Multi-objective optimization design method for grey-green coupling facility

Abstract

The invention discloses a multi-objective optimization design method for a grey-green coupling facility. The method comprises the following steps: S100, collecting and cleaning multi-source data, performing coordinate unification, establishing a digital base data set, and setting regional parameter correction and historical disaster fingerprint calibration; s200, quantifying the regulation, storage and drainage operation rule into an optimizable dynamic variable; s300, establishing a parameterization mechanism model of the grey green facility and a connecting component, coupling life period evolution of degradation-operation and maintenance recovery, and outputting hydraulic power and risk indexes; s400, setting a Level-0 hard constraint and a robust constraint by taking the minimum flood response speed and the minimum full life cycle net cost as a core target; s500, carrying out iteration by adopting NSGA-II / III (Non-dominated Sorting Genetic Algorithm-II / III) to obtain a Pareto non-dominated feasible solution set; s600, calculating a comprehensive risk index based on the risk result and generating a partition and differentiation layout template; and S700, screening a unique recommendation scheme according to the grading evaluation system in combination with AHP-entropy weight coupling weight and landing correction, and outputting engineering accessories such as a parameter list and an operation and maintenance plan.

Description

Multi-objective optimization design method for gray-green coupled facilities Technical Field

[0001] This invention relates to the fields of smart water conservancy, rural stormwater management, sponge city infrastructure, environmental geotechnical engineering, operations research and optimization and artificial intelligence applications, and in particular to a multi-objective optimization design method for gray-green coupled facilities. Background Technology

[0002] Mountainous rural areas are characterized by undulating terrain, short runoff times, and uneven spatial and temporal distribution of rainfall, often accompanied by slope erosion and sediment-laden runoff. This makes them prone to a disaster chain: "torrential rain—runoff generation—sediment transport—siltation / erosion—facility performance degradation or failure—waterlogging / destruction." Existing stormwater management and sponge city infrastructure designs are mostly based on static scale configurations, supplemented by fixed operational rules, which present the following technical shortcomings in mountainous rural settings:

[0003] (1) Multi-source heterogeneous data (topography and geology, hydrology and meteorology, current status of facilities, red line constraints, historical disaster sites, etc.) lack unified coordinates, scale and quality control, and key parameters lack verifiable regional correction and historical fingerprint calibration, resulting in insufficient cross-regional applicability and prediction stability.

[0004] (2) The key node drainage allocation, storage target water level / available reservoir capacity, opening and closing thresholds and rules, etc., are not quantified into optimizable variables in the design stage and boundary and calibration methods are not given, making it difficult to achieve gray-green synergy and dynamic scheduling optimization.

[0005] (3) The gray-green coupling interface components (diversion / overflow / sand settling / energy dissipation, etc.) were not modeled by system parameters, and the life cycle degradation-operation and maintenance-repair process was not adequately depicted, making it difficult to uniformly calculate the full life cycle cost and safety margin.

[0006] (4) The constraints such as farmland and ecological red lines, river management, cultural relics protection / ancient tree avoidance, geological safety, construction and operation and maintenance accessibility are mostly described in principle, lacking calculable hard constraint functions, threshold sources and unified judgment criteria, which easily leads to theoretically feasible but engineering-unfeasible solutions;

[0007] (5) Even if a multi-objective solution set is formed, there is often a lack of a hierarchical evaluation framework and verifiable weight determination rules for finally selecting a unique solution from the Pareto feasible solution set. Furthermore, there is a lack of robust control over opportunity constraints or risk measures (such as CVaR) for factors such as rainfall enhancement, blockage chain events, and monitoring uncertainties.

[0008] Therefore, there is an urgent need for a multi-objective optimization design method for gray-green coupled facilities that integrates dynamic decision variable quantification, gray-green coupling mechanism and life cycle process modeling, multi-objective optimization solution, hard constraint / robust constraint determination, risk partitioning and templated output, and verifiable hierarchical final selection decision on a unified data foundation. Summary of the Invention

[0009] In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention is to provide a multi-objective optimization design method for gray-green coupling facilities, which is used for flood control in mountainous rural areas and solves problems such as gray-green coupling failure, rigid dynamic scheduling, and poor regional adaptability.

[0010] To achieve the above objectives, the present invention provides a multi-objective optimization design method for gray-green coupling facilities, comprising the following steps:

[0011] S100, Construction of Mountainous Rural Environmental Perception and Digital Base Dataset

[0012] Collect multi-source data to construct a digital foundation dataset for rainwater management in mountainous rural areas; establish a set of regional parameter correction parameters or correction functions, and introduce historical disaster fingerprints for model calibration and consistency verification;

[0013] S200, Determination and Quantitative Representation of Dynamic Decision Variables

[0014] The control rule parameters during the operation period are quantified into decision variables that can be optimized in the design phase, and the value boundaries, constraints and calibration criteria are given for each decision variable.

[0015] S300, gray-green coupled system mechanism model and full life cycle dynamic evolution modeling

[0016] Establish a parameterized model of gray-green facilities and connecting components, output hydraulic response, risk indicators, operation and maintenance burden and life-cycle performance degradation trajectory under different scenarios and control rules, and generate a verifiable time series output table or statistical output table.

[0017] S400, Multi-objective Optimization Model Construction

[0018] A multi-objective optimization model is constructed based on the output table of S300, a Level-0 hard constraint system is established, and the constraint function judgment criteria are given; and the objectives and constraint criteria are updated after the model parameters are calibrated based on newly added monitoring data.

[0019] S500, Solving and Obtaining the Pareto Non-Dominated Feasible Solution Set

[0020] We propose a constraint handling strategy for Level-0 hard constraints and probabilistic design, and control convergence and diversity through mechanisms such as reference points or crowding; outputting a Pareto non-dominated feasible solution set that satisfies the constraints.

[0021] S600, Generation of facility classification and differentiated layout templates based on risk outcomes

[0022] Based on the risk-related output of S300, the comprehensive risk index RI is calculated and classified, and the region is divided into zones; gray and green facilities are classified by function; and the "coordinate-scale-connection structure-dynamic rules" of the Pareto non-dominated feasible solution are mapped into differentiated layout templates, and gray and green facility layout and connection structure schemes that meet the risk level requirements and red line constraints are matched or generated from the template library.

[0023] S700, Target Classification and Comprehensive Evaluation and Decision-Making of Solutions

[0024] Establish a hierarchical evaluation system; adopt AHP-entropy weight coupling dynamic weight and set weight interval constraints in combination with village type. First, positively normalize each indicator according to "the larger the better / the smaller the better" and normalize it to [0,1], and then calculate the information entropy and dispersion to obtain the objective weight; take the approval / implementation difficulty as the implementation correction factor, and select the unique recommended solution from the Pareto non-dominated feasible solution set;

[0025] Specifically, the digital base dataset, red line constraints, and regional correction results output by S100 are used to define the domain, boundary conditions, and calibration benchmarks of the decision variables in S200; the dynamic decision variables and static construction parameters output by S200 are used as inputs to the S300 mechanistic model to obtain hydraulic response, risk indicators, and life-cycle degradation trajectories, and are fed back to S400 to construct the objective function and constraint judgment criteria; S500 iteratively searches and iteratively calls S300 under the objectives and constraints defined by S400 to complete the scheme evaluation and feasibility judgment to obtain the Pareto non-dominated feasible solution set; S600 calculates RI based on the solution set and completes risk zoning, facility classification, and templated mapping; S700 selects the unique recommended scheme from the solution set based on hierarchical evaluation, coupling weights, and feasibility correction.

[0026] The beneficial effects of this invention are:

[0027] (1) Unified data base and more stable regional adaptation. By constructing a digital base dataset that includes topography and geology, hydrology and meteorology, current status of facilities, red line constraints and historical disaster fingerprints, and introducing regional parameter correction and fingerprint calibration mechanisms, the input data coordinates, scale and parameter calibrators are unified, thereby improving the cross-regional applicability and computational stability of the model.

[0028] (2) Quantify the operation scheduling into design variables to achieve dynamic optimization of gray-green coordination. Move the operation control variables such as drainage allocation function parameters of key nodes, water level / available storage sequence, opening and closing threshold and rule function to the design stage as optimizable decision variables and give the boundary and calibration calibrator, so that the scheme can achieve gray-green coordinated scheduling optimization for the whole process of rising water level - peak water level - receding water level, and avoid scheduling rigidity caused by static design.

[0029] (3) The full life cycle evolution evaluation is more realistic, reducing the risk of interface fragility and durability degradation. By parametrically modeling gray-green facilities and connecting components, and introducing life cycle processes such as sediment blockage degradation, scouring loss, operation and maintenance recovery, water quality energy consumption and structural durability aging, verifiable hydraulic response, risk indicators, operation and maintenance burden and degradation trajectory are output, thereby more accurately identifying long-term failure mechanisms and life cycle costs.

[0030] (4) Unified criteria for objectives and constraints, results can be reviewed and verified and are more feasible. A multi-objective model is constructed with flood response speed (F1) and net life cycle cost (F2) as the core objectives, and a unified judgment criteria for Level-0 hard constraints and probabilistic / robust constraints is established, so that candidate solutions can be calculated and screened under compliance, safety and uncertainty conditions, reducing the number of theoretically feasible but engineering-infeasible implementation schemes.

[0031] (5) Obtain a set of non-dominated feasible solutions that satisfy the constraints, taking into account both convergence and diversity while maintaining controllable efficiency. By jointly encoding static construction parameters and dynamic rule parameters, NSGA-II or NSGA-III iterative search is performed according to the target dimension. Combined with constraint handling, crowding / reference point diversity preservation, and surrogate / multi-fidelity acceleration, a set of Pareto non-dominated feasible solutions that satisfy the constraints is output within a given computational budget.

[0032] (6) The transformation chain from solution set to engineering template and final selection scheme is complete and the decision is interpretable. Based on the simulation output, risk zoning is calculated and facility function classification is formed, and feasible solutions are mapped to differentiated layout templates; by superimposing approval / implementation difficulty correction through hierarchical evaluation system and AHP-entropy weight coupling weight (entropy weight is calculated after normalization of feasible solution set index matrix), a unique recommended scheme is selected from the solution set, and engineering attachments such as parameter list, layout table, rule table and operation and maintenance plan are output. Attached Figure Description

[0033] Figure 1 is a schematic diagram of the process of S100-S700 of the present invention. Detailed Implementation

[0034] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0035] Referring to Figure 1, the multi-objective optimization design method for gray-green coupling facilities in this embodiment includes the following steps:

[0036] S100, Construction of Mountainous Rural Environmental Perception and Digital Base Dataset

[0037] Data from multiple sources, including topography, geology, hydrology, meteorology, current infrastructure, red line constraints, and historical disaster fingerprints, were collected. Data cleaning, unified coordinate system, and scale transformation were completed to construct a digital foundation dataset for rainwater management in mountainous villages. A set of regional parameter correction parameters or correction functions were established to correct parameters related to rainfall, soil infiltration, temperature freeze-thaw cycles, and vegetation seasonality. Historical disaster fingerprints were introduced for model calibration and consistency verification, forming a data structure and field model that can be uniformly called upon for subsequent modeling, optimization, and operation control.

[0038] S200, Determination and Quantitative Representation of Dynamic Decision Variables

[0039] The control rule parameters during the operation period are quantified into decision variables that can be optimized in the design phase. The decision variables include at least: drainage flow distribution function parameters of key nodes, available storage capacity or target water level sequence of storage facilities, opening and closing thresholds and rule function parameters; and the value boundaries, constraints and calibration standards are given for each decision variable.

[0040] S300, gray-green coupled system mechanism model and full life cycle dynamic evolution modeling

[0041] The simulation of the full life cycle dynamic evolution model is performed according to the following steps: 1) Input static parameters and dynamic rule parameters; 2) For each simulation scenario (s) and each year (y), perform the following cyclical execution with a time step Δt: a) Calculate hydrodynamics based on the current state (S310); b) Calculate sediment transport and deposition based on flow rate and sediment concentration (S330-S340); c) Update the facility flow capacity (S320, S340); d) Determine whether maintenance is triggered (S370), and update the state if so; e) Calculate scour risk (S350) and durability index (S380); 3) Accumulate annualized costs and risk indexes; 4) Output time series results and statistical reports.

