Dynamic regulation line optimization method based on river regime stability analysis

Abstract

The application discloses a dynamic regulation line optimization method based on river regime stability analysis, namely: obtaining basic water and sediment data of a target river section, constructing a water dynamic and sediment mathematical model based on the data, simulating an evolution process and extracting river regime characteristics such as flow velocity, shear force and thalweg swing. Based on the extracted river regime characteristics, a comprehensive river regime stability index of the target river section is determined. On this basis, taking improvement of the comprehensive river regime stability index as a target, a dynamic regulation line optimization model is established with regulation line parameters to be optimized as decision variables. The dynamic regulation line optimization model is iteratively solved, and an optimal regulation line scheme is output. Based on the above scheme, the application can quantify the river regime stability state of the whole river section, improve the global optimization efficiency of the spatial form design of the river regulation line and the scientific nature of the engineering planning under the premise of meeting the complex water and sediment environment and multiple engineering constraints.

Description

Technical Field

[0001] This invention relates to the fields of water conservancy engineering and river dynamics, and in particular to a dynamic regulation line optimization method based on river stability analysis. Background Technology

[0002] River regulation projects are key physical interventions for maintaining shoreline stability, improving navigation hydraulic conditions, and controlling riverbed evolution. The planning of the regulation line morphology directly determines the flow field structure distribution of the cross-section and the shear force response at the bottom of the riverbed, which helps to suppress abnormal scouring, guide smooth flow transitions, and maintain long-term water-sediment dynamic balance.

[0003] Currently, when planning river regulation lines, it is common practice to construct a static hydrodynamic numerical model with a pre-designed frequency, and then combine this with human experience to formulate a limited number of schemes for numerical simulation and comparison. When evaluating the merits of the schemes, quantitative control references are often based on the local flow velocity, extreme water level, or local scouring and sedimentation volume at a single cross-section.

[0004] Existing planning methods still fall short in addressing long-term, complex hydrodynamic environments, lacking sufficient physical objectivity in their evolution evaluation mechanisms and spatial optimization capabilities when solving multi-objective engineering constraints. Therefore, it is necessary to research a method for river regulation assessment and optimization. Summary of the Invention

[0005] The purpose of this invention is to design a dynamic regulation line optimization method based on river stability analysis, in order to at least solve one problem of the existing technology.

[0006] Technical solution: A dynamic regulation line optimization method based on river stability analysis, comprising:

[0007] Obtain basic water and sediment data for the target river section;

[0008] A hydrodynamic sediment mathematical model is constructed based on basic water and sediment data to simulate the evolution process and extract river morphology features.

[0009] The comprehensive river stability index of the target river section is determined based on river morphology characteristics;

[0010] With the goal of improving the comprehensive river stability index, a dynamic regulation line optimization model is established, with the regulation line parameters to be optimized as decision variables.

[0011] The dynamic remediation line optimization model is iteratively solved to output a better remediation line scheme.

[0012] According to one aspect of this application, a hydrodynamic sediment mathematical model is constructed based on basic water and sediment data to simulate the evolution process and extract river morphology features, including:

[0013] A digital elevation model was constructed by integrating multi-source topographic bathymetry data and photogrammetric data, and discretized using an unstructured grid.

[0014] Boundary conditions containing flood hydrographs of different frequencies were set, and a two-way coupled simulation of water flow-driven sediment transport and riverbed deformation feedback flow was performed.

[0015] The velocity characteristics, shear force characteristics, and thorium oscillation characteristics of key river morphology control nodes are extracted as river morphology features.

[0016] According to another aspect of this application, the comprehensive river stability index for a target river segment is determined based on river morphology characteristics, including:

[0017] Based on the river morphology characteristics, the flow compliance index, bank erosion risk index, flow turbulence index, and thorium swing amplitude index of the target river section are calculated.

[0018] The indicators of flow compliance, bank erosion risk, flow turbulence, and thalassus oscillation amplitude were dimensionless and synthesized into a comprehensive river stability index.

[0019] According to another aspect of this application, in synthesizing a comprehensive river stability index by dimensionless processing of the flow compliance index, bank erosion risk index, flow turbulence index, and thalex swing amplitude index, a nonlinear coupling approach is adopted, specifically as follows:

[0020] Calculate the geometrically weighted product of the dimensionless indicators and use it as the main index term;

[0021] Calculate the coordinated degradation terms of normalized scouring and disordered oscillation;

[0022] The positive values ​​are extracted by combining the main index term with the coordinated deterioration terms of compliant scouring and disordered oscillation to obtain the comprehensive river stability index.

[0023] According to another aspect of this application, a dynamic remediation line optimization model is established, using the remediation line parameters to be optimized as decision variables, including:

[0024] The treatment line is discretized into a sequence of control nodes, and a spline curve is used to fit the sequence of control nodes. The coordinates of the control nodes are set as the parameters of the treatment line to be optimized.

[0025] Set multiple constraints and transform them into an external penalty function;

[0026] Establish a multi-objective assessment system that includes minimizing the variation rate of the comprehensive river stability index.

[0027] According to another aspect of this application, the dynamic remediation line optimization model is iteratively solved, including:

[0028] A multi-objective hybrid evolutionary algorithm is used for the search, and the mutation probability of the population is dynamically adjusted according to the diversity and dispersion of the decision variable set.

[0029] Based on the current parameters of the regulation line to be optimized, the river channel calculation grid is dynamically reconstructed. When the displacement of the grid boundary node does not exceed the preset scale threshold, the incremental hot-start hydrodynamic simulation is called to update the boundary conditions.

[0030] According to another aspect of this application, before iteratively solving the dynamic remediation line optimization model, a pre-built multi-fidelity proxy model is invoked for pre-screening. The multi-fidelity proxy model is trained in the following way:

[0031] A low-fidelity prior solution set is obtained using a one-dimensional hydraulic equation set.

[0032] A high-fidelity benchmark solution set was obtained using a two-dimensional hydrodynamic sediment mathematical model.

[0033] The residual distribution between the high-fidelity benchmark solution set and the low-fidelity prior solution set is fitted using a co-kriging network.

[0034] According to another aspect of this application, the pre-configured homing scour co-degradation term is calculated as follows:

[0035] Obtain the deviation between the preset ideal flow compliance state and the flow compliance index, obtain the deviation between the preset ideal bank slope erosion resistance state and the bank slope erosion risk index, and multiply the compliance deviation and the erosion deviation to obtain the compliance-erosion synergistic degradation term.

