A cross-regional flood control and waterlogging elimination coupling simulation method

Abstract

The application relates to the technical field of flood control simulation, and discloses a cross-region flood control and waterlogging elimination coupling simulation method. The application guarantees the uniformity and reliability of input data by obtaining standardized input data sets. Further, the cross-region water dynamic interaction influence simulation is carried out, and a correlation matrix is calculated, the water force influence degree and propagation time between regions are quantified, the space-time correlation between regions is established, and the decision-making island mode is broken. Further, the asynchronous game model is used to optimize the strategy and generate a time sequence, the physical propagation law can be integrated into the decision-making, the decision-making conflict caused by the traditional synchronous assumption is solved, the change from local optimization to global coordination of the flood control and waterlogging elimination decision-making is realized, and the decision-making efficiency and scientificity of the whole basin are improved.

Description

Technical Field

[0001] This invention relates to the field of flood control simulation technology, and specifically to a cross-regional flood control and drainage coupled simulation method. Background Technology

[0002] With the acceleration of urbanization and the frequent occurrence of extreme weather events, regional rainstorm and flood disasters are characterized by high frequency of occurrence, wide impact range, and enhanced coupling. Especially in complex river systems that span multiple basins or administrative regions, significant hydrodynamic correlations exist between upstream and downstream areas, main streams and tributaries, and different drainage units, making traditional flood control and drainage scheduling methods that focus on a single region ineffective in addressing the needs of multi-regional coordinated prevention and control. Therefore, how to construct simulation and decision-making methods that can reflect the hydrodynamic coupling relationships between regions with the support of multi-source data has become an important research direction in the field of flood control and disaster reduction.

[0003] To address the aforementioned issues, existing research primarily focuses on hydrodynamic numerical simulation, multi-source data fusion, and scheduling optimization methods. In hydrodynamic simulation, river network hydrodynamic models based on the Saint-Venant equations are widely used to numerically solve the flood evolution process. In data processing, the accuracy of input information is improved by fusing real-time monitoring data with meteorological forecast data. In scheduling optimization, optimization algorithms or game theory models are introduced to solve drainage strategies under multiple objectives and constraints. Furthermore, some studies have begun to explore multi-scale grids, adaptive computation, and uncertainty analysis to improve simulation accuracy and computational efficiency.

[0004] However, existing technologies still face several deep-seated technical bottlenecks in achieving truly cross-regional, highly dynamic, and refined collaborative scheduling, particularly in the deep coupling of accurate simulation of physical processes and intelligent group decision-making. These problems are manifested in the distortion of physical causal timelines, the contradiction between simulation computational efficiency and accuracy, and the limitations in setting decision-making objectives. Specifically, these include: the traditional Saint-Venant equation assumes that river channels are independent, neglects the pulse-like disturbances of sluice gate and pump drainage, the time-varying delay in the impact of upstream drainage on downstream areas, and the simplification of the superposition effect when multiple polder areas simultaneously drain water. Summary of the Invention

[0005] This invention provides a cross-regional flood control and drainage coupled simulation method to address the significant shortcomings of existing technologies in achieving truly cross-regional, highly dynamic, and refined collaborative scheduling in terms of accurate simulation of physical processes and deep coupling of intelligent group decision-making.

[0006] In a first aspect, the present invention provides a cross-regional flood control and drainage coupled simulation method, the method comprising: Obtain a standardized input dataset containing real-time hydrological monitoring data and meteorological forecast data; based on the standardized input dataset and a preset river network topology model, conduct cross-regional hydrodynamic interaction simulations and calculate the regional interaction intensity distribution matrix and flood wave propagation delay matrix; based on the regional interaction intensity distribution matrix and flood wave propagation delay matrix, use a time-delay-based asynchronous game model to optimize strategies and generate a real-time target drainage strategy time series.

[0007] The cross-regional flood control and drainage coupled simulation method provided by this invention ensures the uniformity and reliability of input data by acquiring a standardized input dataset. Furthermore, by conducting cross-regional hydrodynamic interaction simulations and calculating correlation matrices, the degree of hydraulic influence and propagation time between regions are quantified, establishing spatiotemporal correlations between regions and breaking the decision-making silo model. Moreover, by utilizing an asynchronous game model to optimize strategies and generate time series, the physical propagation laws can be integrated into decision-making, resolving decision conflicts caused by traditional synchronous assumptions, and realizing the transformation of flood control and drainage decision-making from local optimization to global coordination, thereby improving the overall decision-making efficiency and scientific rigor of the watershed.

