for all effective time windows during the calibration period decreases monotonically. In this embodiment, the update step size of various parameters is 0.05, and the value range is from 0.02 to 0.1.

[0043] S4: Based on the parameters to be screened, the three-phase migration model is executed again, and the three-phase mass conservation residual is calculated in conjunction with the reference mass. If it is less than or equal to the mass conservation residual limit, the parameter to be screened is used as the conservation screening parameter; otherwise, multiple candidate parameter groups are generated based on the parameters to be screened, and the corresponding three-phase mass conservation residuals are calculated for each group. Within the physical range of the parameters, the candidate parameter group whose mass conservation residual does not exceed the mass conservation residual limit and whose distance from the parameter to be screened is the smallest is selected as the conservation screening parameter.

[0044] The parameters of the three-phase migration model are updated using the parameters to be screened. Then, the mass changes of each phase and the external source-sink terms of the river section (cumulative input mass, output mass, degradation loss mass and gas-water exchange loss mass within the same time window) are calculated again using the data obtained in step S2.

[0045] The three-phase mass conservation residual is determined by the difference between the total three-phase mass change and the algebraic sum of the input mass, output mass, degradation loss mass, and gas-water exchange loss mass within the same time window, relative to a reference mass. The three-phase mass conservation residual is used to verify the cumulative error of numerical integration and the degree of closure of external source-sink terms within the discrete time step. Essentially, it is a numerical closure test quantity, used to filter out parameter combinations that are not numerically conserved due to parameter out-of-bounds errors or excessive integration errors.

[0046] To determine whether the parameters to be screened can close the mass state within the same time window, the formula for calculating the three-phase mass conservation residual is: in, Let be the three-phase mass conservation residual for the i-th river segment; The total mass change of the three phases in the i-th river segment is obtained by summing the mass changes of the three phases. Let i be the input quality for the i-th river segment. For the output quality of the i-th river segment, Let be the degradation loss mass of the i-th river segment. Let be the mass of air-water exchange loss in the i-th river segment; This is the reference mass for the i-th river segment. Each mass item ( , , and The parameters to be screened are substituted into the three-phase migration model and accumulated within the same river segment and the same time window.

[0047] If the three-phase mass conservation residual is less than or equal to the mass conservation residual limit, then the parameter to be screened is used as the conservation screening parameter. In this embodiment, the parameter to be screened that cannot close the mass state is screened out by comparing the three-phase mass conservation residual with the mass conservation residual limit.

[0048] If the three-phase mass conservation residual is greater than the mass conservation residual limit, multiple candidate parameter groups are generated within the physical range of each parameter by enumeration or iteration based on the parameter to be screened. Each candidate parameter group is then re-substituted into the three-phase migration model to calculate the corresponding three-phase mass conservation residual. Among the candidate parameter groups whose three-phase mass conservation residual is less than or equal to the mass conservation residual limit, the candidate parameter group with the smallest distance to the parameter to be screened is selected as the conservation screening parameter. In this embodiment, a Latin hypercube iteration strategy is used. The sample size of the Latin hypercube sampling is set to 10 to 50 times the normalized dynamic parameter dimension, and the dimension division adopts equal probability stratification. Euclidean distance is used for distance calculation. When the parameter to be screened exceeds the physical range, the nearest projection point of the parameter on the legal boundary is used as the distance calculation benchmark.

[0049] The conservation screening parameters are determined within the upper and lower limits of the normalized parameters, and the three-phase mass conservation residual does not exceed the mass conservation residual limit. When there is no parameter group that meets the conditions, the parameter update of the current time window is cancelled and frozen to the conservation screening parameters of the previous valid time window.

[0050] S5: The normalized hydrodynamic state is calculated based on the hydrodynamic boundary and the normalized hydrodynamic reference. The particle settling parameter and resuspension parameter in the normalized dynamic parameters are continuously updated according to the change of the normalized hydrodynamic state in adjacent time windows to obtain the evolution parameters. The evolution parameters are used to back-substitute to the three-phase migration model of the next time window.