[0036] According to another aspect of this application, the parameters of the regulation line to be optimized are set as a coordinate sequence of the groynes group connecting line, a coordinate sequence of the revetment outline line, or a coordinate sequence of the guide dike axis. The optimal regulation line scheme is output simultaneously with an assessment list of regulation building material consumption that meets the engineering constraints.

[0037] According to another aspect of this application, outputting the optimal regulation line scheme also includes mapping the optimal regulation line scheme to the workspace of the digital twin system, and generating a three-dimensional information model of the river engineering with spatial geometric attributes based on the coordinate control node distribution of the optimal regulation line scheme.

[0038] Beneficial effects: This invention can quantify the overall stability of a river section, and improve the global optimization efficiency and scientific nature of engineering planning in the spatial morphology design of river regulation lines, while meeting the requirements of complex water and sediment environments and multiple engineering constraints. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating a dynamic regulation line optimization method based on river stability analysis, as described in this application.

[0040] Figure 2 This is a schematic diagram illustrating the process of constructing a hydrodynamic sediment mathematical model based on basic water and sediment data, simulating the evolution process, and extracting river morphology features, as described in this application.

[0041] Figure 3 This is a flowchart illustrating a process for determining a comprehensive river stability index for a target river segment based on river characteristics, as described in this application.

[0042] Figure 4 This is a flowchart illustrating a dynamic treatment line optimization model based on the treatment line parameters to be optimized, as described in this application.

[0043] Figure 5 This is a schematic diagram of an iterative solution process for an optimization model of dynamic remediation lines in this application. Detailed Implementation

[0044] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0045] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in sequences other than those illustrated or described herein. Furthermore, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0046] In view of the aforementioned problems, the applicant conducted in-depth searches and analyses and found that:

[0047] Existing static and local evaluation models are unable to capture the dynamic cumulative effect of water flow over time on the evolution of complex riverbeds. At the same time, when faced with nonlinear hydraulic phenomena involving multidimensional coupling such as lateral oscillation of the deep thalve, turbulent flow, and bank erosion, the judgment system of local extreme values ​​often tends to mask the risk of systemic local imbalance.

[0048] Furthermore, when faced with the dynamic process of water flow driving sediment and riverbed deformation feeding back to the water flow, the iterative approach of relying on conventional parameter calculations suffers from low evolutionary search efficiency and difficulty in overcoming local optimal solutions when solving for the boundary morphology of high-dimensional continuous space.

[0049] Accordingly, some nouns or terms appearing in the description of the embodiments of this application shall be interpreted as follows:

[0050] The regulated guideline is the control line for guiding river flow, determined by the regulated guideline.

[0051] The basic water and sediment data for the target river section serve as the boundary conditions for subsequent simulations.

[0052] To overcome the above shortcomings, combined with Figures 1 to 5 The present application will be further illustrated with reference to specific embodiments.

[0053] As an example, a dynamic regulation line optimization method based on river stability analysis includes the following steps:

[0054] Step 101: Obtain basic water and sediment data for the target river section.

[0055] Alternatively, it can retrieve hydrological survey data (hydrological data) and topographic survey data from the historical reference window.

[0056] Among them, hydrological data mainly cover historical flow process lines, observed water levels at water level control stations along the route, suspended and bedload sediment gradation and sediment content processes, which can provide driving loads for reconstructing long-term water and sediment evolution processes.

[0057] Topographic data includes the measured elevation distribution of the deep water area, shallow water area, and slopes on both banks of the main river.

[0058] In practice, the data retrieval source for water flow and meteorological boundary conditions can be limited to historical reference window data. For example, measured flood sequences from previous years, such as 10 or 20 years, can be obtained.

[0059] It should be understood that the above acquisition must be based on physical records that have already occurred.

[0060] Step 102: Construct a hydrodynamic sediment mathematical model based on basic water and sediment data to simulate the evolution process and extract river morphology features.

[0061] The process of constructing a hydrodynamic sediment mathematical model based on basic water and sediment data involves transforming multi-source discrete geographic mapping information into a continuous computational control domain, and establishing a two-way feedback solution mechanism for water flow control and sediment balance within the computational domain.

[0062] Furthermore, this step can be expanded as follows:

[0063] Step 1021: Integrate multi-source topographic depth data and photogrammetric data from the basic water and sediment data to construct a digital elevation model, and discretize it using an unstructured grid.

[0064] In practice, multi-source data obtained by technologies such as multibeam bathymetry in shallow water, single-beam bathymetry in deep water, and oblique photography of bank slopes by UAVs are used to generate a digital elevation model containing the characteristics of the main channel, side beaches, and existing control structures through coordinate system 1 and outlier removal processing.

[0065] Furthermore, the model is discretized using unstructured meshes, such as irregular triangular meshes. Local mesh refinement can be performed in areas with drastic changes in flow gradient, such as riverbank boundaries, areas surrounding existing regulation structures, and near the thalweg, while meshing can be appropriately reduced in gentler areas such as wide beaches.

[0066] Furthermore, quality constraints for mesh reconstruction can be set, such as controlling the interior angles of mesh cells within a preset angle range and ensuring that the area ratio of adjacent mesh cells does not exceed a predetermined ratio.

[0067] In some scenarios, the preset angle range can be set to 15°-150°, and the preset ratio can be set to 1:4;

[0068] This is merely an illustrative example and not the only limitation. Other numerical examples in this application are similar and will not be described one by one.

[0069] For non-conforming units that do not meet this quality limit, local node repositioning can be performed for repair.

[0070] If the standard is still not met after reaching the maximum number of repairs, a global maximum penalty value is assigned to the current grid scheme to skip the hydrodynamic simulation.

[0071] Step 1022: Set boundary conditions including flood process lines of different frequencies, and perform a two-way coupled simulation of water flow-driven sediment transport and riverbed deformation feedback flow.

[0072] Accordingly, the solver for the hydrodynamic sediment mathematical model is based on the two-dimensional shallow water equation and the hydrodynamic equilibrium equation.

[0073] The flood process lines of different frequencies, such as once in 5 years, once in 20 years, once in 50 years, and once in 100 years, are used as the upstream flow boundary input, while the downstream water level boundary is controlled by the corresponding measured relationship curve.

[0074] Within each computational step, the solver calculates the flow field and obtains the near-bottom shear force, which drives the initiation and transport of sediment on the bed surface. Sediment scouring and deposition, in turn, update the grid node elevations, thereby realizing the bidirectional real-time coupled evolution of water flow-driven sediment transport and riverbed deformation feedback water flow.

[0075] Step 1023: Extract the velocity characteristics, shear force characteristics, and thalweg oscillation characteristics of key river morphology control nodes as river morphology features.