[0008] In one optional implementation, a standardized input dataset containing real-time hydrological monitoring data and meteorological forecast data is obtained, including: Acquire real-time hydrological monitoring datasets and meteorological forecast datasets; determine forecast error index values ​​based on the real-time hydrological monitoring datasets and meteorological forecast datasets; calculate dynamic fusion weight values ​​that characterize the reliability of forecast data using the forecast error index values; perform weighted fusion of the real-time hydrological monitoring datasets and meteorological forecast datasets based on the dynamic fusion weights to obtain a fused rainfall input sequence; generate a standardized input dataset based on the fused rainfall input sequence.

[0009] The cross-regional flood control and drainage coupled simulation method provided by this invention can assess the reliability of forecast data in real time by determining the forecast error index value. Furthermore, by calculating dynamic fusion weight values, it can adaptively match forecast quality with data weights, improving the rationality of data fusion. Moreover, by generating a fused rainfall input sequence through weighted fusion, it combines the advantages of both measured and forecasted data, reducing data uncertainty and improving the reliability of rainfall data.

[0010] In one optional implementation, based on a standardized input dataset and a pre-defined river network topology model, a cross-regional hydrodynamic interaction simulation is performed, and the regional interaction intensity distribution matrix and flood wave propagation delay matrix are calculated, including: Obtain the impulse response function matrix; based on the impulse response function matrix, generate a three-layer asynchronous adaptive grid containing a global coarse grid, an influence zone fine grid, and an impulse zone ultrafine grid; in the three-layer asynchronous adaptive grid, perform hydrodynamic calculations using a standardized input dataset, and obtain calculation results at multiple levels; prioritize and fuse the calculation results at multiple levels, and generate a multi-temporal and spatial resolution water level field; based on the multi-temporal and spatial resolution water level field and a preset river network topology model, determine the regional interaction influence intensity distribution matrix and the flood wave propagation delay matrix.

[0011] The cross-regional flood control and drainage coupled simulation method provided by this invention can quantify the spatiotemporal interaction of regional drainage behavior by obtaining the impulse response function matrix, thus providing a basis for grid generation. Furthermore, by combining the impulse response function matrix to generate a three-layer asynchronous adaptive grid, computational resources are allocated on demand, balancing simulation accuracy and efficiency. Furthermore, by using a standardized input dataset to perform hydrodynamic calculations within the three-layer asynchronous adaptive grid and obtaining multi-level results, the hydraulic characteristics of different regions are accurately captured, providing multi-resolution data for subsequent fusion. Furthermore, by fusing and generating multi-spatial-temporal resolution water level fields, high resolution is ensured in key areas while maintaining overall computational efficiency. Furthermore, by determining the regional interaction intensity distribution matrix and the flood wave propagation delay matrix, the intensity of inter-regional influence and propagation delay are clarified, providing physical constraints and information input for the asynchronous game model.

[0012] In one optional implementation, obtaining the impulse response function matrix includes: By adding an impulse source term to the standard Saint-Venant hydrodynamic equations, an impulse-embedded hydrodynamic equations are obtained. The impulse source term is used to simulate the instantaneous start-up behavior of the pumping station. The Green's function method is used to solve the impulse-embedded hydrodynamic equations and generate the impulse response function matrix.

[0013] The cross-regional flood control and drainage coupled simulation method provided by this invention, by constructing a pulse-embedded hydrodynamic equation system, can more realistically simulate the hydraulic effects of instantaneous pump station activation, thus improving the physical fidelity of the model. Furthermore, by solving the pulse-embedded hydrodynamic equation system using the Green's function method and generating an impulse response function matrix, a physical law library of spatiotemporal interactive influences across the entire river network is established, providing core parameters for subsequent simulations.

[0014] In an optional implementation, the method further includes: modifying the impulse response function matrix using a nonlinear correction function, wherein the nonlinear correction function is used to reflect the nonlinear effects of the actual cross-sectional morphology of the river channel.

[0015] The cross-regional flood control and drainage coupled simulation method provided by this invention corrects the impulse response function matrix by using a nonlinear correction function, which makes up for the neglect of the nonlinear effects of the actual river channel by the linear model and improves the physical reality and accuracy of the matrix.

[0016] In one optional implementation, based on a multi-temporal resolution water level field and a preset river network topology model, the regional interaction influence intensity distribution matrix and flood wave propagation delay matrix are determined, including: Based on the multi-temporal and spatial resolution water level field, a real-time water level dataset is obtained. A time-varying propagation velocity field is generated based on the real-time water level dataset and a pre-defined river cross-section parameter library. This time-varying propagation velocity field reflects real-time hydrological conditions and river geometry. A decision-oriented directed graph is generated based on a pre-defined river network topology model. The flood wave propagation delay matrix uses the decision region as nodes and the direction of water flow influence as directed edges. Using the reciprocal of the time-varying propagation velocity field as path weights, the decision-oriented directed graph is processed using a shortest path search algorithm to obtain a set of target propagation paths. A flood wave propagation delay matrix is ​​constructed based on this set of target propagation paths. The impulse response function matrix and the historical target drainage strategy time series are convolved and integrated to determine the regional interaction intensity distribution matrix.