[0051] The normalized hydrodynamic state is the weighted sum of the quotients obtained by dividing the flow velocity and water depth by the normalized reference values ​​for flow velocity and water depth, respectively. Flow velocity affects bottom shear stress and turbulence intensity to drive sedimentation / resuspension, while water depth affects particle settling distance and bottom shear transfer efficiency. Both contribute equally to sediment dynamics in conventional rivers. In this embodiment, the initial weights of the two quotients (flow velocity to normalized reference value, and water depth to normalized reference value) are both set to 0.5; adaptive adjustments can be made based on channel type and historical data.

[0052] The particle settling parameters and resuspension parameters continuously regress to the conservation screening parameters based on the normalized hydrodynamic state changes in adjacent time windows, and do not perform discrete parameter switching according to normal flow, rising water, receding water or event flow.

[0053] This step first generates a normalized hydrodynamic state using flow velocity and water depth relative to a hydrodynamic normalized reference, and then calculates the increment of the normalized hydrodynamic state between adjacent time windows. For particle settling and resuspension parameters, the continuous update relationship can be expressed as: in, The parameters for the next time window of the k-th class in the i-th river segment; For the i-th river segment, the k-th type of current normalized dynamic parameters; For the k-th type of conservation screening parameters in the i-th river segment; The first response coefficient of the k-th parameter in the i-th river segment; The second response coefficient for the k-th parameter of the i-th river segment; The normalized hydrodynamic state increment for the i-th segment in adjacent time windows is calculated based on the normalized hydrodynamic state of the previous and current time windows. The first and second response coefficients are estimated using the hydrodynamic state-sediment response sequence during the calibration period, with least squares fitting as the estimation method.

[0054] This step adjusts the particle settling and resuspension parameters in the conservation screening parameters; the water-particle phase exchange parameters and sediment-water phase desorption parameters from step S4 are directly transferred to the evolution parameters. That is, the directly transferred values ​​from step S4 (water-particle phase exchange and sediment-water phase desorption parameters) and the calculated values ​​from step S5 (particle settling and resuspension parameters) are combined into complete normalized kinetic parameters for the next time window, used when executing the three-phase migration model in the next time window.

[0055] When the mineral composition and pollutant types remain essentially unchanged, the drift of exchange and desorption parameters determined by minerals / pollutants under hydrodynamic driving is negligible. Compared to water-particle exchange parameters and sediment-water desorption parameters, particle settling and resuspension parameters are more susceptible to hydrodynamic influences. This method uses the combined changes in flow velocity and water depth to drive the continuous updating of particle settling and resuspension parameters.

[0056] Furthermore, when any observation record is invalid, automatic differentiation is unavailable, conservation screening fails, or model input exceeds the calibration range, the conservation screening parameters of the previous effective time window are frozen, and the corresponding river segment and time window are marked as pending measurement. In the pending measurement state, updates using new parameters are stopped, and the conservation screening parameters of the previous effective time window are retained. Only after obtaining complete measured results for river segment identification, time window, sampling depth, and quality control identification, or after completing high-precision simulation results using the same boundary conditions and undergoing quality control, are the corresponding observation records and calibration references updated with these results. In the pending measurement state, the model continues to integrate forward using the conservation screening parameters of the previous effective time window; upon recovery, the sedimentary facies state is reinitialized using the latest acquired measured sedimentary facies inventory as the initial value, and the aqueous and suspended particle facies states are reset by the corresponding measured concentrations to eliminate long-term integration drift. The freezing and recovery processes described above are performed based on the calibration benchmark values ​​that are solidified and remain unchanged during the operation period in step S1, as well as the reference inventory, reference quality, hydrodynamic normalized reference, and quality conservation residual limit values ​​that do not change with parameter updates. The above benchmarks are not recalibrated in the state awaiting supplementary testing.

[0057] Repeat steps S2 to S5 continuously to calculate the mass change of each phase in each time window, the sediment stock calculated by the model and the external source-sink terms of the river section (output results when the three-phase migration model is executed again in step S4), and the evolution parameters (calculated in step S5 and used for the three-phase migration model in the next time window).