[0076] During the simulation evolution process, for sensitive areas of river morphology control, namely key river morphology control nodes, such as bend apex, branching gate, and control engineering head, the physical quantity response of grid nodes is tracked and recorded.

[0077] Flow velocity characteristics can be represented as a time series of the magnitude and directional deflection of the flow velocity at the recording node;

[0078] Shear force characteristics, specifically the distribution of bottom friction shear force at the node;

[0079] The oscillation characteristics of the thalweg are determined by tracking the coordinate sequence of the maximum water depth points on each cross-section, i.e., the thalweg points, to form the oscillation trajectory of the thalweg as it evolves over time.

[0080] Step 103: Determine the comprehensive river stability index of the target river section based on river characteristics.

[0081] River stability is not a localized characteristic of a single section, but a comprehensive reflection of the flow pattern, the erosion resistance of the bank slopes, and the planar position of the main channel.

[0082] This step can also be performed in the following way:

[0083] Based on the river morphology characteristics, the flow compliance index, bank erosion risk index, flow turbulence index, and thalassus swing amplitude index of the target river section are calculated.

[0084] The indicators of flow compliance, bank erosion risk, flow turbulence, and thalassus oscillation amplitude were dimensionless and synthesized into a comprehensive river stability index.

[0085] Among them, the water flow compliance index can be obtained by extracting the angle between the velocity vector direction of each grid cell in the simulated flow field and the tangent direction of the corresponding treatment line, and calculating the cosine area weighted mean of the angle.

[0086] The bank slope erosion risk index can be calculated by extracting the bottom friction shear force of each grid cell in the bank slope area and the pre-configured critical erosion shear force of the soil, calculating the ratio of the bottom friction shear force exceeding the critical erosion shear force of the soil, and using the area-weighted average of the ratio as the index.

[0087] The flow turbulence index can be weighted by the distribution ratio of grid cells exhibiting countercurrent flow along the mainstream direction and the information entropy of the velocity direction distribution; the distribution ratio of grid cells is the area ratio of the recirculation zone.

[0088] The thalweg swing amplitude index is calculated as the ratio of the maximum offset width of the thalweg point to the width of the river at that cross section.

[0089] Furthermore, a dimensionless mapping needs to be performed on the four indicators with different physical meanings.

[0090] In practical implementation, a piecewise linear mapping with threshold sensitivity or a sigmoid continuous mapping function can be used to uniformly map the index values ​​to a preset numerical range.

[0091] By mathematically synthesizing the mapped sub-indices, for example by using a linear weighted combination, a comprehensive river stability index reflecting the overall stability of the target river section can be obtained.

[0092] Step 104: With the goal of improving the comprehensive river stability index, a dynamic regulation line optimization model is established, with the regulation line parameters to be optimized as decision variables.

[0093] Among them, the dynamic rectification line optimization model is used to transform the layout of physical engineering projects into a computer-searchable parameter space;

[0094] Establish a dynamic remediation line optimization model with the remediation line parameters to be optimized as decision variables, including:

[0095] Step 1041: Discretize the treatment line into a sequence of control nodes, fit the sequence of control nodes with spline curves, and set the coordinates of the control node sequence as the parameters of the treatment line to be optimized.

[0096] The physical structure to be optimized is discretized into several spatial control nodes, such as the coordinate sequence of the groynes group connection line, the coordinate sequence of the revetment outline line, or the coordinate sequence of the guide dike axis. These can also be referred to as the regulation line parameters to be optimized.

[0097] By using smooth spline curves to connect the above control node sequence, not only can the continuous and smooth shape of the treatment line be made and the water-blocking sharp corners be avoided, but the shape change of the entire treatment line can also be guided by the movement of the control point coordinates.

[0098] The set of spatial coordinates of the control nodes constitutes the decision variables for the optimization algorithm to perform the search.

[0099] Step 1042: Set multiple constraints and transform them into an external penalty function.

[0100] Accordingly, the flood control safety water level limit, the navigation clearance elevation limit, and the curvature radius limit of the regulation line are obtained as multiple constraints.

[0101] For example, the highest flood level obtained from the simulation must not exceed the design elevation of the dike minus the safety superelevation.

[0102] These rigid engineering constraints are transformed into external penalty functions and applied to the objective function.

[0103] Furthermore, a penalty factor based on an adaptive incremental strategy can be constructed and periodically multiplied as the number of generations increases.

[0104] In some cases, the initial value of the penalty factor can be set to 10.

[0105] Furthermore, an external penalty function is constructed by multiplying the penalty factor with the deviation of the current candidate solution from violating multiple constraints.

[0106] It should be understood that this design allows the algorithm to explore the infeasible region to a certain extent in the early stages of optimization, while imposing extremely severe penalties on solutions that violate the constraints in the later stages, forcing the population to converge to the feasible region.

[0107] Step 1043: Establish a multi-objective evaluation system that includes minimizing the variation rate of the comprehensive river morphology stability index and an external penalty function.

[0108] It should be understood that the optimization objective is not merely to increase the absolute value of stability in a specific locality, but rather to seek balanced governance of the entire river section. Therefore, calculating the spatial variability rate of the comprehensive river stability index along the affected area of ​​the regulation line, with minimizing this variability rate as one of the objectives, can be used to eliminate local weak links; at the same time, combining maximizing the mainstream conformity and minimizing the consumption of engineering construction materials, a multi-objective evaluation system is formed.

[0109] Step 105: Iteratively solve the dynamic remediation line optimization model and output the optimal remediation line scheme.

[0110] The solution process relies on an evolutionary computation engine to conduct long-term exploration and development in the decision variable space.

[0111] Furthermore, this step can be expanded as follows:

[0112] Step 1051: A multi-objective hybrid evolutionary algorithm is used for searching, and the mutation probability of the population is dynamically adjusted according to the diversity and dispersion of the decision variable set.

[0113] In the solution engine, a hybrid strategy of using the non-dominated sorting genetic algorithm NSGA-III in combination with the particle swarm optimization algorithm PSO can be adopted.

[0114] The initial population can be generated by Latin hypercube sampling to ensure the uniformity of spatial distribution.

[0115] During the iteration process, the normalized dispersion of the decision variables in the multidimensional space is measured;

[0116] Based on this dispersion, the exponential decay formula is introduced to calculate the adaptive mutation probability;

[0117] When a highly concentrated population may fall into a local extreme, increase the mutation probability to increase the perturbation.

[0118] When the population is widely dispersed, reducing the mutation probability can facilitate convergence.