[0042] Establish a parameterized model of gray-green facilities and connecting components, and construct a dynamic evolution model of the entire life cycle, including sedimentation and degradation, scouring loss, operation and maintenance recovery, water quality indicators and energy consumption, and structural durability aging. Output hydraulic response, risk indicators, operation and maintenance burden and life-cycle performance degradation trajectory under different scenarios and control rules, and form a verifiable time series output table or statistical output table.

[0043] S400, Multi-objective Optimization Model Construction

[0044] A multi-objective optimization model is constructed based on the output table of S300, with at least "optimal flood response speed (minimum F1)" and "minimum net cost over the entire life cycle (minimum F2)" as the core objective functions, and optional extended comprehensive benefit-related objectives; a Level-0 hard constraint system is established and the constraint function judgment criteria are given, while probabilistic constraints or robust constraints are constructed to deal with uncertainties; and the objectives and constraint criteria are updated after the model parameters are calibrated based on newly added monitoring data.

[0045] S500, Solving and Obtaining the Pareto Non-Dominated Feasible Solution Set

[0046] Joint encoding is used to describe static construction parameters and dynamic rule parameters. NSGA-II or NSGA-III is selected for iterative search according to the target dimension. Constraint handling strategies are designed for Level-0 hard constraints and probabilistic / robust constraints, and convergence and diversity are controlled by mechanisms such as reference points or crowding. Proxy models or multifidelity simulations can be used to accelerate computation and improve efficiency. The output is a Pareto non-dominated feasible solution set (approximate Pareto front solution set) that satisfies the constraints.

[0047] S600, Generation of facility classification and differentiated layout templates based on risk outcomes

[0048] Based on the risk-related output of S300, the comprehensive risk index RI is calculated and classified, and the region is divided into high / medium / low risk zones; gray-green facilities are classified into backbone defense type, peak shaving buffer type and source emission reduction type; and the "coordinate-scale-connection structure-dynamic rule" of Pareto non-dominated feasible solution is mapped into differentiated layout templates, and gray-green facility layout and connection structure schemes that meet the risk level requirements and red line constraints are matched or generated from the template library.

[0049] S700, Target Classification and Comprehensive Evaluation and Decision-Making of Solutions

[0050] A tiered evaluation system is established, consisting of "Level-0 veto, Level-1 core objective-driven, Level-2 important objective threshold bonus, and Level-3 auxiliary indicator deduction." An AHP-entropy weight coupling dynamic weighting method is adopted, with weight range constraints set based on village type. The entropy weighting method uses the index value matrix within the feasible solution set output by S500 as input. Each index is first positively oriented according to "larger is better / smaller is better" and normalized to [0,1] before calculating information entropy and dispersion to obtain objective weights. Approval / implementation difficulty is used as a feasibility correction factor. A unique recommended solution is selected from the Pareto non-dominated feasible solution set (and its templated expression), and an engineering attachment is output. This attachment includes at least a parameter list, layout table, rule table, and operation and maintenance plan.

[0051] Specifically, the digital base dataset, red line constraints, and regional correction results output by S100 are used to define the domain, boundary conditions, and calibration benchmarks of the decision variables in S200; the dynamic decision variables and static construction parameters output by S200 are used as inputs to the S300 mechanistic model to obtain hydraulic response, risk indicators, and life-cycle degradation trajectories, and are fed back to S400 to construct the objective function and constraint judgment criteria; S500 iteratively searches and iteratively calls S300 under the objectives and constraints defined by S400 to complete the scheme evaluation and feasibility judgment to obtain the Pareto non-dominated feasible solution set; S600 calculates RI based on the solution set and completes risk zoning, facility classification, and templated mapping; S700 selects the unique recommended scheme from the solution set based on hierarchical evaluation, coupling weights, and feasibility correction.

[0052] Specifically:

[0053] The digital base dataset, red line constraint layer, and regional parameter correction results output by S100 are used to define the domain, boundary conditions, and calibration benchmarks of the dynamic decision variables (function parameters, thresholds, and sequences) of S200. The dynamic decision variables output by S200 serve as control inputs and, together with the static construction parameters, serve as inputs to the S300 mechanism model and the full life cycle dynamic evolution model. This yields evaluation quantities such as hydraulic response, risk indicators, operation and maintenance triggers, and life cycle degradation trajectories. These evaluation quantities are then fed back to S400 to establish an objective function system centered on the flood response speed target F1 and the full life cycle net cost target F2, as well as a unified calculation method for Level-0 hard constraints, probability / robust constraints, constraint functions and judgment criteria, life cycle, and discounting.

[0054] Under the objective function and constraints defined by S400, S500 uses joint encoding and NSGA-II / NSGA-III for iterative search. During the iteration process, S300 is called repeatedly to complete the evaluation of candidate solutions and feasibility determination. Combined with constraint handling, surrogate model or multi-fidelity simulation acceleration and convergence-diversity control, it outputs a Pareto non-dominated feasible solution set (approximate Pareto front solution set) that satisfies the constraints.

[0055] Based on the risk and performance evaluation results of the feasible solution set, S600 completes the classification of high / medium / low risk zones and facilities such as backbone defense type, peak shaving buffer type, and source emission reduction type. It also maps the "coordinate-scale-connection structure-dynamic rules" of the feasible solution into differentiated layout templates, and matches them with the template library or generates gray-green coupling engineering schemes that meet the risk level requirements and red line constraints.

[0056] S700 performs a graded evaluation on the templated engineering scheme and introduces approval / implementation difficulty as a feasibility correction factor, selecting a unique recommended scheme from the Pareto non-dominated feasible solution set and its templated expression.

[0057] Preferably, S100 further includes:

[0058] S110, Multi-source heterogeneous data acquisition and cleaning

[0059] S111. Collect and construct a full-element database. It must include at least:

[0060] (1) Topographic data: acquire high-precision DEM (preferably ≤0.5m resolution), calculate slope raster (S(x,y)), curvature, and runoff accumulation. Valley lines and potential runoff paths; extraction of village micro-topography (roads, low-lying areas in courtyards, gully heads, steep slopes);

[0061] (2) Hydrometeorological data: Rainfall stations or radar rainfall grid points, forming a spatiotemporal field of rainfall (I(x,y,t)); water level / discharge at key sections; historical design rainstorm parameters (recurrence period, duration, rainfall pattern);

[0062] (3) Sediment data (optional high sediment module): Turbidity sensor or sampling monitoring to obtain (Turb(t)); sediment content is obtained through inversion model. And obtain the particle size distribution according to the particle size group. With component mass fraction ;

[0063] (4) Facilities and topology data: existing pipe networks, ditches, culverts, ponds, dams, pumping stations, gate valves, existing greening facilities, etc.; form a directed topology graph (G) and label the hydraulic properties and controllable properties of each edge;

[0064] (5) Geological and disaster data: lithology, soil layer thickness, permeability coefficient, fault fracture zone; boundaries of landslide / collapse / debris flow hazard points ; as well as sections damaged by historical flooding, culvert blockages, etc.;

[0065] (6) Red Line and Human Data: Permanent Basic Farmland Boundary Ecological red line Scope of River Management Cultural Relics Protection Scope Collection of Ancient and Famous Trees and protection radius Historical and cultural corridors and visually sensitive areas;

[0066] (7) Construction and operation conditions: road grade, width, slope, bridge and culvert load limits, types of machinery that can enter; material sources and transportation distance.

[0067] Data cleaning employs filtering and consistency checks: Kalman filtering (KF) or robust filtering (Huber-KF) is used to denoise time series data, piecewise interpolation and uncertainty labeling are used for missing data, and a unified coordinate system is used for spatial data and topological consistency checks are performed (node ​​connectivity, edge direction, and integrity of cross-sectional attributes).

[0068] S120, Regional Parameter Correction Model

[0069] To address the issue of the general model failing to adapt to different regions, regional adjustments are made to key parameters. Let the baseline value for any parameter to be adjusted be... The reference values ​​for the key parameters in a preset standard reference area or general model can be determined by consulting relevant technical manuals, standards and specifications, or based on typical regional test data (e.g., storage capacity coefficient, infiltration-related parameters, roughness, maintenance cycle, etc.). Local parameters are obtained by correcting for factors such as rainfall, soil infiltration, temperature freeze-thaw cycles, and vegetation phenology. :

[0070] ;in, This is a rainfall correction factor (determined by annual rainfall, extreme rainfall intensity, etc.); Soil / permeability correction factor (determined by saturated hydraulic conductivity, soil texture, karst development, etc.); Temperature / freeze-thaw correction factor (determined from meteorological data); The vegetation cover / phenology correction factor (determined by vegetation cover and LAI, etc.).

[0071] To ensure repeatability, it is stipulated that local and ref must be valued under the same statistical caliber (same unit, same return period and duration, same statistical time window), and all correction factors must be positive; in this embodiment, each factor can be limited to To avoid distortion caused by extreme data.

[0072] S121, Rainfall Correction Factor

[0073] For parameters such as the storage capacity coefficient that are positively correlated with the design rainfall intensity, the following values ​​can be used:

[0074] ;in:

[0075] For the same return period Simultaneous calendar Design rainfall intensity under the given conditions (units are based on the adopted rainstorm intensity formula, usually in mm / h).

[0076] The results were calculated using the local storm intensity formula (IDF) for the study area, or obtained through frequency analysis based on local rain gauge sequences.

[0077] Take and Reference regions (refs) with consistent sources, in the same The results are obtained using the local IDF formula or frequency analysis. In this embodiment... As given as design conditions (e.g., once every 5 years, duration of 1 hour, or determined according to project standards).

[0078] S122, Soil / Permeability Correction Factor

[0079] For permeation-related parameters, the following values ​​are acceptable:

[0080] ;in:

[0081] Saturated hydraulic conductivity / saturated permeability coefficient (local and ref units should be consistent, such as m / s or mm / h); The penetration sensitivity index is dimensionless and ranges from 0.5 to 1.2.

[0082] The data is obtained from geotechnical laboratory permeability tests, soil databases, or field infiltration tests (such as double-ring infiltration).

[0083] The acquisition of: comes from and Corresponding reference area (ref) with test / database / field test data of the same caliber;

[0084] Determination: Calibration is preferably performed using historical rainfall-runoff / infiltration observations; when historical observations are lacking, this embodiment uses... .

[0085] when When the area is large and there is a risk of landslides, an "infiltration inhibition factor" (see S400 hard constraint) can be added to avoid excessive infiltration leading to an increase in geological risk.

[0086] S123, Temperature / Freeze-Thaw Correction Factor

[0087] This is used to characterize the effects of temperature and freeze-thaw cycles on parameters such as material durability, permeability decay, and maintenance frequency. The data source is weather station data or reanalysis meteorological data. This example uses the "annual freeze-thaw cycle count" indicator:

[0088] ;in:

[0089] This refers to the number of freeze-thaw cycles that cause the temperature to rise above 0°C within a year; in areas without freeze-thaw cycles, the following can be used: .

[0090] S124, Vegetation / Phenology Correction Factor

[0091] This is used to characterize the effects of vegetation cover and seasonality on parameters such as roughness, interception, infiltration attenuation, and maintenance frequency. This example uses the Leaf Area Index (LAI).

[0092] ;in:

[0093] LAI stands for Leaf Area Index (dimensionless).

[0094] Sourced from remote sensing inversion LAI products or quadrat monitoring / surveys;

[0095] LAI data derived from the reference region (ref) within the same seasonal window.

[0096] Through the above-mentioned regional corrections, the model parameters can be adapted to regional differences in rainfall, soil infiltration, climate freeze-thaw cycles, and vegetation phenology, thereby improving cross-regional applicability and prediction stability.

[0097] S130, Historical Disaster Fingerprint Calibration

[0098] Collection of historical flood points , old disaster spot collection Destroyed by history And select a set of representative historical rainstorm events. As a calibration sample, the objective function is calibrated based on the "error of the highest water level at critical nodes":

[0099] ;in:

[0100] For the first Planar coordinates of historical flood points (unified coordinate system);

[0101] The highest observed water level at that point (the water level recorded by flood markers or water level stations, with a unified elevation datum).

[0102] This refers to the time corresponding to the highest water level at that point during the event; when historical data cannot provide precise... At that time, this embodiment specifies The absolute value of the difference between the simulated maximum water level and the observed highest water level.