[0058] This embodiment selects the receiving section of a tributary as the test object. Multiple sections are divided along the river's length, and the experiment is conducted on one of these sections. The section is 1000m long, with an average width of 30m, an average depth of approximately 2.5m, and a water volume of approximately 75,000 cubic meters. 3 The riverbed area is 30,000 m².2 Table 1 presents the input mass, degradation loss mass, total mass change of the three phases, reference mass, and model-calculated sediment stock values ​​for this river section over 12 consecutive time windows.

[0059] Table 1

[0060] Table 2 shows examples of the baseline values ​​for the calibration stage of the normalized kinetic parameters and the normalized values ​​for the first time window.

[0061] Table 2

[0062] Table 3 shows the normalized kinetic parameters, three-phase mass conservation residuals, and inventory deviation losses for each time window in this method.

[0063] Table 3

[0064] Figure 2 The continuous evolution trajectory of normalized kinetic parameters, represented by the settling rate constant of suspended particles, over a time window is presented. Existing techniques discretely switch parameters in the order of normal flow, rising water, receding water, and rising water, resulting in significant jumps between adjacent time windows and failing to guarantee parameter continuity. By employing the method of this invention, because the parameters are selected based on conservation as the regression target and the evolution is continuous according to the normalized hydrodynamic state increments of adjacent time windows, the settling rate constant of suspended particles smoothly increases and tends to stabilize within the continuous time window.

[0065] Figure 3 The convergence comparison curves of the three-phase mass conservation residuals over time windows are presented. Existing techniques update only based on inventory deviation gradients, resulting in large residual fluctuations and the inability to eliminate candidate parameters whose mass states cannot be closed. By employing the method of this invention, the parameters to be screened are substituted back into the three-phase migration model, and the conservation parameters are screened using the three-phase mass conservation residuals and physical range within the same time window. This results in an overall decreasing trend in the residuals, which gradually converge.

[0066] Figure 4 A comparative curve of inventory deviation loss over a time window is presented. Existing techniques update the reference inventory without sufficient quality control of the supplementary data, resulting in significant loss fluctuations and compromising the traceability of the reference records. By adopting the method of this invention, due to the quality control of the observation records, the overall inventory deviation loss shows a downward trend and tends to stabilize.