[0119] Step 1052: Dynamically reconstruct the river channel calculation grid based on the current parameters of the regulation line to be optimized. When the displacement of the grid boundary node does not exceed the preset scale threshold, call the incremental hot-start hydrodynamic simulation to update the boundary conditions.

[0120] Accordingly, when the displacement exceeds the preset scale threshold, a full-domain restart hydrodynamic simulation is performed to obtain the updated river morphology characteristics in order to calculate the fitness evaluation value of the current scheme.

[0121] Each iteration generates new control point coordinates, indicating that the physical boundary of the river channel has changed;

[0122] The unstructured mesh generator reconstructs the mesh topology and maps the hydrodynamic boundary based on the new coordinates.

[0123] To improve efficiency, a warm start operation is performed when it is determined that the displacement between the old and new control points does not exceed a preset multiple of the local average grid size.

[0124] That is, the stable flow field of the previous generation is directly extracted as the initial calculation field of the current step, and only transient incremental calculation is performed on the local disturbance region, thus avoiding the convergence process of the global initial field starting from zero.

[0125] The algorithm terminates when the set convergence criteria are met, such as the rate of change of the solution space hypervolume being less than a specified limit.

[0126] In the output stage, the TOPSIS approximation ideal solution ranking method can be used, combined with the expert's preference weights for different focuses of flood control or navigation, to recommend a better treatment line scheme from the non-dominated solution set.

[0127] In addition to outputting coordinate data, the optimal remediation line scheme also outputs a list of remediation building material consumption assessments that meet the preset engineering constraints.

[0128] Furthermore, the optimal regulation line scheme can be mapped to the workspace of the digital twin system, and a three-dimensional information model of the river engineering with spatial geometric attributes can be generated based on the coordinate control node distribution of the optimal regulation line scheme, providing digital support for subsequent visualization simulation and engineering design.

[0129] According to one aspect of this application, the calculation of the material consumption list can be based on the geometric parameters of the optimal treatment line and the pre-configured material quota standards, and the engineering quantity can be calculated by multiplication.

[0130] The geometric parameters, such as length and cross-sectional dimensions.

[0131] According to one aspect of this application, the generation of a three-dimensional information model can be achieved by using a parametric modeling engine to sweep along the treatment line to generate solid geometry based on the elevation and planar position of coordinate control nodes.

[0132] According to one aspect of this application, the normalized dispersion can be the mean of the normalized variance of the individual decision variables of the population in each dimension, and the normalized mean of the Euclidean distance between individuals.

[0133] Alternatively, other conventional methods for measuring population diversity can be used to calculate it;

[0134] It should be understood that an appropriate metric can be selected based on the actual performance of the evolutionary algorithm.

[0135] In other embodiments, a different implementation of the synthetic river morphology stability index is described, which uses a nonlinear coupling approach to process the indicators of each dimension. The remaining steps are the same as described above and will not be repeated here.

[0136] In one possible implementation, this embodiment includes the following steps:

[0137] Step 201: Calculate the geometrically weighted product of the dimensionless indicators and use it as the main index term.

[0138] Accordingly, the supplementary values ​​of the bank slope erosion risk index are transformed into positive bank slope erosion resistance index;

[0139] The complement value of the flow turbulence index is taken and transformed into a positive flow smoothness index;

[0140] The complementary value of the deep-water oscillation amplitude index is converted into a positive deep-water stability index.

[0141] Take the complement value, that is, 1 minus the corresponding index after normalization;

[0142] In traditional river course assessment, various indicators are usually superimposed with linear weights. This method can mask a low value in a single dimension indicator by a high score in other dimensions.

[0143] Therefore, this embodiment introduces a geometrically weighted average to construct the backbone term of the index.

[0144] Furthermore, the four positive indicators, after being dimensionless and homogenized, are multiplied together by exponentiation according to their respective weight coefficients.

[0145] On the one hand, this calculation form maps the shortest board effect in river geomorphological evolution. That is, when any one of the following factors—flow compliance, bank slope erosion resistance, flow regime smoothness, or thalassus stability—is biased towards zero, no matter how high the performance of the other indicators is, the product of these factors will approach zero, thus determining that the overall river morphology of the grid unit is in an unstable state.

[0146] For example, HSI _GM =(I _1 * ) w_1 ×(I _2 * )w _2 ×(I _3 * ) w_3 ×(I _4 * ) w_4 ;

[0147] Among them, HSI _GM As the main term of the index, I _1 * I is the normalized flow compliance index. _2 * I is the normalized slope erosion resistance index. _3 * I is the normalized flow smoothness index. _4 * w is the normalized deep-water stability index. _1 w _2 w _3 w _4 These are the weighting coefficients for each corresponding indicator, and the sum of the weighting coefficients is limited to 1.

[0148] The weighting coefficients w1 to w4 were determined by the Analytic Hierarchy Process (AHP), with river morphology sensitivity as the judgment criterion. They were obtained by those skilled in the art through pairwise comparison of the water and sediment characteristics of the target river section with the treatment target and by calculating the eigenvector.

[0149] Step 202: Based on the various dimensionless indicators, calculate the coordinated degradation terms of shunting and turbulent oscillation; or, calculate the coordinated degradation terms of shunting and turbulent oscillation according to the preset formula structure.

[0150] Alternatively, this step can also be performed in the following manner:

[0151] Obtain the deviation between the preset ideal flow compliance state and the flow compliance index, obtain the deviation between the preset ideal bank slope erosion resistance state and the bank slope erosion risk index, and multiply the compliance deviation and the erosion deviation to obtain the compliance-erosion synergistic degradation term.

[0152] Obtain the turbulence deviation between the preset ideal smooth flow state and the flow turbulence index, and obtain the oscillation deviation between the preset ideal deep-water stability state and the deep-water oscillation amplitude index. Multiply the turbulence deviation and the oscillation deviation to obtain the turbulence-oscillation co-deterioration term.

[0153] Among these, there are physical coupling phenomena among the various dimensions of river stability.

[0154] For example, when the compliance of the water flow decreases, the water flow impacts the riverbank at a larger angle, which increases the risk of bank erosion.

[0155] Meanwhile, turbulent flow patterns are often accompanied by disordered lateral oscillations of the thalweg, which in turn disrupt the mainstream structure.

[0156] To quantify this cooperative coupling mechanism, this embodiment defines two sets of cooperative degradation terms. By introducing a constant 1 to represent the upper limit of the ideal state, the difference between each index and the ideal state is calculated.

[0157] When two deviations within the same group have large values, their product forms a larger penalty factor.