[0103] For the simulation model at point With time The simulated water level; the point value extraction adopts either "nearest grid value" (selected in this embodiment) or bilinear interpolation, and the one selected is determined to be unique.

[0104] The weights are point weights (dimensionless). This embodiment uses equal weights. Optionally, when the point belongs to the old disaster point set. Or located in the historical erosion section When within the range of influence, take Take the rest .

[0105] To ensure the reliability of the calibration results within the interpretable range of measurement and terrain errors, a mean absolute error threshold is used as a constraint:

[0106] ;

[0107] Where: threshold (Unit: m) represents the upper limit of acceptable calibration mean absolute error, determined by the observed measurement error. Error with Digital Elevation Model (DEM) The overall composition is as follows: ,in:

[0108] The value can be determined based on the nominal accuracy of the measuring instrument used. For example, 0.02-0.05m can be used when using RTK measurement, and 0.005-0.01m can be used when using a level.

[0109] The value can be determined based on the product accuracy level of the DEM data used. For example, when using an airborne LiDARD DEM with a resolution of 0.5m, 0.05-0.1m can be used, and when using an ALOS DEM with a resolution of 12.5m, 2-5m can be used.

[0110] In this embodiment, a 0.5m LiDAR DEM and a level were used for measurement. , Therefore .

[0111] Preferably, S200 further includes:

[0112] S210, Drainage flow distribution variable Functional expression

[0113] This invention defines the "inflow ratio to green space" of key nodes within rural areas (interception wells, diversion weirs, inlets of regulating reservoirs, and gray-green space connections) as an optimizable dynamic function. For any split node (i), let the gray-side inflow be... The inflow of green space is The main ash inlet channel flow rate is .Will Parameterized as a composite Sigmoid function of "water level-turbidity-capacity constraints":

[0114] ,in:

[0115] ;

[0116] This is the maximum split ratio;

[0117] To activate the midpoint water level;

[0118] This is the water level sensitivity coefficient;

[0119] The turbidity suppression threshold;

[0120] This is the turbidity sensitivity coefficient;

[0121] Capacity constraint mapping (automatically reduce the inflow diversion when the overflow capacity of the connection port is reduced due to blockage).

[0122] Recommended: ,in This represents the initial current carrying capacity. The decision variable vector contains... These parameters enable "rules to be optimized".

[0123] The initial range of values ​​for the decision variables can be preliminarily set based on engineering experience, facility physical limits, and local hydrogeological conditions, and will serve as the search space during the optimization process. Their specific values ​​will be determined through subsequent optimization searches in S500.

[0124] S220, Dynamic water storage capacity / target water level Quantification

[0125] For the water storage facility (k) (grey pond or green pond), define the target water level sequence. Or available storage capacity As a dynamic decision variable, the "forecast level - segmented ladder" function is adopted:

[0126] ,in:

[0127] High water level during normal times (resource utilization / landscape / ecology);

[0128] The height that can be vacated;

[0129] It is a step function that monotonically increases with the forecast level (the higher the warning level, the lower the target water level).

[0130] Define the advance time of leukorrhea As a variable, so that at the time of warning triggering Pre-ejaculation:

[0131] ;in Low water levels were observed under the heavy rain warning. The capacity of the pumping station and the downstream load-bearing capacity can be jointly limited. This enables a dynamic balance of "storing water during normal times, releasing water before rain, reducing flood peaks, and returning water after floodwater recedes".

[0132] S230, Dynamic Variable Sets and Joint Encoding

[0133] Final set of dynamic decision variables:

[0134] In the optimization, instead of directly solving point-by-point on the time series, the solution is performed on... and Solving for "function parameters" and "rule parameters" achieves dimensional controllability and engineering feasibility.

[0135] Preferably, S300 further includes:

[0136] S310, Hydrodynamics and Confluence Model (Rapid Confluence in Mountainous Areas)

[0137] A distributed or semi-distributed rainfall-runoff-catchment model is used to divide the watershed into sub-catchment units. and with a unified time step Discrete calculations are performed (1–10 min in this embodiment). Flow generation can be represented using SCS-CN or a modified Green-Ampt; confluence can be represented using kinematic waves or simplified Saint-Venant equations.

[0138] S320, Green Connector Component Parameterization

[0139] The gray-green connection interface is considered as a set of key nodes. For each connection node, define an optimizable geometric parameter: sediment volume. Tangential inflow angle Grid spacing , dam height Overflow outlet width Length of energy dissipation pool Flood level And so on, and the hydraulic capacity of the connecting nodes is expressed as:

[0140] ;in:

[0141] : The set of connecting nodes is determined by the planning layout (interface locations such as canal-wetland, pipe-storage tank, ditch-sedimentation well, etc.).

[0142] (m³) (° or rad) (mm) (m) (m) (m) (each): Geometric parameters for connection construction, which can be used as optimization decision variables, and their boundaries are given by the scope of the engineering construction (e.g., according to standard drawings / construction process constraints).

[0143] Materials: Lining type, concrete grade, masonry / slope protection form, etc., from design selection or standard drawings.

[0144] Siltation status: Use subsequent S340 Or grid blockage rate Characterization.

[0145] Capacity function. To ensure repeatability, this embodiment uses a combination of rule formulas: the broad-crested weir formula is applied to the weir flow. For orifice flow, the orifice free outflow formula is used: For the bar screen, the local resistance coefficient method is used (see "Code for Design of Water Supply and Drainage" GB 50015), and for the energy dissipation tank / drop structure, the energy loss coefficient method is used (see "Hydraulic Engineering Design Manual" Volume II). The system is dynamically updated under the degradation terms of S340 / S350. .

[0146] S330, sediment module (optional high sediment module) and turbidity inversion

[0147] When the watershed has high sediment content or debris flow gullies, the high sediment module should be activated (the high sediment module should be activated when historical monitoring or geological surveys indicate that the average sediment content at the watershed outlet section under design rainfall is greater than 10 kg / m³, or when there are identified debris flow gullies within the watershed). Establish the relationship between turbidity and sediment content: ;in:

[0148] Sediment content / suspended sediment concentration, unit: kg / m³;

[0149] Turbidity, measured in NTU, is derived from observations by a turbidity sensor.

[0150] The regression coefficients were obtained using a "synchronous sampling-regression calibration" method: water samples were taken at multiple points during representative rise and fall water events, and the coefficients were measured using a filtration and weighing method. , and synchronization The sample is composed of coefficients, which are obtained by fitting the sample using nonlinear least squares or robust regression.

[0151] Settling velocity by particle size group: Stokes' law is used for fine particles (Rubey's formula can be used for coarse particles):

[0152] ;in:

[0153] Particle size group Settlement velocity (m / s);

[0154] Sediment density (kg / m³) is derived from material testing or empirical values ​​and can be set.

[0155] Water density (kg / m³), refer to the table from the water temperature or take 1000 kg / m³;

[0156] Particle size (m), derived from sieving / laser particle size analyzer detection;

[0157] Dynamic viscosity (Pa·s) is calculated from a table or empirical formula based on water temperature.

[0158] 9.81 m / s².

[0159] This step converts the "rapid increase in turbidity" into a sediment content input, providing a driving force for subsequent siltation degradation and sediment discharge control.

[0160] S340, Stagnation and Degradation Model (Functional Aging)

[0161] Define the available overcurrent capacity of the connected nodes. With the cumulative silt volume attenuation:

[0162] ,in:

[0163] Initial capacity (m³ / s or equivalent capacity index) is calculated by S320 under "no siltation / no blockage" conditions or given by design capacity;

[0164] This is the clogging sensitivity coefficient (dimensionless), reflecting the intensity of the impact of clogging volume on the attenuation of flow capacity. Recommended values ​​are as follows:

[0165] (1) Calibration method: If there are historical dredging records and corresponding flow capacity test data, the formula can be used for inversion fitting;

[0166] (2) Empirical values: Values ​​are selected based on the type of sediment transport fluid and the flow regime in the sedimentation tank. For sedimentation processes dominated by fine sand with stable flow, a value of 1.0-1.5 can be used; for scouring environments containing gravel and with rapid flow, a value of 0.5-1.0 can be used; for easily compacted cohesive silt, a value of 1.5-2.0 or higher can be used. In this embodiment, a uniform value is used for sediment-laden runoff commonly found in mountainous rural areas. .

[0167] The cumulative sediment volume is determined by the inflow sediment load and the capture efficiency:

[0168] ;in:

[0169] Inflow rate (m³ / s), output from S310 / S320;

[0170] : Sand content (kg / m³), obtained from S330 inversion or sampling;

[0171] : Sediment density (kg / m³);

[0172] The capture efficiency (0–1) is related to the sediment structure, residence time, and flow regime. In this embodiment, the following method is used: ;in The equivalent settling velocity is obtained by weighting the settling velocities of particle size groups according to their proportions or by taking the dominant particle size. The effective settlement area (m², calculated from the effective geometry of the settling zone). Q is the flow rate (inflow or through flow) of this settling / retention unit.

[0173] When a grid is present, define the grid clogging rate. And add a penalty to the overcurrent capacity:

[0174] ;in It can be empirically calibrated based on floating debris load and cleaning frequency, or obtained through statistical regression from inspection / pressure difference / water level difference; this embodiment allows for... Set it as a piecewise constant and reduce it to the preset level after a "grid cleaning" maintenance action occurs. For the connection port / bar / pipe section at the water level The available flow capacity (m³ / s) is obtained by hydraulic calculations (such as Manning's formula / local loss) and by adding the blockage reduction factor.

[0175] S350, scouring loss model and structural failure risk

[0176] Calculate the bed shear stress for structures such as interfaces, drop structures, and energy dissipation pools: ;in:

[0177] The stress is the shear stress (Pa). The density of water is (kg / m³). The hydraulic radius is (m). The energy slope (dimensionless) is obtained from hydrodynamic output or cross-sectional geometry calculations.

[0178] when Erosion occurs at (critical shear force, Pa), erosion depth / loss amount (m) Growth takes the form of overthreshold power: ;in, This is the scouring rate coefficient. These are empirical indices, and their values ​​depend on the properties of the bed material, which can be determined through indoor erosion tests or engineering analogies. For example, for loose sandy soil, Approximately 1.5 is acceptable.

[0179] Mapping structural losses to failure probabilities: ;in: (1 / m) is obtained from historical damage statistical regression or reliability target calibration. This module transforms "scour damage" into quantifiable risk and cost items, which are then incorporated into F1 / F2 and subject to constraints.

[0180] S360, Vegetation Phenology and Hydraulic Response (Seasonalization of Green Facilities)

[0181] The roughness and infiltration capacity of green facilities vary with the seasons. The Manning roughness is defined as a variation with leaf area index (LAI): ;in:

[0182] The equivalent roughness (dimensionless);

[0183] Determined by consulting tables or conducting on-site assessments based on vegetation type;

[0184] Obtained by remote sensing inversion or quadrat monitoring. Determined by statistical analysis of historical year sequences.

[0185] Define the effective permeability coefficient as a function of blockage / water content: ;in:

[0186] (m / s or mm / h) is the effective permeability coefficient;

[0187] The initial permeability coefficient (obtained from geotechnical tests / infiltration tests / soil database);

[0188] For the blockage / siltation index (in this embodiment, we take...) (or equivalent indicators);

[0189] Sensitivity coefficient (determined by on-site infiltration attenuation observations or experience from similar projects);

[0190] The cumulative siltation volume (m³)

[0191] For reference volume (m³), the effective volume of the sediment or the characteristic volume of the connection can be taken and obtained by geometric calculation.

[0192] This module is used to improve the credibility of continuous events and full lifecycle simulations.

[0193] S370, Operation and Maintenance Recovery Model (Dredging / Inspection / Replacement)

[0194] Define the set of operation and maintenance actions : Dredging, screen cleaning, repair, component replacement, etc. Triggering rule: When or Dredging / cleaning is triggered when the structure is damaged. Repair is triggered at specific times. The threshold is among them. Based on operational procedures / reliability targets or experience settings, this embodiment can be set to quantifiable rules such as "capacity drops to 70% of the initial value" or "accumulation reaches 50% of the effective volume". The effective thickness loss or scour reduction depth (m) of the critical section of the connecting component relative to the initial state is obtained from maintenance measurements or durability / scour models.