[0067] This embodiment also provides a continuous adaptive calibration system for the three-phase nonequilibrium migration parameters of a river, the technical solution of which is as follows: including: The parameter calibration unit is used to acquire historical data on the concentration of water phase, suspended particulate phase, surface sediment concentration, surface sediment thickness, dry bulk density, sampling area, hydrodynamic boundary, discharge load and upstream boundary concentration of the receiving river, and to calibrate the reference inventory, reference mass, hydrodynamic normalized reference, mass conservation residual limit, normalized kinetic parameters and their physical ranges. The three-phase migration calculation unit is used to obtain the water phase concentration, suspended particle phase concentration, surface sediment concentration, surface sediment thickness, dry bulk density, sampling area, hydrodynamic boundary, discharge load and upstream boundary concentration within the time window, calculate the current sediment phase measured stock, and at the same time use the three-phase migration model to calculate the model calculated sediment stock. The automatic differential update unit is used to calculate the inventory deviation loss based on the reference inventory, the sediment inventory calculated by the model and the current measured sedimentary facies inventory, and to perform automatic differential update on the normalized kinetic parameters to obtain the parameters to be screened. The conservation screening unit is used to re-execute the three-phase migration model based on the parameters to be screened, and calculate the three-phase mass conservation residuals in conjunction with the reference mass. If the residuals are less than or equal to the mass conservation residual limit, the parameters to be screened are used as conservation screening parameters; otherwise, multiple candidate parameter groups are generated based on the parameters to be screened, and the corresponding three-phase mass conservation residuals are calculated for each group. Within the physical range of the parameters, the candidate parameter group whose mass conservation residuals do not exceed the mass conservation residual limit and whose distance from the parameters to be screened is the smallest is selected as the conservation screening parameters. The parameter evolution unit is used to calculate the normalized hydrodynamic state based on the hydrodynamic boundary and the hydrodynamic normalized reference. With the conservation screening parameters as the regression target, the particle settling parameters and resuspension parameters in the normalized dynamic parameters are updated according to the changes in the normalized hydrodynamic state in adjacent time windows to obtain the evolution parameters. The evolution parameters are used to substitute back into the three-phase migration model of the next time window.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A continuous adaptive calibration method for three-phase nonequilibrium migration parameters of a river, characterized in that, include: S1: Obtain historical data on the concentration of water phase, suspended particulate phase, surface sediment concentration, surface sediment thickness, dry bulk density, sampling area, hydrodynamic boundary, discharge load and upstream boundary concentration of the receiving river, and calibrate to obtain reference inventory, reference mass, hydrodynamic normalized reference, mass conservation residual limit, normalized kinetic parameters and their physical ranges. S2: Obtain the concentration of water phase, concentration of suspended particles phase, concentration of surface sediment, thickness of surface sediment, dry bulk density, sampling area, hydrodynamic boundary, discharge load and upstream boundary concentration, calculate the current sediment phase inventory, and use the three-phase migration model to calculate the model-calculated sediment inventory. S3: Calculate the inventory deviation loss based on the reference inventory, the model-calculated sediment inventory, and the current measured sedimentary facies inventory, and perform automatic differential updates on the normalized kinetic parameters to obtain the parameters to be screened; S4: Based on the parameters to be screened, the three-phase migration model is executed again, and the three-phase mass conservation residual is calculated in conjunction with the reference mass. If it is less than or equal to the mass conservation residual limit, the parameters to be screened are used as conservation screening parameters. Otherwise, generate multiple candidate parameter groups based on the parameters to be screened, calculate the corresponding three-phase mass conservation residuals for each group, and select the candidate parameter group whose mass conservation residual does not exceed the mass conservation residual limit and is the smallest distance from the parameters to be screened within the physical range of the parameters, as the conservation screening parameters. S5: The normalized hydrodynamic state is calculated based on the hydrodynamic boundary and the normalized hydrodynamic reference. The particle settling parameter and resuspension parameter in the normalized dynamic parameters are updated according to the change of the normalized hydrodynamic state in adjacent time windows to obtain the evolution parameters. The evolution parameters are used to back-substitute to the three-phase migration model of the next time window.

2. The continuous adaptive calibration method for river three-phase non-equilibrium migration parameters as described in claim 1, characterized in that, In step S3, the formula for calculating inventory deviation loss is: in, For the inventory deviation loss of the i-th segment of the receiving river; This represents the measured sedimentary facies inventory for the i-th river segment. Calculate the sediment stock for the model of the i-th river segment. This is the reference inventory for the i-th river segment.

3. The continuous adaptive calibration method for river three-phase non-equilibrium migration parameters as described in claim 2, characterized in that, In step S3, the formula for calculating the updated value of the normalized kinetic parameter of type k is: in, This is the k-th parameter to be filtered for the i-th river segment; For the i-th river segment, the k-th type of current normalized dynamic parameters; Let be the update step size for the k-th type of parameter; for right The gradient value.

4. The continuous adaptive calibration method for river three-phase non-equilibrium migration parameters as described in claim 1, characterized in that, Normalized kinetic parameters include: water-particle phase exchange parameters, particle settling parameters, resuspension parameters, and sediment-water phase desorption parameters.

5. The continuous adaptive calibration method for river three-phase non-equilibrium migration parameters as described in claim 1, characterized in that, In step S4, the three-phase mass conservation residual is the difference between the total three-phase mass change and the algebraic sum of the input mass, output mass, degradation loss mass, and gas-water exchange loss mass, and then divided by the reference mass.