[0158] Conversely, if any one of the indicators is in an ideal state and the deviation approaches zero, then the product term loses its penalizing effect.

[0159] As an example, the calculation of the coordinated degradation term of normalization scour and the coordinated degradation term of disordered oscillation can be expressed by the following formula:

[0160] Φ _12 =(1-I _1 * )×(1-I _2 * );

[0161] Φ _34 =(1-I _3 * )×(1-I _4 * );

[0162] Where, Φ _12 To align with the scouring and deterioration items, Φ _34 For the disordered oscillation co-deterioration term, 1 represents the upper limit of the ideal state preset by the index.

[0163] Step 203: Combine the main index term with the co-deterioration term of compliant scouring and the co-deterioration term of disordered oscillation to extract positive values ​​and obtain the comprehensive river stability index.

[0164] After obtaining the exponential principal term and two sets of co-degradation terms, the final nonlinear mapping function needs to be constructed. This mapping mechanism uses the exponential principal term as a reference, scales the co-degradation terms using a coupling strength coefficient, and subtracts the scaled penalty value from the reference constant 1.

[0165] Considering that the deduction result may penetrate the zero line under the condition of multiple indicators superimposed and deteriorated, a positive operator is introduced to truncate the calculation result to ensure that the final output stability score meets the physical constraints of the system.

[0166] This evaluation method can be quantified using the following formula:

[0167] HSI=HSI _GM ×max(1-c _12 ×Φ _12 -c _34 ×Φ _34 ,0);

[0168] The HSI above is the Comprehensive River Context Stability Index. _GM Here, c is the exponential term, max is the operator for extracting positive values, and c is the exponential term. _12 c is the coupling strength coefficient between compliance and scouring. _34 Φ is the coupling strength coefficient between disorder and oscillation. _12 With Φ _34 These are two sets of co-degradation terms generated from the preceding calculations.

[0169] In actual engineering configurations, c _12 With c _34 The sum of its values ​​is usually less than or equal to 1;

[0170] On the other hand, the specific value of the coupling strength coefficient can be determined based on the historical statistical correlation between the river morphology characteristics of each dimension of the target river section, or it can be adjusted through sensitivity analysis under the premise of satisfying the non-negativity constraint of the exponent.

[0171] This embodiment reveals the negative impact of local instability on the overall river morphology, which can be used to correct the tendency of conventional linear evaluation to miss certain cases.

[0172] On the other hand, the indicators of flow compliance, bank erosion risk, flow turbulence, and thalassus oscillation amplitude are dimensionless and synthesized into a comprehensive river stability index using a time integration approach.

[0173] Step 301: Calculate the transient river stability index for multiple time steps in the simulation process.

[0174] Conventional river morphology assessment methods use extreme values ​​for flood events, which fail to reflect the continuous and cumulative impact of flood duration on riverbed scouring and deposition.

[0175] In this embodiment, for any given flood condition, the evaluation index is calculated once within each calculation time step of its hydrodynamic simulation;

[0176] When processing the input sequence, a pre-configured historical reference window needs to be called as the retrieval source to obtain the hydrological and boundary data of the corresponding working conditions, so as to ensure the determinism of water and sediment time series extrapolation.

[0177] At each output time step, based on the local flow field, water depth, and bottom shear force physical field at the current moment, the basic indicators such as the flow compliance degree and flow turbulence degree at the corresponding moment are calculated and obtained, thus obtaining the time series of transient river stability index that evolves continuously over time.

[0178] Step 302: Extract the over-limit amplitude value of the bottom friction shear force exceeding the pre-configured critical starting shear force, and construct the effective work weighting weight of the terrain.

[0179] Among them, the bottom friction shear force can be output by the hydrodynamic sediment mathematical model.

[0180] The initiation and transport of sediment particles in the riverbed are subject to rigid physical threshold constraints, resulting in a significant difference in the effective contribution of low-flow periods and high-flood periods to the reshaping of river channel geometry. Therefore, this step introduces geomorphological effective work parameters to weight the transient evaluation results.

[0181] Due to the existence of a physical threshold, this process includes a state determination action regarding whether the bottom friction shear force exceeds a pre-configured critical starting shear force. At this point, the system splits into two parallel execution paths:

[0182] Under a certain hydrodynamic flow condition, when the bottom friction shear force does not exceed the critical starting shear force, the weighted weight of the effective geomorphic work at the corresponding time step is set to a zero contribution reference value.

[0183] When the kinetic energy of the water body is insufficient to overcome the self-weight of the sediment particles on the bed surface and the cohesion between the particles, the local riverbed is in a relatively static and non-scouring state, and no sediment transport work is generated. At this time, the effective work weight of the geomorphology at this time step is forcibly assigned to 0, so that the transient river stability index at this moment is removed in the subsequent time integration.

[0184] Under another hydrodynamic flow condition, when the bottom friction shear force exceeds the critical starting shear force, the specified nonlinear exponent power of the difference between the bottom friction shear force and the critical starting shear force is obtained, and the result of the nonlinear exponent power is used as the weight of the effective work of the geomorphology. The specified nonlinear exponent is the empirical power law constant of the bedload sediment transport intensity.

[0185] Once the hydrodynamic force exceeds the critical initiation threshold for sediment transport, sediment from the bed surface is carried into the internal flow, and the water body begins to consume kinetic energy to perform sediment transport work. This nonlinear weighted mechanism can be described by the following formula:

[0186] W _geo (t)=(τ _b (t)-τ _c ) m ;

[0187] In the formula, W _geo (t) represents the weighted average of the effective geomorphic work at time step t, where τ is the weighted average of the effective geomorphic work at time step t. _b (t) represents the bottom friction shear force calculated by the model at time step t, where τ is the computational mesh. _c denoted as the critical starting shear force of the bed surface sediment at that location, and m is a specified nonlinear exponent controlling the power scaling amplitude.

[0188] In some scenarios, based on the statistical relationship of bedload sediment transport intensity, the parameter m can be assigned a value of 1.5;

[0189] τ of bed surface mud and sand _c The parameters were obtained by sieving and statistically analyzing the particle size distribution data of sediment particles.

[0190] Step 303: The transient river stability index is integraled over time using the effective work weighting of the geomorphology to obtain the single-condition stability index.

[0191] After obtaining the transient indices and corresponding effective work weights for all time steps, the equivalent indices for the entire cycle of the flood condition are obtained through discrete integration.

[0192] The specific implementation involves multiplying the transient river stability index calculated at each time step, the corresponding geomorphic effective work weighting, and the time step, and then summing them over time. Simultaneously, the sum of the products of the geomorphic effective work weighting and the time step is calculated. The ratio of the sum of the former to the sum of the latter is then calculated to obtain the single-condition stability index.