[0195] Dredging and Restoration:

[0196]

[0197]

[0198] In the formula:

[0199] For the connection port / bar / pipe section at the water level The available flow capacity (m³ / s) is obtained from hydraulic calculations and can be corrected according to the blockage reduction factor;

[0200] The initial flow capacity (m³ / s) is determined by structural dimensions and design conditions through calculation or verification.

[0201] The cumulative sediment volume (m³) is obtained by integration from a sediment deposition model or by inversion from dredging measurements / cross-sectional measurements.

[0202] For reference volume (m³), the effective volume of the sedimentation zone or the characteristic volume of the connection port is taken and obtained by geometric calculation;

[0203] For the volume removed in a single dredging operation (m³ / operation), it can be calculated as follows: Calculation, where The sludge removal capacity of the equipment per unit time (m³ / h). The operation duration (h) is determined by equipment parameters or maintenance records;

[0204] The blockage sensitivity coefficient (dimensionless) is determined by regression analysis of historical data on "flow capacity - siltation volume" before and after dredging or by experience from similar projects.

[0205] , These represent the moments before and after the maintenance actions.

[0206] S380, Structural durability and aging dimension ( )

[0207] Key components of gray facilities (pool walls, well chambers, channel linings, culverts, pump station foundations, etc.) are included in the durability aging model.

[0208] S381, Carbonization Depth Model

[0209] ;in The carbonization depth is shown in mm. Service time (in years) Carbonization coefficient ( The value can be determined based on the concrete grade, ambient humidity, and other conditions, referring to the recommended values ​​in the appendix of the "Standard for Durability Design of Concrete Structures" or through testing.

[0210] S382, Steel Corrosion Rate Model

[0211] when Achieve protective layer thickness Or corrosion is initiated when chloride ions reach a threshold:

[0212] ;in:

[0213] The corrosion rate is an indicator (set to 1 / year in this embodiment);

[0214] The baseline corrosion rate;

[0215] (%)and (°C) is from meteorological data or on-site monitoring;

[0216] Data from environmental monitoring / soil and water testing or regional environmental data;

[0217] To correct the dimensionless function, this embodiment uses a piecewise function or empirical formula with default parameters:

[0218] ,

[0219] .

[0220] S383, Load Degradation and Failure Probability

[0221]

[0222]

[0223] In the formula:

[0224] The load-bearing / stiffness reduction factor (0-1) is related to durability.

[0225] The probability of durability failure (0-1); The carbonization depth is shown in mm. The ratio of the two values ​​represents the thickness of the protective layer (mm). Dimensionless carbonization transmittance;

[0226] To ensure dimensional consistency, a dimensionless corrosion damage index is defined as the corrosion rate (1 / year). ,in To calculate the time step (in years), the parentheses should be used. It is a dimensionless durability degradation index;

[0227] The durability degradation sensitivity coefficient (dimensionless) can be obtained by statistical fitting of historical maintenance / failure data, or based on a given design life. Probability of target failure Under the conditions, according to Rate determined.

[0228] S384, Cost Linkage (Enter F2)

[0229] ;in:

[0230] For the first Year, Lifespan (years);

[0231] The probability threshold for triggering repair (set to 0.05–0.1 in this embodiment);

[0232] For the first The annual durability failure probability was calculated according to the load degradation and failure probability model of S383.

[0233] This is an indicator function that takes the value 1 if the condition within the parentheses is true, and 0 otherwise.

[0234] The unit repair cost is derived from cost quotas / bill of quantities or settlement data of similar projects;

[0235] The amount of repair work is determined according to the size of the component and the rules governing the extent of damage (e.g., calculated based on the percentage of damaged area or damaged length).

[0236] As an expectation operator, this embodiment uses the scenario set. By weight Take the weighted average, that is If we disregard uncertainty, It can degenerate into a deterministic predicted value;

[0237] For the discount rate, this embodiment takes... (Or within the range of 3%–5%), the value can be based on the long-term government bond yield or the project's benchmark discount rate. This linkage causes the algorithm to favor more durable structures or reduce exposure intensity in humid, hot, acid rain, or freeze-thaw environments, thereby reducing total life-cycle cost and improving reliability.

[0238] Preferably, S400 further includes:

[0239] S410, Objective F1: Optimal (minimum) flood response speed.

[0240] The response rate is characterized by "excessive water level exposure + time to recede":

[0241] ;in:

[0242] The limit water level (m, determined by ground elevation + flood control standard or warning water level) for the key node (i) is written into the parameter table.

[0243] The time (h) required for the water levels of all critical nodes in the system to first fall below the safe water level from the peak moment, calculated from the discrete simulation output table.

[0244] The weight of the time term (dimensionless or converted to the same dimension). and It can be determined by the dimensional unification method: Let (typical order of magnitude) / (typical order of magnitude) (typical order of magnitude) / (typical order of magnitude), where and for The adjustment factor between these values ​​reflects the degree of emphasis on either the rate of receding water or the rate of rise. The default value can be set to [value missing]. , ,Right now typical typical , typical typical .

[0245] Discretization and Verifiable Calculation of S411 and F1 "Flood Response Speed" Indicators

[0246] To ensure repeatability during the review process, this invention discretizes the continuous-time integral index, allowing it to be directly calculated and verified from the simulation output table. Let the simulation step size be... Simulation time set:

[0247] ,in For any critical node The water level sequence is The limit water level is .

[0248] Definition of water level exceeding limit: ;

[0249] Define node overexposure (discrete integral): ;

[0250] Considering the importance of different nodes (such as schools, clinics, main street intersections, bridge and culvert entrances, etc.), set weights. (satisfy (or normalization can be used), then the overall over-limit exposure is:

[0251]

[0252] Receding water time Defined as: from the peak time Starting from this point, the water levels at all key nodes of the system have fallen back to safe levels for the first time. The following is the required time.

[0253] Discrete representation:

[0254] Since the peak time may not fall on discrete time intervals, we can approximate the peak time as follows:

[0255]

[0256] Then F1 discretization is: ;

[0257] To accommodate the risk of "rapid price increase," a "rate of increase penalty" can be added. A rapid rise in water level during the initial rise phase is considered unfavorable (for easier response and early warning). The average rate of rise is defined as follows:

[0258] ;in This set of moments during the initial rise phase can be determined from the start of rainfall to the period before the peak. Therefore: ;

[0259] F1 with starting penalty: ,in: The acceptable water level rise rate threshold (between 0.3 and 0.5 m / h) is determined by flood control response capabilities and road traffic requirements. This discrete form allows any reviewer or auditor to calculate F1 simply by obtaining the "Node Water Level Timing Output Table," and each threshold has a clear engineering meaning (limit water level, water level safety margin, rise rate threshold), ensuring repeatability.

[0260] in, The set of key nodes is determined by the facility / road / building database and the list of historical waterlogging points / bottleneck sections of S100. In this embodiment, 10 to 50 key nodes are selected for each village area.

[0261] For node limit water levels, it can be calculated according to... Calculation, where Obtained from DEM or measured elevation. Determined by road traffic / building flood protection standards, this embodiment uses 0.15 to 0.30m (road-type nodes) or values ​​according to building standards. To ensure a safe margin for water receding, this embodiment uses 0.05–0.10 m or is determined according to local flood control plans. Assigning weights to nodes based on their importance can be done hierarchically according to node type; if this is not possible, assign equal weights. ;

[0262] , The weighting factor (used to unify the dimensions of the time and water depth terms, its value can be obtained by regression analysis of the contribution ratio of the receding time to the total loss in historical events, with a preliminary suggested range of 0.05~0.2) can be fitted based on historical events or calibrated according to dimensional consistency.

[0263] This example uses the timeframe of the initial rise phase, where rainfall intensity exceeds a threshold. From the first moment to the peak moment The set of moments in between.

[0264] S420, Objective F2: Economic optimization of gray-green facility retrofit (minimization of net life-cycle cost).

[0265] Lifecycle cost (LCC) is defined as:

[0266]

[0267] in:

[0268] Construction cost: ,in:

[0269] This refers to the quantity of work (pipe length, pool volume, paved area, etc.). This is the comprehensive unit price. The construction difficulty coefficient.

[0270] Operation and maintenance costs (coupled with sediment degradation and operation and maintenance triggers):

[0271] ,in:

[0272] The number of times S370 is triggered is obtained. For energy consumption, This refers to the amount of plant maintenance.

[0273] Expected loss at risk: ,in:

[0274] The scenario set includes different rain types, traffic jams, and enhanced weather conditions. For scenario probabilities, Estimated from depth-loss curves and exposed assets.

[0275] Expected loss at risk: ,in:

[0276] A collection of scenarios, For scenario probabilities, Estimated from depth-loss curves and exposed assets.

[0277] Specifically, regarding asset types A piecewise linear loss rate function is used. :

[0278] ,in:

[0279] Characteristic water depths (e.g., 0.1m, 0.5m, 1.0m);

[0280] For the corresponding loss rates (e.g., 0, 0.3, 0.8), the values ​​can be referenced in Appendix A of the "Flood Disaster Loss Assessment Standard" (SL 579-2012);

[0281] Benefits (negative costs): Rainwater reuse, increased irrigation yields, improved road accessibility, etc. (converted to other benefits)

[0282] Structural repair costs From S384.

[0283] S421. Discounting, splitting, and uncertainty correction of net life cost (LCC)

[0284] F2 is expressed as the net present value over the entire life cycle, assuming a lifespan of 1200 years. In that year, the discount rate was (A percentage of 3% to 8% may be taken, and the basis for this percentage should be explained in the appendix.) Annual expenses Discounted to present value: .

[0285] Construction costs This usually occurs in year 0; to avoid confusion, it can be written as: ;

[0286] Operation and maintenance costs are broken down into: routine maintenance, dredging and obstacle removal, energy consumption, durability repair and emergency response. Annual maintenance costs: ;

[0287] Routine maintenance can be estimated linearly based on the facility size:

[0288] , The annual maintenance cost per unit area is (RMB / meter, RMB / m², RMB / unit·year).

[0289] The cost of dredging and clearing obstacles is determined by the number of triggers, let's say the number of triggers is 1. The number of dredging operations per year is: Cost of each dredging operation ,but: ;

[0290] Energy consumption cost: For equipment such as pumping stations, gates, and electrically controlled sand-draining systems, the annual energy consumption can be obtained by integrating the power consumed during operation. Discrete representation:

[0291]

[0292] but: ,in The electricity price (current electricity price can be used and a trend can be set).

[0293] Risk loss cost: scenario set Includes different rainfall types, climate enhancement, congestion chain events, etc., scenario probabilities satisfy Then the first Expected annual loss:

[0294]

[0295] Depth-loss curve: for asset types (Residential buildings, shops, roads, bridges, farmland, etc.), with a water depth of [missing information]. Maximum water depth .

[0296] The loss rate function is The asset value is ,but:

[0297] ,in:

[0298] Piecewise linear or S-shaped functions can be used. For piecewise linear functions, the relationship between the inflection point water depth d and the loss rate g can be determined by referring to typical curves in the "Flood Disaster Loss Assessment Standard" or similar publicly available literature. For residential buildings, the following settings can be used: loss rate 0 when water depth is 0m; loss rate 0.3 when water depth is 0.5m; loss rate 0.8 when water depth is 1.0m; and loss rate 1.0 when water depth is ≥1.5m.

[0299] Benefits (negative costs): These include rainwater reuse replacing tap water / pumping, increased irrigation yields, and tourism landscape enhancements. Let the... Annual reuse volume Unit revenue Increased production and income ,but: ;

[0300] Because "operational uncertainty" can lead to incomplete dredging and induce failures, this invention introduces a failure probability correction for operation and maintenance related items. Let the first... The annual failure probability of critical connection nodes is: (Calculated from the S300 degradation-recovery process), the risk loss can be corrected using the amplification factor:

[0301]

[0302] Meanwhile, maintenance costs will also increase due to sudden emergency response. Let the expected emergency cost be:

[0303]

[0304] The time required to further increase the carbonization depth to the thickness of the protective layer Considered as the starting point of corrosion:

[0305] ,when Afterwards, the corrosion rate takes effect and introduces repair costs.

[0306] The net present value over the entire life cycle is:

[0307]

[0308] If carbon price or carbon emission costs are considered, the carbon footprint can be converted into a cost item and added to F2. Let carbon emissions be... The carbon price is ,but:

[0309] And order:

[0310] ;

[0311] All the above-mentioned life-cycle cost items are measured in "yuan" and discounted to net present value based on year 0.