6. The continuous adaptive calibration method for river three-phase non-equilibrium migration parameters as described in claim 1, characterized in that, In step S4, if the three-phase mass conservation residual is greater than the mass conservation residual limit, multiple candidate parameter groups are generated within the physical range of each parameter by enumeration or iteration based on the parameter to be screened, and each candidate parameter group is re-substituted into the three-phase migration model to calculate the corresponding three-phase mass conservation residual; among the candidate parameter groups whose three-phase mass conservation residual is less than or equal to the mass conservation residual limit, the candidate parameter group with the smallest distance from the parameter to be screened is selected as the conservation screening parameter.

7. The continuous adaptive calibration method for river three-phase non-equilibrium migration parameters as described in claim 1, characterized in that, In step S5, the normalized hydrodynamic state is the weighted sum of the two quotients obtained by dividing the flow velocity and water depth by the normalized reference values ​​for flow velocity and water depth, respectively. The hydrodynamic boundary includes flow velocity and water depth; the hydrodynamic normalization reference includes normalized reference values ​​for flow velocity and water depth.

8. The continuous adaptive calibration method for river three-phase non-equilibrium migration parameters as described in claim 1, characterized in that, The update relationship between particle settling parameters and resuspension parameters is expressed as follows: in, For the i-th river segment, the parameters for the k-th class of the next time window; For the i-th river segment, the k-th type of current normalized dynamic parameters; For the k-th type of conservation screening parameters in the i-th river segment; The first response coefficient of the k-th parameter in the i-th river segment; The second response coefficient for the k-th parameter of the i-th river segment; It represents the normalized hydrodynamic state increment for the adjacent time window of the i-th river segment.

9. The continuous adaptive calibration method for river three-phase non-equilibrium migration parameters as described in claim 1, characterized in that, The particle settling parameters and resuspension parameters continuously regress to the conservation screening parameters based on the normalized hydrodynamic state changes in adjacent time windows, and do not perform discrete parameter switching according to normal flow, rising water, receding water or event flow.

10. A continuous adaptive calibration system for three-phase nonequilibrium migration parameters of a river, characterized in that, A continuous adaptive calibration method for river three-phase nonequilibrium migration parameters as described in any one of claims 1 to 9, comprising: The parameter calibration unit is used to acquire historical data on the concentration of water phase, suspended particulate phase, surface sediment concentration, surface sediment thickness, dry bulk density, sampling area, hydrodynamic boundary, discharge load and upstream boundary concentration of the receiving river, and to calibrate the reference inventory, reference mass, hydrodynamic normalized reference, mass conservation residual limit, normalized kinetic parameters and their physical ranges. The three-phase migration calculation unit is used to obtain the concentration of water phase, suspended particle phase, surface sediment concentration, surface sediment thickness, dry bulk density, sampling area, hydrodynamic boundary, discharge load and upstream boundary concentration, and calculate the current sediment phase measured stock. At the same time, the model-calculated sediment stock is obtained by using the three-phase migration model. The automatic differential update unit is used to calculate the inventory deviation loss based on the reference inventory, the sediment inventory calculated by the model, and the current measured sedimentary facies inventory, and to perform automatic differential update on the normalized kinetic parameters to obtain the parameters to be screened. The conservation screening unit is used to re-execute the three-phase migration model based on the parameters to be screened, and calculate the three-phase mass conservation residuals in conjunction with the reference mass. If the residuals are less than or equal to the mass conservation residual limit, the parameters to be screened are used as conservation screening parameters; otherwise, multiple candidate parameter groups are generated based on the parameters to be screened, and the corresponding three-phase mass conservation residuals are calculated for each group. Within the physical range of the parameters, the candidate parameter group whose mass conservation residuals do not exceed the mass conservation residual limit and whose distance from the parameters to be screened is the smallest is selected as the conservation screening parameters. The parameter evolution unit is used to calculate the normalized hydrodynamic state based on the hydrodynamic boundary and the hydrodynamic normalized reference. With the conservation screening parameters as the regression target, the particle settling parameters and resuspension parameters in the normalized dynamic parameters are updated according to the changes in the normalized hydrodynamic state in adjacent time windows to obtain the evolution parameters. The evolution parameters are used to substitute back into the three-phase migration model of the next time window.

Images