[0193] That is, HSI _q =∑(HSI(t)×W _geo (t)×dt) / ∑(W _geo (t)×dt);

[0194] In the formula, HSI _q Let W be the single-condition stability index under the q-th flood condition, where ∑ represents the summation operator over all output time steps within the complete simulation duration, HSI(t) is the transient river morphology stability index at the corresponding time step t, and W is the single-condition stability index under the q-th flood condition. _geo (t) represents the weighted weight of the effective work of the terrain calculated at time step t, and dt is the scalar increment of the time step.

[0195] According to one aspect of this application, there exists an operating state in which, if the sum of the denominators of a certain computational grid cell remains zero throughout the entire simulation cycle, the HSI of that grid cell is directly reduced. _q A value of 1 indicates that no riverbed deformation work response occurred at this location during the current working cycle.

[0196] Step 304: Calculate the frequency weights based on the reciprocal of the return period of floods of different frequencies, and use the frequency weights to weight and integrate multiple single-condition stability indices to obtain the comprehensive river stability index.

[0197] The long-term planar morphology of a river is shaped by the alternating superposition of flood events with different return periods over long periods. The probabilities of different flood conditions occurring over long periods are inversely related. To obtain a comprehensive stability index with long-term characterization capabilities, it is necessary to combine multiple previously obtained single-condition stability indices using a probabilistic dimension.

[0198] Furthermore, the frequency weight extraction process can be calculated using the following equivalent relationship:

[0199] p _q =1 / T _R ;

[0200] Above, p _q T represents the equivalent annual contribution frequency weight corresponding to the difference in occurrence frequency for the q-th flood condition. _R This refers to the design recurrence period of the flood condition.

[0201] The above can output a long-term comprehensive river morphology stability index that integrates the dynamic evolution effect over time and the probability dimension of the return period.

[0202] On the other hand, conventional solutions typically use a fixed number of control points, which can easily lead to insufficient fitting ability in sensitive areas and redundant search dimensions in flat areas. This embodiment discloses a spatial gradient-driven adaptive parameterization strategy.

[0203] Specifically, the treatment line is discretized into a sequence of control nodes, as follows:

[0204] Step 401: Calculate the spatial gradient of the comprehensive river stability index along the regulation line.

[0205] The current optimization iteration number is obtained. When the preset adaptive adjustment period is met, the coordinates of the control nodes of the regulation line for the optimal entity scheme in the current population are extracted. The cumulative arc length sequence is extracted along the curve of this regulation line, and the comprehensive river stability index of the nearby riverbed computational grid mapped to each control node is extracted. By calculating the ratio of the absolute value of the difference between evaluation indices between adjacent nodes to the arc length increment between nodes, gradient information reflecting the rate of change in the spatial distribution of river stability is obtained.

[0206] As an example, G(s) _k =abs(HSI(s) _k+1 )-HSI(s _k-1 )) / (s _k+1 -s _k-1 );

[0207] In the formula, G(s) _k That is, the spatial gradient at the k-th control node, s _k That is, the cumulative arc length coordinates of the k-th control node along the treatment line, s _k+1 With s _k-1 These are the cumulative arc length coordinates of its preceding and following nodes, respectively, HSI(s) _k+1 ) and HSI(s _k-1 ) are the comprehensive river stability indexes for the corresponding locations, and abs is the absolute value operator.

[0208] Step 402: Identify the intervals where the spatial gradient is higher than a preset first mean multiple, and insert new control nodes in the corresponding intervals.

[0209] Accordingly, calculate the global average value of the spatial gradient of all control nodes along the entire current remediation line;

[0210] For locally adjacent nodes, calculate the local mean of their gradients and determine whether the local mean is greater than the product of the global mean and a first mean multiple. The first mean multiple serves as a sensitive threshold for triggering encryption operations, and in some embodiments, it can be set to 1.5.

[0211] When the judgment result is yes, it indicates that the river morphology stability changes drastically within this interval, belonging to a region of water flow ramming or complex flow patterns, requiring a higher parameter resolution.

[0212] A new control node is inserted between the two original control nodes. The initial coordinates of the new control node are set to the midpoint of the arc length of the curve segment where the original control nodes are located on both sides, thereby improving the degree of freedom in optimizing the local spatial form.

[0213] Step 403: Identify the intervals where the spatial gradient is lower than a preset second mean multiple, and remove the original control nodes in the corresponding intervals.

[0214] Furthermore, in addition to identifying high-gradient sensitive regions, redundant dimension identification is also performed in low-gradient flat regions;

[0215] Determine whether the local spatial gradients of multiple consecutive control nodes are all lower than the product of the global average and a multiple of the second average. The multiple of the second average serves as a passivation threshold for triggering sparse operations, and in some cases can be set to 0.3.

[0216] When the judgment result is yes, it indicates that the fine-tuning of the morphology of the regulation line within this continuous interval does not have a substantial impact on the comprehensive river stability index, and there is redundancy in the decision-making dimension. Optionally, the original control nodes that are not endpoints within this low-gradient interval will be removed, thereby reducing the parameter dimension of this region and saving the overall search computing power of the intelligent optimization algorithm.

[0217] Step 404: The original set of decision variables is mapped to the adjusted parameter space using least squares fitting.

[0218] After inserting new control nodes and removing existing control nodes, the number of control points corresponding to the current rectification line changes, causing a change in the dimensional space of the decision variables of the evolutionary algorithm.

[0219] Optionally, it may be necessary to reconstruct the parameters of all individuals in the current evolutionary population.

[0220] That is, extract the spline curve discrete point set of each individual under the original decision variable set, and use the least squares fitting algorithm to inversely obtain the new control node coordinates within the updated control node number framework.

[0221] It should be understood that this operation smoothly maps the original set of decision variables to the adjusted parameter space, forming a new generation of evolutionary population.

[0222] This projection process introduces a small numerical fitting bias, which is equivalent to applying a spatial dimension variation perturbation to the population, thus helping to maintain the population diversity of the hybrid evolutionary search algorithm.

[0223] This embodiment further describes the iterative solution of the dynamic remediation line optimization model, which involves the pre-screening and training process of calling a pre-built multi-fidelity proxy model. Specifically, it includes:

[0224] Step 501: For the sampled candidate parameter set of the treatment line, a low-fidelity prior solution set is obtained using a one-dimensional hydraulic equation set.

[0225] Optionally, when the optimization algorithm performs large-scale population iterations, directly calling the global two-dimensional hydrodynamic sediment mathematical model will generate a dense computational load.