[0312] It occurred in year 0;

[0313] It is calculated from the facility type, scale, and geometric parameters (pipe length, pool volume, paved area, etc.) output by S600;

[0314] The prices are derived from cost quotas / bills of quantities or market inquiries and are standardized as comprehensive unit prices including tax.

[0315] To determine the construction difficulty coefficient, this embodiment uses segmented values ​​based on terrain slope and geologically sensitive areas (either through manual assignment or testing).

[0316] Maintenance trigger count The results were obtained from statistical analysis of the S370 triggering rules during lifetime simulation.

[0317] Equipment energy consumption Calculated from the power integral during the operating period within the year. The power rating is determined by the equipment's rated power and operating status (opening / closing / extraction / sand removal). The power integration time step (h);

[0318] Electricity price Use the local current electricity price or set a trend based on the annual growth rate. Scenario set. Scenario probabilities are constructed from combinations of different rainfall patterns, congestion levels, and climate enhancement coefficients. Determined by historical frequency, design rainstorm return period, or equal probability assumptions and satisfying ;

[0319] Asset value Sourced from asset ledgers or statistical data;

[0320] Loss rate function Piecewise linear or sigmoid functions are used, and breakpoint values ​​are obtained through manual assignment or testing;

[0321] Reuse Water balance output from S300;

[0322] Unit revenue Calculated using tap water price or pumping electricity price;

[0323] Increased production revenue Calculation, where For agricultural product prices, To increase production or to estimate based on empirical coefficients.

[0324] The probability of failure of critical connection node functions (0~1) is obtained from the S300 degradation-recovery process; The risk amplification factor (dimensionless) is given in this embodiment, and its default range can be determined by historical emergency events.

[0325] S430, Level-0 hard constraint system (one-vote veto)

[0326] Establish a set of hard constraints Violation will result in removal:

[0327] (1) Red line compliance: The area occupied is 0; ecological red line Hardened ash facilities are prohibited within the area; points will be deducted if the river-related area meets management requirements or triggers approval difficulties (hard constraints can be set as prohibited).

[0328] (2) Geological safety: landslide hazard area Internal anti-infiltration facilities or setting an upper limit on infiltration intensity; slope safety factor (For ordinary areas, e.g., 1.25; for important areas, e.g., 1.35). FS is calculated based on geological data input from S100, using either a simplified Bishop method or an infinite slope model. According to the "Technical Specification for Building Slope Engineering" (GB 50330) or the value determined by the geological disaster assessment report, 1.25 can be used for general areas and 1.35 can be used for important areas.

[0329] (3) Cultural and ancient tree avoidance: distance from facility boundary to ancient tree location Excavation or alteration of the landscape and visual corridors is prohibited within the protected cultural heritage area.

[0330] (4) Hydraulic safety: full flow duration at key nodes Interface shear stress .

[0331] (5) Historical disaster fingerprint constraint: For historical disaster sites, the simulated peak water level shall not exceed the historical flood mark minus the safety margin. This step ensures compliance, safety, and constructability through hard constraints.

[0332] , , All of these are spatially constrained areas, obtained from GIS layers such as permanent basic farmland, ecological red lines, geological disaster hazard areas, cultural heritage protection areas, and ancient and famous trees imported from S100; the area occupied by facilities is calculated as the "intersection area of ​​the facility boundary polygon and the constraint layer".

[0333] The slope safety factor FS is given by the slope stability analysis or geological hazard assessment report. The recommended value is 1.25 (general area) or 1.35 (important area) according to the specifications or reports. The slope safety factor FS should be provided by the accompanying geological survey report or calculated based on a simplified formula (such as an infinite slope model) and included as input data in the digital base dataset of S100.

[0334] Ancient tree avoidance radius Values ​​are taken according to the regulations for the protection of ancient and famous trees (5-10m in this embodiment). Full-flow duration at key nodes. The threshold value is obtained by determining the full flow state from the hydraulic simulation output and accumulating the results. Determined by drainage standards (10-30 min in this embodiment); interface shear stress according to (Or obtained through local structural calculations) It is determined by the material's allowable impact stress or empirical values ​​for the retaining wall.

[0335] Historical Traces From historical surveys / measurements, For safety margin (0.05-0.10m in this embodiment).

[0336] S440, Robust / Probabilistic Constraints (Data and Multi-Hazard, Climate Enhancement)

[0337] In scenarios involving data uncertainty, blockage / landslide chain events, and climate enhancement, opportunity constraints or CVaR constraints are employed:

[0338] Opportunity constraint example: ;

[0339] Example of CVaR constraint: ;

[0340] This step can improve the robustness of the solution under uncertain conditions.

[0341] In this embodiment, uncertainty is represented by a discrete scenario set S (combinations of different rain patterns / parameter perturbations / blockage levels / climate enhancement coefficients), and scenario probabilities. satisfy In this embodiment, uncertainty is represented by a discrete scenario set ( The scenario is characterized by a combination of rainfall pattern, parameter perturbations, congestion level, and climate enhancement coefficient, with a scenario probability ( ),and( In the absence of statistical data, this embodiment assumes equal probability. When frequency data is available, it is normalized according to rainfall pattern or historical frequency to obtain ( ).

[0342] Opportunity Constraints: Using scenario approximation:

[0343] ;in,

[0344] The risk level is set to 0.05 or 0.10 in this embodiment and written into the parameter table. The water accumulation threshold is (e.g., 0.15m or determined according to road traffic standards).

[0345] CVaR constraints on loss random variables (Calculated from the depth-loss curve and exposed assets under scenario(s)) CVaR is presented in a weighted discrete form:

[0346] ;in:

[0347] As a quantile auxiliary variable, it represents the estimated value at risk (VaR), i.e., the value corresponding to the loss distribution in the quantile distribution. (This embodiment takes) Loss threshold for quantiles;

[0348] The acceptable loss limit (set by reliability targets / investment constraints);

[0349] In a discrete scenario set The optimal value can be obtained by initializing the loss values ​​after sorting them and taking the corresponding quantile values, and then solving the problem through linear programming or iteration.

[0350] Preferably, S500 further includes:

[0351] S510, Feasibility Prioritization and Constraint Repair Strategies

[0352] This invention adopts a "feasibility-first" constraint processing strategy: when comparing any two schemes (A, B), if (A) satisfies all Level-0 hard constraints while (B) does not, then (A) takes priority; if both satisfy the hard constraints, then they are compared according to non-dominated sorting and crowding degree (NSGA-II) or reference point association distance (NSGA-III); if neither satisfies the hard constraints, then the one with the smaller degree of violation takes priority.

[0353] When the number of optimization objectives is 2 or 3, the NSGA-II algorithm is used; when the number of optimization objectives is greater than or equal to 4, the NSGA-III algorithm is used to better maintain the diversity of the solution set in the high-dimensional space.

[0354] Number of reference points for NSGA-III It is recommended to consider the population size Configure and ensure the number of reference points. Slightly smaller It is usually advisable Make closest to between.

[0355] The degree of violation is defined as the normalized overlimit sum of the hard constraints: ;in:

[0356] This is the set of Level-0 hard constraints;

[0357] The constraint function value for constraint (c) is uniformly written as ( ), exceeding the time limit ( );

[0358] The scale constant (and ( (Same dimensions, used for normalization), threshold constraints () When taking () ) or safety margin (such as ( ), ( ), ( ), distance / area categories are categorized by ( ) / unit area value; when there is no clear engineering scale, take () on the near boundary sample. ), and set a lower limit ( );

[0359] z represents the decision variable vector / encoding vector of the candidate schemes (including gray-green facility scale and layout parameters, and dynamic rule parameters). For any threshold constraint... , uniformly written as , making To meet constraints, This exceeds the limit;

[0360] Scale constant Taken from hard constraint parameters (such as water level). Shear stress is adopted Full flow duration adopts This is used to normalize and unify out-of-limit values ​​of different dimensions. A mildly injured individual is defined as... (In this embodiment, the value is 0.5), and deterministic repair is performed only on it to improve the proportion of effective solutions.

[0361] Space under repair The signed distance field is the boundary of the penalty area (negative for inside the penalty area, positive for the feasible area), along... The directional translation ensures the shortest path out of the restricted area, with the minimum translation distance being one grid cell side length. Repair examples (all are deterministic and reproducible rules):

[0362] (1) Space touches the red line: the distance from the field marked by the penalty area symbol The gradient direction is translated to outside the nearest feasible boundary, with the minimum translation distance. Take the side length of one grid cell (source: grid resolution);

[0363] (2) Geological infiltration restricted areas: Replace infiltration-type facilities with "infiltration-drainage combination (drainage as the main method) / seepage-proof lining" type, and update the permeability coefficient at the same time. Cost (Source: Type Library and Cost List);

[0364] (3) Excessive shear stress: Increase the number of energy dissipation stages. Or increase the size of the energy dissipation pool Alternatively, the lining grade may be upgraded.

[0365] S520, NSGA-III Reference Points and Diversity Preservation

[0366] When the target dimension is high, NSGA-III uses a pre-defined set of reference points. Maintain Pareto front coverage. Reference points are uniformly generated in the normalized target space, and individuals are associated with the nearest reference point during selection; for each reference point, individuals with the smallest distance to the reference line are preferentially retained to avoid the solution set being concentrated in a few regions.

[0367] To ensure repeatability, this embodiment uses Das–Dennis reference point generation: given the target dimension. With the number of partitions . For the number of reference points, It is the number of combinations; For the target dimension, For the number of partitions (in this embodiment, we take...) The number of partitions, p, can be set to 1 / 10 to 1 / 20 of the population size, ensuring that the number of reference points generated is no less than 1 / 2 of the population size.

[0368] The Euclidean distance between the reference point and the target vector is calculated in the dimensionless space after range normalization. Used to prevent the target range from being This results in division by zero. This invention suggests that when the target... At that time, Break it down into 2-3 sub-objectives (e.g., ecology (E), production (Pr), low carbon (LC)), and Together they form a 5-6 dimensional target space, which is then maintained by NSGA-III to maintain diversity, and finally the S700 classification evaluation selects the unique solution.

[0369] S530, Multi-fidelity / Agent Update Downtime Conditions and Security Backup

[0370] To avoid unsafe solutions being mistakenly selected due to proxy errors, this invention stipulates that any solution entering the "final candidate set" must undergo high-fidelity simulation verification (at least once for the design rainstorm and the climate-enhanced rainstorm, and at least once for the chain blockage scenario), and the final selection shall be based on the high-fidelity results.

[0371] The termination condition is either "hypervolute index convergence + proportional stability of feasible solutions" or reaching the maximum algebra.

[0372] : Maximum algebra (100–300 in this example);

[0373] Continuous decision window (10–30 generations in this embodiment);

[0374] : No. The hypervolume index (dimensionless) of feasible nondominated sets in the standardized target space.

[0375] : Overvolume enhancement threshold (dimensionless, in this embodiment) );

[0376] : Proportion of feasible solutions (number of individuals satisfying Level-0 / population size);

[0377] Feasibility stability threshold (0.02–0.05 in this example).

[0378] Stop when any of the following conditions are met:

[0379] (1) Reaching the maximum algebra ; or (2) continuous The substitution satisfies the following conditions: the overvolume increase is less than the threshold and the feasibility is stable.

[0380]

[0381]

[0382] If the proxy model is enabled, an average error threshold (as a safety net) is also set: for critical outputs (such as... or key objectives ) Calculate the mean absolute error (MAE) on the validation set / cross-validation, when A real simulation sample correction agent must be added at this time; this embodiment takes The water level is 0.05–0.10 m or 3%–5% of the index range. Any scheme entering the final candidate set will be verified using high-fidelity simulation results to avoid proxy errors introducing unsafe final selection. High-fidelity simulation refers to verification calculations using a finer time step (e.g., Δt ≤ 1 min), a more detailed hydrodynamic model (e.g., a complete solution of the Saint-Venant equations), or more complete coupled modules (e.g., detailed sediment transport) than the default simulation settings in the optimization iteration.