[0226] Furthermore, using the one-dimensional Saint-Venant equation and the one-dimensional sediment transport equation, a one-dimensional geometric boundary of the channel cross section is constructed for the given candidate treatment line schemes.

[0227] Solve the one-dimensional hydraulic equations to obtain the physical quantities of average water level, average flow velocity, and average scour and sedimentation thickness at each evaluation section;

[0228] This model is suitable for low-cost, large-scale sample computation. With specific parameter configurations, several sets of candidate coordinate sequences for the remediation line can be generated, and corresponding low-fidelity prior solution sets can be generated based on a one-dimensional computation process.

[0229] Step 502: For a subset of the candidate parameters of the treatment line, a two-dimensional hydrodynamic sediment mathematical model is used to obtain a high-fidelity benchmark solution set.

[0230] Based on obtaining a low-fidelity prior solution set, it is necessary to construct a high-precision comparison target to characterize complex two-dimensional hydraulic phenomena such as transverse circulation and local backflow.

[0231] A two-dimensional hydrodynamic sediment mathematical model is used to perform a complete solution under the same water flow and sediment boundary conditions. Due to the large computational resource consumption of the two-dimensional model, this step only extracts a small sample set; in some embodiments, this may be set to dozens of sets.

[0232] Specifically, several sets of sampled candidate treatment lines can be configured, the corresponding velocity vectors and bed shear forces can be extracted, and the corresponding comprehensive river stability index can be calculated and generated, which can be used as a high-fidelity benchmark solution set containing all two-dimensional water and sediment evolution characteristics.

[0233] Step 503 involves fitting the residual distribution between the high-fidelity baseline solution set and the low-fidelity prior solution set using a co-kriging network to obtain the trained multi-fidelity surrogate model; specifically including:

[0234] Step 5031: Extract local Froude numbers from the low-fidelity prior solution set to assess the prior compliance bias of the water flow deviating from the guide line.

[0235] One-dimensional hydrodynamic models assume that the water flow always moves smoothly along the longitudinal axis of the river channel, making it impossible to directly output the physical parameter of the lateral deflection angle of the flow. To map the compliance characteristics of two-dimensional flow into a low-fidelity model, the ratio of inertial force to gravity in the local flow is extracted as a substitute mapping basis. In regions with high Froude numbers, the inertia of the flow motion dominates, making fluid particles more prone to deviating from the pre-defined compliance guideline.

[0236] For example, I _align,LF =1-β _align ×Fr _local ;

[0237] In the formula, I_align,LF To assess one-dimensional flow compliance using one-dimensional parameters, β _align Fr is an empirical correction factor that reflects the trend of water flow deviation. _local The local Froude number extracted from the model.

[0238] Under normal plain river hydrodynamic conditions, the Froude number is usually below 0.5. By setting an empirical correction coefficient, the product of the two is made less than or equal to 1, so that the output correction deviation is non-negative.

[0239] For indicators such as flow turbulence and thalphon swing amplitude, which are difficult to deduce using one-dimensional longitudinal variables, a uniform value of 0 can be assigned in the low-fidelity solution set, and the system missing bias they represent can be compensated by the subsequent difference correction function.

[0240] In some scenarios, empirical correction coefficients can be determined by comparing the one-dimensional compliance estimate with the actual compliance index calculated by the two-dimensional model, and regression analysis can be used to determine empirical correction coefficients suitable for the hydraulic conditions of the specific river section.

[0241] Step 5032: Multiply the low-fidelity prior solution set containing prior correction bias by a preset scaling factor to obtain the physical prior basis.

[0242] After approximating the various one-dimensional indicators, the global scaling adjustment scalar is obtained;

[0243] Multiply the low-fidelity prior solution set with the scalar;

[0244] This scalar, as a scaling factor, is used to level the absolute numerical magnitude of the deviation between the low-fidelity prediction distribution and the high-fidelity baseline distribution on a global scale.

[0245] The output of the multiplication operation forms a priori basis with one-dimensional hydraulic physical framework characteristics, providing a reference surface with mechanism support for the residual fitting of machine learning algorithms.

[0246] Step 5033: Train a co-kriging network to fit the difference correction function between the high-fidelity benchmark solution set and the physical prior basis, and construct a multi-fidelity proxy mapping relationship that includes the prior subject and the difference correction function.

[0247] Furthermore, using a shared sample point set, the numerical residuals between the high-fidelity benchmark solution set and the physical prior basis at the same remediation line coordinate nodes are calculated.

[0248] An interpolation network model with a spatial correlation learning mechanism is constructed. The decision variable sequence of the governance line is used as the input feature sequence. The network model is trained to converge and fit the numerical residuals to form the corresponding difference correction function.

[0249] Among them, the co-Kriging network can optionally adopt a Matérn-type kernel function, and optimize the kernel hyperparameters by maximizing the marginal log-likelihood function;

[0250] The output of the difference correction function is a scalar prediction value in the same form as the comprehensive river morphology stability index.

[0251] The final surrogate model prediction process is a linear superposition of the prior subject output and the difference correction function output.

[0252] This proxy mapping relationship can be expressed as:

[0253] HSI _hat =ρ×HSI _LF +δ;

[0254] The above is HSI _hat The comprehensive river stability index (HSI) is the output of the surrogate model prediction. ρ is a pre-defined scaling factor, which can be determined by linear regression fitting of the systematic bias between the low-fidelity prior solution set and the high-fidelity benchmark solution set on the shared sample using the least squares method. _LF δ represents the low-fidelity prior solution set, and δ is the difference correction function of the fitted output of the co-Kriging network.

[0255] In a simplified implementation, the scaling factor ρ can be set to 1, in which case the physical prior basis is the low-fidelity prior solution set itself.

[0256] In this mechanism, the one-dimensional hydraulic solution already captures the main hydrodynamic evolution trends and physical constraints, thus reducing the fitting task of the machine learning network from predicting complex global river morphology responses to predicting only local residual distributions containing two-dimensional lateral and nonlinear effects. This reduces the surrogate model's dependence on the number and density of training samples, mitigating to some extent the dimensionality explosion effect caused by high-dimensional regulation line parameter optimization.

[0257] As an alternative, in systems with low control node dimensionality or redundant computing server resources, a purely data-driven proxy model based on radial basis functions can also be constructed. This approach directly trains the radial basis function network model using the sample feature set generated by Latin hypercube sampling and the high-fidelity evaluation index of the two-dimensional hydrodynamic mathematical model as input-output pairs.