[0383] Super volume Based on the standardized target space calculation, this embodiment uses a reference point. (Each dimension is set to 1 or to an acceptable upper bound normalized constant) to ensure HV is verifiable. High-fidelity verification scenarios are selected from scenario set S: at least one design storm (in this embodiment, a storm with a return period of R years), one climate-enhanced storm (scaled up by an enhancement factor based on the design storm), and one chain-like congestion scenario (congestion level is medium to high). The surrogate model error (MAE) is calculated on the validation set, which is obtained by randomly partitioning the training samples or through K-fold cross-validation; when MAE exceeds a threshold... High-fidelity sample correction agents must be added at times, and the final selection shall be based on the high-fidelity results.

[0384] Preferably, S600 further includes:

[0385] S610. Calculate and classify the Comprehensive Risk Index (RI) to guide facility function classification and engineering template output. For any evaluation unit (grid unit) within the study area... or sub-water collection unit Define the comprehensive risk index:

[0386] ;in:

[0387] Comprehensive risk index (dimensionless), with a uniform value range in this embodiment. ;

[0388] Risk level index (dimensionless);

[0389] Vulnerability index (dimensionless);

[0390] Exposure index (dimensionless);

[0391] Historical disaster fingerprint indicators (dimensionless);

[0392] Four types of weights (dimensionless), satisfying the following constraints: The weights were determined by expert assignment (or AHP): 5–15 experts from fields such as water conservancy, municipal engineering, geological disaster prevention, and rural planning were organized to evaluate the relative importance of the four indicators. .

[0393] To ensure repeatability, All are mapped to according to a unified normalization rule. For any positive indicator (The larger the range, the higher the risk), using range normalization:

[0394] ;in:

[0395] Take the minimum / maximum value of all evaluation units in the study area (obtained from the statistical results of the calculation); To prevent division by zero constant (in this embodiment, we take...) ).

[0396] (1) Risk level The intensity of the disaster is obtained by normalizing the simulation output, which characterizes the disaster intensity based on factors such as water depth, flow velocity, and duration. This embodiment uses: ;in:

[0397] The maximum water depth (m). The maximum flow velocity (m / s) is taken from the maximum value of the S300 hydrodynamic simulation output sequence.

[0398] The duration of the overthreshold (h), for example The cumulative duration, The threshold for determining water accumulation is 0.15m in this embodiment or as determined by road traffic standards.

[0399] Internal weights (dimensionless) In this embodiment, the weights are equal (1 / 3 each).

[0400] (2) Vulnerability The result is obtained by normalizing slope, soil permeability, and geological sensitivity. This example uses: ;in:

[0401] The slope (° or %) is calculated from the DEM using the GISSlope tool.

[0402] The saturated hydraulic conductivity (mm / h or m / s, obtained by manual assignment or testing) is derived from soil databases, geotechnical tests, or field infiltration tests.

[0403] Geological sensitivity (dimensionless) is derived from the geological hazard zoning map / potential hazard points and buffer zones, and is assigned a value according to the classification table;

[0404] Internal weights (dimensionless) In this embodiment, the weights are set to equal (1 / 3 each).

[0405] (3) Exposure : Obtained by normalizing land use importance and population / asset exposure. This embodiment uses: ;in:

[0406] The land use exposure score (dimensionless) is obtained by looking up the land use type table (e.g., residential / school / hospital areas are high, commercial / main roads are high, general cultivated land is medium, and forest land is low; the specific score is obtained by manual assignment or testing). Source: remote sensing interpretation / current land use map / third national land survey data.

[0407] Population density (people / km² or people / grid) is derived from village records, census grids, or statistical data.

[0408] The asset value density (yuan / km² or yuan / grid) is derived from real estate / building ledgers, public facility lists, or estimated according to "land type - unit price".

[0409] Internal weights (dimensionless) In this embodiment, the weights are set to equal (1 / 3 each).

[0410] (4) Historical fingerprints This is obtained by normalizing historical disaster empirical information. This embodiment uses a "weighted density of historical disaster points" approach. ;in:

[0411] This is a collection of historical disaster sites, derived from historical flood / landslide / debris flow records, on-site investigations, and flood traces.

[0412] The distance is in planar dimension (m), calculated by GIS. The radius of influence (m) is 100m (or 50–200m) in this embodiment.

[0413] The disaster point weight (dimensionless) can be taken in this embodiment. (Based on frequency of occurrence only), or by disaster severity level Hierarchical assignment (assignments are obtained manually or through testing). Normalization symbols. This indicates that the above range normalization is mapped to... .

[0414] Finish After calculation, the areas are divided into high-risk (Red), medium-risk (Yellow), and low-risk (Green) categories based on threshold values, directly represented by color. This embodiment uses quantile thresholds (verifiable and applicable to different areas):

[0415] ;in For all evaluation units in the study area The 33% and 67% quantiles (obtained directly from the calculation results).

[0416] Based on risk level, facilities are classified by function as follows: Backbone defense type (high-risk area Red): mainly gray, with strong drainage / large storage / strong slope protection / sedimentation and energy dissipation; Peak shaving and buffer type (medium-risk area Yellow): gray and green co-location, focusing on cascade transmission, nodal sedimentation and moderate storage; Source reduction type (low-risk area Green): mainly green, with on-site disposal, infiltration and storage and resource reuse.

[0417] Template library: Define templates for each risk level. ( The template must include at least "a list of facilities combinations + a range of connection construction parameters + dynamic rule preferences". The range of connection construction parameters includes... The upper and lower bounds and default values, dynamic rule preferences include the default direction of thresholds / rule parameters (such as inhibiting infiltration during the high sediment load incubation period, prioritizing sediment discharge, etc., mapped to...). (Parameter range). This step is used to achieve reproducible engineering output, making it easy to quickly apply and optimize different villages.

[0418] Normalization , In this embodiment, all evaluation units are statistically analyzed based on the design baseline scenario (or scenario expectation value). If a multi-scenario evaluation is used, each evaluation unit is first weighted according to the scenario set S. Find the expected value, and then perform range normalization.

[0419] Unless otherwise specified, the internal weights of Haz are taken as follows: The internal weights of Vul and Exp are also equal. The larger the value, the stronger the infiltration capacity and the lower the risk of runoff and waterlogging; therefore, it is adopted. As a positive risk factor, this embodiment only counts historical fingerprints (His) in the calculation. Disaster points are identified to control computational load. (In the template library) These represent the effective volume of sediment, slope ratio / angle, grid spacing, weir height, outlet width, energy dissipation pool length, and number of sluice gates, respectively. (The parameters required for Haz calculations are...) (Maximum water depth) (Maximum flow velocity) is taken from the "Nodal Hydraulic Characteristic Statistics Table" output by step S300; (Duration exceeding threshold) The water depth data from the same source's "Node Water Depth Time Series Table" exceeds the threshold. The cumulative time is calculated.

[0420] S620, Mapping rules to the template

[0421] Based on the Pareto non-dominated feasible solution set P obtained from S500, and the risk partitions {Red, Yellow, Green} and facility function classifications defined in S630, the mapping is performed according to the following rules:

[0422] Template selection: For each solution Based on the spatial distribution of its facilities, the dominant risk zones of its impact are determined. (If more than 60% of the facility capacity is located in the Red Zone, then) =Red).

[0423] Parameter extraction and normalization:

[0424] Facility coordinates: to be solved The facility coordinates are aligned with the center of the nearest land use patch or planning grid and regularized to the standard construction positioning point.

[0425] Size and parameters: The solution Continuous scale parameters (such as pool volume) Round up to the nearest standard modulus (such as an integer multiple of 50m³).

[0426] Dynamic rules: solutions The regular function parameters in (e.g.) Based on its numerical range, it is mapped to a preset rule level (such as "high warning", "medium warning" and "low warning").

[0427] Template instantiation: Invocation and dominant risk partitioning Corresponding basic template Fill in the parameters that were normalized in step 2. The corresponding placeholders in the code form the instantiation template for the solution. .

[0428] Template library update: Update all generated instantiated templates Store in the template library. For new design projects, the template with the same risk partition and the closest key parameters can be directly matched from the library as the initial solution.

[0429] Preferably, S700, based on the differentiated layout template set generated by S600 and corresponding to the Pareto non-dominated feasible solution set, performs the following comprehensive evaluation and decision-making steps to select the unique recommended solution. To address the problem of too many Pareto solutions making decision-making difficult, this invention establishes a four-level target hierarchical system and introduces AHP-entropy weight coupling dynamic weights and implementation-oriented corrections to ensure that the final selected solution combines "compliance, security, economy, feasibility, manageability, and fairness".

[0430] S710, Four-Level Target Classification System

[0431] S711, Level-0: Hard constraints veto power

[0432] Establish Level-0 hard constraint set (Red line, geology, culture, critical hydraulics, construction and operation accessibility, historical flood marks, etc.). For any candidate scheme If any hard constraint is not satisfied, the algorithm is directly eliminated. The constraints are uniformly expressed as: for each... All require (Exceeding the limit is positive). Sources: (1) GIS / spatial constraints: red line overlay, occupied area, avoidance distance, etc., are calculated from spatial overlay or distance field; (2) hydraulic safety constraints: full flow duration, peak water level, shear stress, etc., are obtained by comparing the S300 simulation output with the threshold; (3) historical flood marks / fingerprint constraints: comparison of simulated water level at historical flood mark points / old disaster points with the measured threshold (from S100). Threshold sources are as follows: Water level related thresholds: determined according to the flood control standards specified in the "Technical Specification for Urban Waterlogging Prevention and Control", or using the warning water level issued by the local flood control command department. Geological safety thresholds According to the "Technical Specification for Building Slope Engineering". Spatial constraint threshold: based on the vector boundaries of permanent basic farmland, ecological protection red lines, etc., as delineated by the natural resources department. Construction capacity threshold: such as... The size is determined based on the maximum operating width of the selected construction machinery or the commercially available standard size of the pipe.

[0433] S712, Level-1: Core objectives take precedence (F1 response speed, F2 total lifecycle cost).

[0434] Level-1 with core objectives As the primary factor, a weighting of 50% to 70% is recommended. This applies to the set of feasible solutions that pass Level-0. For indicators that are "the smaller the better", a unified normalization is applied (converting them into scores that are "the larger the better"). ;in:

[0435] Normalized score (dimensionless, [0,1]);

[0436] To prevent division by zero constant (in this embodiment, we take ( ));

[0437] This is the set of feasible solutions that pass the Level-0 hard constraints. The calculation is derived from the S421 discretization formula and the simulated water level time series output table; The net present value (NPV) of the entire life cycle of S422 (including discounting, operation and maintenance, and risk breakdown) is calculated from the cost and simulation output table.

[0438] Level-1 score (linear weighting in this example; TOPSIS can also be used, but this example requires determining one): ;in:

[0439] Core weights (dimensionless) satisfy... The weights are derived from the S720 coupling weights; this embodiment can also provide default values ​​(such as...). When it is safest, it can be taken Note: The values ​​of w1 and w2 are derived from the AHP-entropy weight coupling weight of S730; in this embodiment, the default value w1=w2=0.5 can also be set (or w1>w2 for a safer approach).

[0440] S713, Level-2: Bonus points for important objectives

[0441] Level-2 is a key bonus point for objectives (recommended 20%–30%), including maintenance frequency, robustness loss, durability margin, resource utilization rate, industry synergy, and microclimate benefits. For each indicator… Unified and normalized to (The larger the better), and adopt a threshold trigger + linear scoring method:

[0442] ;in:

[0443] Level-2 indicator set (each) (The definitions, units, and calculation methods are obtained through manual assignment or testing).

[0444] Normalization rule: If The original meaning was "the bigger the better": ;like The original idea was "the smaller the better": ; Pick Internal statistics (verifiable);

[0445] The trigger threshold (dimensionless) is implemented in this embodiment using one of the following two methods: mapping a specification / engineering threshold to... The value after; or take The quantile (e.g., 0.5 quantile) is used as the threshold;

[0446] : Bonus weight (dimensionless), satisfying Source: S720 or the default table in this embodiment.

[0447] In this embodiment, Level-2 includes at least the following verifiable indicators: number of dredging triggers per year. (Times / Year, Source: S370 Statistics), Robust Loss Indicators (Unit: Yuan, Source: S440 Discrete Scenario Calculation), Durability Margin (mm, source S381). Original thresholds for each indicator (e.g.) , , First, map the image to [0,1] according to the corresponding normalization rule to obtain the result. Then the bonus points will be triggered again.