[0258] The above omits the steps of generating physical priors, but the prediction error and computational cost under high-dimensional search space conditions are relatively higher than those of the multi-fidelity mechanisms mentioned above.

[0259] In some embodiments, the collaboration mechanism between NSGA-III and PSO can be specifically described as follows:

[0260] In each generation of evolution, the first half of the population is updated using the reference point-guided non-dominated sorting and crowding distance selection operator of NSGA-III, while the second half is updated using the velocity update formula guided by the global optimal position and the individual historical optimal position of PSO.

[0261] The two updated individuals are merged into a candidate pool and uniformly subjected to non-dominated sorting to enter the next generation of the population.

[0262] In some embodiments, the effective geomorphic work constituted by the supercritical shear force of the bottom water flow is introduced for time-series integration, transforming the isolated static extreme value evaluation into a continuous dynamic process that conforms to the physical work law of sediment, thus eliminating the interference of invalid hydrological duration on long-period evolution characteristics.

[0263] Furthermore, a nonlinear synthesis path based on extreme value penalty product is constructed, so that the deterioration of any single safety index can intuitively drive the overall evaluation result downward, which helps to restore the shortest board constraint and cross-cooperative deterioration mechanism in river dynamics.

[0264] In terms of improving the ability to find multiple objectives in space, redundant control nodes are dynamically stripped away and highly sensitive areas are adaptively densified by using the spatial gradient of evaluation indicators, thereby realizing the on-demand allocation of decision variable dimensions.

[0265] Meanwhile, by utilizing a one-dimensional hydraulic physics prior structure to support a two-dimensional residual fitting network, the nonlinear optimization computational load of the high-dimensional search space is reduced, thereby achieving the approximation and convergence of complex boundary morphologies while preserving the high-fidelity physical flow field analysis depth.

[0266] In this application, the operating parameters such as population size, maximum number of iterations, and convergence threshold are adaptively configured according to the dimensions of decision variables and computational resource conditions;

[0267] The specific values ​​of parameters such as the first and second mean multiples and the initial value of the penalty factor can be adaptively adjusted according to the actual river conditions, and appropriate values ​​can be determined without creative effort.

[0268] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A dynamic regulation line optimization method based on river morphology stability analysis, characterized in that, include: Obtain basic water and sediment data for the target river section; A hydrodynamic sediment mathematical model is constructed based on basic water and sediment data to simulate the evolution process and extract river morphology features. The comprehensive river stability index of the target river section is determined based on river morphology characteristics; With the goal of improving the comprehensive river stability index, a dynamic regulation line optimization model is established, with the regulation line parameters to be optimized as decision variables. The dynamic remediation line optimization model is iteratively solved to output a better remediation line scheme. Among them, a hydrodynamic sediment mathematical model was constructed based on basic water and sediment data to simulate the evolution process and extract river morphology features, including: integrating multi-source topographic depth data and photogrammetric data to construct a digital elevation model, and discretizing it using unstructured grids; setting boundary conditions that include flood process lines of different frequencies, and performing a two-way coupled simulation of water flow-driven sediment transport and riverbed deformation feedback water flow; and extracting the velocity characteristics, shear force characteristics, and thalweg oscillation characteristics of key river morphology control nodes as river morphology features. Among them, the comprehensive river stability index for the target river section is determined based on river morphology characteristics, including: calculating the flow compliance index, bank erosion risk index, flow turbulence index, and thalex swing amplitude index for the target river section based on river morphology characteristics; and performing dimensionless processing on the flow compliance index, bank erosion risk index, flow turbulence index, and thalex swing amplitude index to synthesize the comprehensive river stability index. In synthesizing the comprehensive river stability index by dimensionlessly processing the flow compliance index, bank erosion risk index, flow turbulence index, and thalex swing amplitude index, a nonlinear coupling approach is adopted. Specifically, the geometric weighted product of each dimensionless index is calculated as the index principal term; the coordinated deterioration terms of compliance and erosion and the coordinated deterioration terms of turbulent swing are calculated; and the index principal term, the coordinated deterioration terms of compliance and erosion and the coordinated deterioration terms of turbulent swing are combined to extract positive values, thus obtaining the comprehensive river stability index. The process of establishing a dynamic regulation line optimization model with the regulation line parameters to be optimized as decision variables includes: discretizing the regulation line into a sequence of control nodes, fitting the control node sequence with spline curves, and setting its coordinates as the regulation line parameters to be optimized; setting multiple constraints and transforming the multiple constraints into external penalty functions; and establishing a multi-objective evaluation system that includes minimizing the variation rate of the comprehensive river morphology stability index.

2. The method according to claim 1, characterized in that, The dynamic remediation line optimization model is solved iteratively, including: A multi-objective hybrid evolutionary algorithm is used for the search, and the mutation probability of the population is dynamically adjusted according to the diversity and dispersion of the decision variable set. Based on the current parameters of the regulation line to be optimized, the river channel calculation grid is dynamically reconstructed. When the displacement of the grid boundary node does not exceed the preset scale threshold, the incremental hot-start hydrodynamic simulation is called to update the boundary conditions.

3. The method according to claim 1, characterized in that, Before iteratively solving the dynamic remediation line optimization model, a pre-built multi-fidelity proxy model is called for pre-screening. The multi-fidelity proxy model is trained in the following way: A low-fidelity prior solution set is obtained using a one-dimensional hydraulic equation set. A high-fidelity benchmark solution set was obtained using a two-dimensional hydrodynamic sediment mathematical model. The residual distribution between the high-fidelity benchmark solution set and the low-fidelity prior solution set is fitted using a co-kriging network.

4. The method according to claim 1, characterized in that, The pre-configured homing-flushing co-degradation term is calculated as follows: Obtain the deviation between the preset ideal flow compliance state and the flow compliance index, obtain the deviation between the preset ideal bank slope erosion resistance state and the bank slope erosion risk index, and multiply the compliance deviation and the erosion deviation to obtain the compliance-erosion synergistic degradation term.

5. The method according to claim 1, characterized in that, The parameters of the regulation line to be optimized are set as the coordinate sequence of the groynes group connection line, the coordinate sequence of the revetment outline line, or the coordinate sequence of the guide dike axis. The optimal regulation line scheme is output simultaneously with the evaluation list of regulation building material consumption that meets the engineering constraints.

6. The method according to claim 1, characterized in that, The output of the optimal regulation line scheme also includes mapping the optimal regulation line scheme to the workspace of the digital twin system, and generating a three-dimensional information model of the river engineering with spatial geometric attributes based on the coordinate control node distribution of the optimal regulation line scheme.

Images