[0448] S714, Level-3: Optimal Selection of Auxiliary Indicators

[0449] Level-3 indicators, such as landscape harmony, construction disturbance, and noise and dust, are handled through a point deduction system. Each deduction indicator... (The larger the value, the worse) First normalize to Deductions:

[0450] ;in: Level-3 indicator set; Deduction weight (dimensionless), satisfying Or it may be given according to the proportions set in this embodiment.

[0451] S720, Comprehensive Basic Score

[0452] For schemes that pass Level-0, the overall basic score is defined as follows:

[0453] ;in:

[0454] Hierarchical weights (dimensionless) satisfy the following conditions: Recommended scope: .

[0455] S730, AHP-Entropy Weight Coupled Dynamic Weight

[0456] For the set of indicators that need to be weighted (which can be used for Level-1) or Level-2 Subjective weights are obtained from AHP, and objective weights are obtained from the entropy weight method, and they are coupled:

[0457] ;in:

[0458] Final weight (dimensionless);

[0459] The subjective weights of the Alternative Hierarchical Method (AHP) are dimensionless. The determination process for AHP subjective weights is as follows: Experts in water conservancy, municipal engineering, and economics are invited to conduct pairwise comparisons of indicators at the same level. A judgment matrix is ​​formed using the 1-9 scaling method. After a consistency test (CR < 0.1), the weights of each indicator are calculated. The objective weights of the entropy (dimensionless) are determined from the candidate scheme set. The dispersion is obtained by calculating the index value matrix; Coupling coefficient (dimensionless), in this embodiment, is taken as... Alternatively, a value of 0.6 to 0.7 can be used when consistency is higher. To avoid weight distortion caused by different village types, range constraints can be set for the weights of key indicators, and normalization can be performed after truncation. .

[0460] Entropy weight method calculation The index value matrix within is used as input. First, each index is positively oriented and normalized according to the rule of "the larger the better / the smaller the better". Then, information entropy and dispersion are calculated to obtain... In this embodiment, the following settings are used: Level-1 uses S730 coupling weights for F1 and F2; Level-2 uses internal weights. The entropy weight method (or equal weight) is adopted to avoid the impact of arbitrary changes in weights on repeatability.

[0461] Application Example 1 (High-risk village in a canyon)

[0462] (1) Scenarios and Data (corresponding to S100)

[0463] Choose canyon-type village A:

[0464] The valley is narrow, with a steep gradient and high sediment load. Historical flood marks are concentrated at the main street entrance and bridge / culvert inlets. Data was collected and constructed using a 0.5m resolution DEM and slope raster. Valley lines / river centerlines, road and building vectors, existing pipeline networks and node elevations; red line constraint masks (ecological red lines / river-related management areas / permanent basic farmland); historical disaster fingerprint point sets. The rainfall scenario set S includes:

[0465] ①Designed for a rainstorm (e.g., a 5-year return period, lasting 1 hour);

[0466] ② Climate-enhanced heavy rainfall (rainfall intensity is amplified by a correction factor for the same duration);

[0467] ③ Chain blockage scenario (high sand content + increased grid blockage rate).

[0468] Simulated step size min, set of key nodes Six key locations were selected, including main street intersections, bridge and culvert entrances, and schools / health clinics, and their water level limits were provided. (Determined by ground elevation + flood control standards).

[0469] (2) Decision variables (corresponding to S200)

[0470] static variables This includes: the diameter, slope, and direction of the main drainage pipes, and the volume of the storage tank. Effective area of ​​sedimentation basin With volume Energy dissipation structural series Outlet width wait;

[0471] Dynamic rule variables Includes: Drainage flow allocation function for critical nodes (Distributed to the two main channels / bypasses in segments according to water level thresholds), target water level sequence for regulation and storage facilities. (Pre-flood / During-flood / Post-flood), Sediment Discharge / Gate Opening / Closing Thresholds and Rule Functions (Prioritize sediment discharge during the initial rise and prioritize receding water levels after the peak).

[0472] The boundaries of each variable are given by the redline mask, construction accessibility, and construction library (e.g., m³, ).

[0473] (3) Mechanism model and lifetime evolution (corresponding to S300)

[0474] Coupled with sediment deposition in hydrodynamic calculations:

[0475] Cumulative silt volume satisfy ,in:

[0476] Output from simulated cross-sectional flow rate Given by sampling or inversion, The density of the sediment;

[0477] , The equivalent settlement velocity is used. For nodes with a grid structure, the clogging rate is introduced. And correct the overcurrent capability Degradation-recovery process: when or Trigger dredging, press Update, in which Single-cycle cleaning capacity (given by equipment capacity and operation duration). Lifespan. Annual discount rate .

[0478] (4) Calculation of objectives and constraints (corresponding to S400)

[0479] F1 uses the "excessive exposure + evaporation time" method:

[0480] For each , , , (weight) Normalized by node importance).

[0481] Receding water time From peak time From the beginning, the first satisfaction of .Pick m, will Converted to m·h units, then F2 based on Net Present Value (LCC): ,in Triggered by dredging number of times Energy consumption With vegetation maintenance calculate, This reflects the amplification of failure.

[0482] Level-0 hard constraints: Zero occupancy of the red line, compliance of the river-involving area, and full flow duration at key nodes. Shear stress wait;

[0483] Robust constraints: Apply chance constraints or CVaR verification to chain blockage and climate enhancement scenarios.

[0484] (5) Solving and final selection (corresponding to S500-S700)

[0485] Using joint coding Target dimension Using NSGA-III: population size 120, 200 generations; reference points generated and partitioned according to the Das–Dennis method. Each generation calls upon S300 for evaluation, and hard constraints are evaluated using a "feasibility priority + violation degree" approach. "Strategy. The proposed solutions that enter the final candidate set must undergo high-fidelity verification (designed rainstorm + climate-enhanced rainstorm + chain blockage at least once each).

[0486] The S700 adopts a tiered evaluation system: Level-0 is a veto, Level-1 is based on F1 / F2, Level-2 adds points, and Level-3 deducts points. The Level-1 weighting uses AHP-entropy weight coupling. ).

[0487] (6) Output schemes and results (corresponding to S600-S700)

[0488] Recommended scheme: The gray infrastructure consists of "main drainage ditch + 2 sedimentation ponds + 1 regulating reservoir (3200m³) + 2 levels of energy dissipation structure"; the green infrastructure includes shallow ditches and ecological retention zones in medium-risk areas and permeable paving in low-risk areas; the dynamic rule is "rainfall intensity exceeding the threshold during the initial rising period". Prioritize sediment removal, and after the peak, allow sediment to recede according to key milestones. Triggers accelerated drainage, automatically reducing the infiltration rate and increasing the priority of strong drainage / sediment removal in high sediment-laden scenarios.

[0489] Results (verified by S300 output): Design under heavy rain conditions h, m·h, m·h; Net present value over lifespan It costs tens of thousands of yuan and meets all hard constraints under scenarios of climate enhancement and chain congestion.

[0490] Application Example 2 (Terrace Agriculture Village)

[0491] (1) Scenarios and Data (corresponding to S100)

[0492] Selected terraced agricultural village B: moderate slope, high proportion of arable land, strong constraints from the basic farmland red line; significant irrigation demand. Data collection: 1m resolution DEM and runoff accumulation. ,soil (On-site infiltration test or soil pool), land use / planting structure, population and asset ledger, red line cover (basic farmland must be completely occupied). Scenario set S includes: designed heavy rainfall and typical rainfall patterns in wet / dry years; vegetation phenology through... Correct the effective roughness and permeability parameters of green facilities.

[0493] (2) Decision variables (corresponding to S200)

[0494] static variables Including: Rain garden area Length of infiltration trench Permeable pavement area Distributed storage tank volume Key nodes and pipe diameters of the reused pipeline network; dynamic rule variables, etc. Includes the target water level sequence for the two stages of the "flood season - irrigation season". Priority rules for reuse (priority replenishment after rain, priority reuse during dry season). Variable boundaries are determined by the land use boundary and the upper limit of construction disturbance (e.g., ㎡, m³).

[0495] (3) Model, Objectives and Constraints (corresponding to S300-S400)

[0496] In S300, infiltration-retention-overflow coupling calculations are used for green facilities, and the effective infiltration permeability coefficient is... Characterizes the impact of blockage / water content, and is updated by S370 maintenance trigger rules (dredging / repair / replacement). .

[0497] F1 is reviewed using the same discretization method as in Application Example 1; F2, in addition to including Capex and maintenance costs, explicitly includes the reuse revenue. The net present value is used for discounting. Level-0 hard constraints ensure zero occupation of basic farmland and compliance with river-related regulations.

[0498] (4) Solving and final selection (corresponding to S500-S700)

[0499] The objective dimension consists of four dimensions: F1, F2, resource utilization rate, and robust loss. NSGA-III (population 120, iterations 150 generations) is used. The final selection also undergoes high-fidelity verification. The village type in S700 is "agricultural". The weight range of the "reuse / resource utilization rate" indicator in Level-2 is constrained and normalized.

[0500] (5) Output schemes and results (corresponding to S600-S700)

[0501] Recommended scheme: 560m infiltration trench, 1800㎡ rain garden, 2200㎡ permeable pavement; 600m³ decentralized storage tanks with valve-controlled reuse regulations; annual reuse volume... m³ is used for irrigation as an alternative to water intake.

[0502] Result: Designed for heavy rain h, m·h, m·h; Net present value over lifespan RMB 10,000 (already included in maintenance triggers, risk losses and reuse revenue deductions).

[0503] Comparative Example 1 (corresponding to Application Example 1: static scale optimization only, fixed control rules, no dynamic decision variables introduced)

[0504] Aside from the differences, the S100 data, scenario set, hard constraint thresholds, and solution budget are consistent with Application Example 1. The difference is that runtime rules are not parameterized as optimizable variables (not using...). , The threshold rule function only optimizes static pipe / pond volume and uses fixed gate and sediment discharge rules; it cannot adaptively switch sediment discharge priority strategy during the high sediment concentration rise period.

[0505] Result (verification): Design for heavy rain h, m·h, m·h; Net present value over lifespan Tens of thousands of yuan; under the scenario of "climate enhancement + chain congestion", the full flow duration of some key nodes is close to / exceeds the threshold, and the robust feasibility boundary may require an increase in the scale of gray facilities afterward.

[0506] Comparative Example 2 (corresponding to Application Example 2: F2 only considers initial investment / constant maintenance, and does not include lifetime evolution and revenue items)

[0507] Except for the differences, it is the same as application example 2. Difference: F2 only uses... Alternatively, a "constant maintenance" approach could be used, excluding maintenance fluctuations and increased failure probability risks caused by dredging, as well as losses from reuse revenue deductions. The proposed solution tends to reduce the size of water storage and reuse facilities, resulting in lower short-term investment but deteriorating flood control and resource recovery performance.

[0508] Result (verification): Design for heavy rain h, m·h, m·h; if calculated based on the actual lifespan (including maintenance triggers, risks and benefits), the equivalent net present value F2 rises to approximately RMB4.75 million and the amount of reuse is insufficient.

[0509] Application examples and comparative results:

[0510]

[0511] Compare using the table above:

[0512] (1) Compared with Comparative Example 1: This invention elevates the operational control (flow allocation, target water level sequence, threshold rules) to an optimizable dynamic decision variable, enabling the same set of hardware to adaptively switch control strategies under different rainfall / sediment conditions. This is reflected in the significant reduction of receding water time and excessive exposure, and remains feasible under uncertain scenarios (enhanced climate + blockage).

[0513] (2) Comparative Example 2: This invention evaluates F2 using the net present value over the entire life cycle, explicitly accounting for operational fluctuations triggered by siltation and degradation, amplified risk losses from failure probability, and deductions from reuse revenue. This prevents optimization from being misled by "minimizing short-term investment," thereby achieving a synergistic result of lower net cost and better flood control performance. The above two sets of comparisons show that this invention, through the collaborative design of "optimizable dynamic rules + life cycle evolution / revenue inclusion + hard constraints / robust verification + multi-objective solution and final selection mechanism," produces a comprehensive technical effect that is quantifiable, verifiable, and implementable, supporting its inventiveness.

Claims

1. A multi-objective optimization design method for gray-green coupled facilities, characterized in that, The process includes the following steps: S100, Construction and correction of the mountainous rural environmental perception and digital foundation dataset, collecting and correcting multi-source data, including red line constraint data; constructing a digital foundation dataset for mountainous rural stormwater management; establishing a set of regional parameter correction parameters or correction functions, and introducing historical disaster fingerprints for regional parameter correction model calibration and consistency verification; S200, Determination and quantitative representation of dynamic decision variables, quantifying the control rule parameters during operation into optimizable decision variables in the design phase, and providing value boundaries, constraints, and calibration standards for each decision variable; S300, Mechanism model and full life cycle dynamic evolution modeling of the gray-green coupling system, establishing a parameterized model of gray-green facilities and connecting components, outputting hydraulic response, risk indicators, operation and maintenance burden, and life-cycle performance degradation trajectory under different scenarios and control rules, and forming a verifiable time-series output table or statistical output table; S400, Construction of a multi-objective optimization model, constructing a multi-objective optimization model based on the output table of S300, establishing a Level-0 hard constraint system, and providing the constraint function judgment standard; It allows updating the target and constraint calibrators after calibrating the model parameters based on newly added monitoring data; S500, solving and obtaining Pareto non-dominated feasible solution sets, designing constraint handling strategies for Level-0 hard constraints and probabilistic constraints, and controlling convergence and diversity through reference points or congestion mechanisms; outputting Pareto non-dominated feasible solution sets that satisfy the constraints; S600, generating facility classification and differentiated layout templates based on risk results, calculating the comprehensive risk index RI based on the risk-related output of S300 and classifying it, and partitioning the region; classifying gray and green facilities by function. It can classify; and map the "coordinate-scale-connection construction-dynamic rules" of Pareto non-dominated feasible solutions into differentiated layout templates, and match or generate gray-green facility layout and connection construction schemes that meet risk level requirements and red line constraints from the template library; establish a hierarchical evaluation system for S700, target classification and scheme comprehensive evaluation decision; adopt AHP-entropy weight coupling dynamic weight and set weight interval constraints in combination with village type, first positiveize each indicator according to "the larger the better / the smaller the better" and normalize it to [0,1], and then calculate information entropy and dispersion to obtain objective weights; The difficulty of approval / implementation is used as a feasibility adjustment factor to select a unique recommended solution from the Pareto non-dominated feasible solution set. Specifically, the digital base dataset, red line constraints, and regional adjustment results output by S100 are used to define the domain, boundary conditions, and calibration benchmarks of the decision variables in S200. The dynamic decision variables and static construction parameters output by S200 are used as inputs to the S300 mechanistic model to obtain hydraulic response, risk indicators, and life-cycle degradation trajectory, which are then fed back to S400 to construct the objective function and constraint judgment criteria. Under the objectives and constraints defined by S400, S500 iteratively searches and repeatedly calls S300 to complete the scheme evaluation and feasibility judgment to obtain the Pareto non-dominated feasible solution set. S600 calculates RI based on the solution set and completes risk zoning, facility classification, and template mapping. S700 selects the unique recommended solution from the solution set based on hierarchical evaluation, coupling weights, and feasibility adjustments.

2. The multi-objective optimization design method for gray-green coupling facilities according to claim 1, characterized in that, S100 also includes: S110, Multi-source heterogeneous data acquisition and cleaning: acquiring and constructing a full-element database. Data cleaning employs filtering and consistency checks; S120, regional parameter correction model: local parameters : Among them: rainfall correction factor ： Soil / permeability correction factor ： Temperature / Freeze-Thaw Correction Factor ： Vegetation / phenology correction factor ： S130, Historical Disaster Fingerprint Calibration: Collection of historical flood data points Old Disaster Collection Destroyed by history And select a set of representative historical rainstorm events. As a calibration sample, the objective function is calibrated based on the "error of the highest water level at critical nodes": The mean absolute error threshold is used as a constraint. ; where the threshold Determined by a combination of observation and measurement errors and topographic elevation errors: 。 3. The multi-objective optimization design method for gray-green coupling facilities according to claim 1, characterized in that, S200 also includes: S210, drainage flow distribution variables. Functional representation: Constructing a composite Sigmoid function: ； S220, Dynamic water storage capacity / target water level Quantification: For the water storage facility (k), define the target water level sequence. Or available storage capacity As a dynamic decision variable, the "forecast level - segmented ladder" function is adopted: S230, Dynamic Variable Set and Joint Encoding, Final Dynamic Decision Variable Set: 。 4. The multi-objective optimization design method for gray-green coupling facilities according to claim 1, characterized in that, S300 also includes: S310, a hydrodynamic and runoff model, which uses a distributed or semi-distributed "rainfall-runoff-catchment" model to divide the watershed into sub-catchment units. and with a unified time step Discrete calculations are performed; S320 and green connection components are parameterized, and the gray-green connection interface is regarded as a set of key nodes. For each connection node, define an optimizable geometric parameter: sediment volume. Tangential inflow angle Grid spacing , dam height Overflow outlet width Length of energy dissipation pool Flood level The hydraulic capacity of the connecting nodes is expressed as: S330, Sediment Module and Turbidity Inversion: When there are high sediment loads or debris flow gully mouths in the watershed, the high sediment load module is activated to establish the relationship between turbidity and sediment load. S340, sludge degradation model, defines the available overcurrent capacity of connected nodes. With the cumulative silt volume attenuation: S350, scour loss model and structural failure risk, for interface, drop, energy dissipation pool construction, calculate bed shear stress: S360, vegetation phenology and hydraulic response, seasonal variations in the roughness and infiltration capacity of green facilities, and the definition of Manning roughness as a function of leaf area index (LAI): S370, Operation and Maintenance Recovery Model, Dredging and Recovery: ； S380, Structural durability and aging dimension: Incorporating key components of gray facilities into durability aging modeling; S381, Carbonization depth model: S382, Steel Corrosion Rate Model: When Achieve protective layer thickness Or corrosion is initiated when chloride ions reach a threshold: S383, Load-bearing degradation and failure probability: ； 。 5. The multi-objective optimization design method for gray-green coupling facilities according to claim 1, characterized in that, S400 also includes: S410, Target F1: Optimal flood response speed; response speed is characterized by "excessive water level exposure + receding time". S420, Objective F2: Economic optimization of gray-green facility transformation; Define life cycle cost (LCC) as: ； Construction cost: Operation and maintenance costs: Expected loss at risk: Benefits: S430, Level-0 hard constraint system: Establishing a set of hard constraints Violations will result in removal; S440, robust / probabilistic constraints, and chance constraints or CVaR constraints are used in scenarios of data uncertainty, blockage / landslide chain events, and climate enhancement.

6. The multi-objective optimization design method for gray-green coupling facilities according to claim 5, characterized in that, S410 also includes: discretization and verifiable calculation of the "flood response speed" index S411 and F1; assuming the simulation step size is... Simulation time set ,in For any key node The water level sequence is The limit water level is ; Define water level exceeding limit: Define node exposure limits: Considering the importance of different nodes, set weights. The overall overexposure level is: ; time for water to recede Defined as: from the peak time Starting from this point, the water levels at all key nodes of the system have fallen back to safe levels for the first time. The following time requirements are expressed discretely as follows: Since the peak time may not fall on discrete time intervals, we can approximate it by taking the peak time as an approximation: Then F1 discretization is: Define the average rate of ascent: ;in The set of moments during the initial rise phase can be determined from the start of rainfall to the period before the peak. Therefore: F1 with initial price increase penalty: 。 7. The multi-objective optimization design method for gray-green coupling facilities according to claim 5, characterized in that, S420 also includes: S421, discounting, splitting, and uncertainty correction of the net life cost (LCC); F2 is expressed as the net present value (NPV) over the entire lifespan, assuming a lifespan of 120 years. In that year, the discount rate was , the Annual expenses Discounted to Construction costs It occurred in year 0: The operation and maintenance costs are: Routine maintenance is estimated linearly based on facility size: ;Dredging and obstacle removal costs: The cost of dredging and clearing obstacles is determined by the number of times the event is triggered. The number of dredging operations per year is: Cost of each dredging operation Discrete representation: ;but: Risk loss cost: Scenario set Includes different rainfall types, climate enhancement, and congestion chain events, with scenario probabilities. satisfy Then the first Expected annual loss: Depth-loss curve: Maximum water depth ， For asset type, , where is the water depth; the loss rate function is The asset value is ,but: ;Benefit item: Let the first Annual reuse volume Unit revenue Increased production and income ,but: Risk losses are adjusted using a magnification factor: ； For the first The probability of functional failure of critical connection nodes in a given year; meanwhile, maintenance costs will also increase due to unforeseen emergencies, with the expected emergency cost set as follows: Further time required to achieve the carbonization depth to the thickness of the protective layer. Considered as the starting point of corrosion: ;when Afterwards, the corrosion rate takes effect and repair costs are introduced; then the net present value over the entire life cycle is: The carbon footprint is converted into a cost item and added to F2: And order: 。 8. The multi-objective optimization design method for gray-green coupling facilities according to claim 1, characterized in that, S500 also includes: S510, Feasibility Priority Ranking and Constraint Repair Strategy: A "feasibility priority" constraint handling strategy is adopted: When comparing any two options (A, B), if (A) satisfies all Level-0 hard constraints while (B) does not, then (A) takes precedence; if both satisfy hard constraints, then they are compared according to non-dominated ranking and crowding degree (NSGA-II) or reference point association distance (NSGA-III); if neither satisfies hard constraints, then the one with the smaller degree of violation takes precedence; the degree of violation is defined as the normalized overlimit sum of the hard constraints. S520, NSGA-III Reference Points and Diversity Preservation: When the target dimension is high, NSGA-III uses a pre-set set of reference points. Maintain Pareto front coverage; reference points are uniformly generated in the normalized target space, and individuals are associated with the nearest reference point during selection; for each reference point, individuals with the smallest distance to the reference line are preferentially retained, using Das–Dennis reference point generation: given the target dimension. With the number of partitions Number of reference points: (i.e., from) Nakadori The number of combinations); S530, shutdown conditions and safety net for multi-fidelity / agent updates: any scheme that enters the "final candidate set" must be verified by high-fidelity simulation, and the final selection shall be based on the high-fidelity result; the termination condition adopts "convergence of hypervolume index and stability of feasible solution proportion" or reaching the maximum algebra.

9. The multi-objective optimization design method for gray-green coupling facilities according to claim 1, characterized in that, S600 also includes: calculating and classifying the Comprehensive Risk Index (RI) to guide facility function classification and engineering template output; defining the Comprehensive Risk Index for any evaluation unit within the study area: ； For any positive indicator Range normalization is used: Among them: (1) Risk level : Obtained by normalizing the simulation output characterizing the intensity of the disaster: (2) Vulnerability The result is obtained by normalizing the slope, soil permeability, and geological sensitivity. (3) Exposure : Obtained by normalizing land use importance and population / asset exposure: (4) Historical fingerprints : Obtained by normalization of historical disaster empirical information: 。 10. The multi-objective optimization design method for gray-green coupling facilities according to claim 1, characterized in that, S700 is used to select a unique engineering solution from the Pareto feasible solution set. S700 also includes: S710, a four-level objective classification system: S711, Level-0: hard constraints veto; establishing a set of Level-0 hard constraints. For any candidate scheme If any hard constraint is not satisfied, the system is directly eliminated; the constraints are uniformly expressed as: for each All require S712, Level-1: Core Objectives Dominant: Level-1 is driven by core objectives. As the primary factor, the set of feasible solutions for Level-0 is considered. For indicators that are "the smaller the better", a unified normalization is adopted: Level-1 score: S713, Level-2: Bonus points for key objectives; Level-2 is a bonus point item for key objectives, for each indicator. Unified and normalized to It employs threshold triggering and linear scoring: S714, Level-3: Optimal selection of auxiliary indicators; Level-3 indicators are processed using a deduction method, with each deducted indicator... First normalize to Deductions: S720, Overall Basic Score: For solutions that pass Level-0, the Overall Basic Score is defined as follows: S730, AHP-Entropy Weight Coupled Dynamic Weights: For the set of indicators requiring weighting, subjective weights are obtained from AHP, and objective weights are obtained from the entropy weight method, and are coupled as follows: 。

Images