[0017] The cross-regional flood control and drainage coupled simulation method provided by this invention extracts multi-temporal and spatial resolution water level information by acquiring real-time water level datasets, providing basic data for wave velocity calculation. Furthermore, by combining a pre-set river cross-section parameter library to generate a time-varying propagation velocity field, the fixed wave velocity assumption is abandoned, making wave velocity estimation more consistent with real-time hydrological conditions and river morphology, thus improving the accuracy of propagation delay calculation. Furthermore, by generating a directed decision graph, the river network topology can be transformed into a decision-related structure, clarifying the regional decision-making logic. Furthermore, by processing the target propagation path set through a shortest path search algorithm, the fastest flood propagation channel can be accurately identified, avoiding delay errors caused by misjudgment of propagation paths. Furthermore, by constructing a flood wave propagation delay matrix, the dynamic connectivity of the river network is comprehensively reflected, providing a core basis for asynchronous decision-making timing. Furthermore, by determining the regional interaction influence intensity distribution matrix through convolution integrals, the time-varying cumulative impact of drainage behavior is quantified, providing support for the weight configuration of the benefit function.

[0018] In one optional implementation, based on the regional interaction influence intensity distribution matrix and the flood wave propagation delay matrix, a time-delay-based asynchronous game model is used for strategy optimization, and a real-time target drainage strategy time series is generated, including: Using the flood wave propagation delay matrix, an asynchronous decision-making time series is determined. This asynchronous decision-making time series is used to ensure that the upstream decision-making influence of each decision-making area can physically propagate to the downstream before downstream decision-making is carried out. A time-varying payoff function set is obtained. Based on the asynchronous decision-making time series, with the time-varying payoff function set as the optimization objective, the Nash equilibrium is solved iteratively to generate a real-time target drainage strategy time series.

[0019] The cross-regional flood control and drainage coupled simulation method provided by this invention, by determining asynchronous decision-making time series, can embed physical propagation delays into game rules, avoiding decision conflicts caused by information asynchrony and ensuring the scientific nature of downstream decisions. Furthermore, by obtaining a time-varying payoff function set, a decision objective that considers current safety, historical impact, and future responsibility is constructed, which helps guide regional collaborative cooperation. Moreover, by iteratively solving for Nash equilibrium, the optimal strategy for global coordination can be found and a real-time target drainage strategy time series can be generated, improving the overall flood control and drainage efficiency of the basin.

[0020] In one optional implementation, obtaining the time-varying revenue function set includes: The system calculates the current item reflecting the safety status of each decision-making area; calculates the upstream historical impact item using the flood wave propagation delay matrix; determines the future expected responsibility item based on the impulse response function matrix; and dynamically configures time-varying weight coefficients for the current item, upstream historical impact item, and future expected responsibility item according to the regional interaction influence intensity distribution matrix, and generates a time-varying benefit function set.

[0021] The cross-regional flood control and drainage coupled simulation method provided by this invention quantifies the region's own flood control safety status by calculating the current term, ensuring the responsiveness of decisions to immediate risks. Furthermore, by calculating the upstream historical impact term, it clarifies the region's responsibility for upstream flood risks, avoiding the neglect of historical cumulative impacts in decision-making. Furthermore, by determining the expected future responsibility term, it quantifies the potential impact of current decisions on downstream areas, suppressing self-serving behavior that harms neighbors. Furthermore, by configuring time-varying weight coefficients and generating a function set, it enables the decision objective to adapt to different operating conditions, balancing the importance of various impact terms and improving the flexibility and scientific nature of decision-making.

[0022] In one alternative implementation, the method further includes: Using the real-time target drainage strategy time series as the historical target drainage strategy time series, a cross-regional hydrodynamic interaction simulation is performed again, and a result water level field is generated. The result water level field and the preset physical constraint set are checked, and any violations are identified in the result water level field. If any violations are found, the real-time target drainage strategy time series is iteratively corrected using a penalty function until no violations are found, and the real-time target drainage strategy time series is determined.

[0023] The cross-regional flood control and drainage coupled simulation method provided by this invention can verify the physical feasibility of a strategy by resimulating and generating the resulting water level field, thereby validating the strategy's adaptability to actual hydraulic environments. Furthermore, by comparing against physical constraints to identify violations, strategy risks are investigated, ensuring the engineering safety of the strategy. Moreover, by iteratively correcting the strategy through a penalty function, violation issues are eliminated, thus transforming the mathematically optimal solution into an engineering-feasible solution, improving the strategy's reliability and robustness.

[0024] In one alternative implementation, the method further includes: Obtain the target multi-temporal resolution water level field corresponding to the real-time target drainage strategy time series; calculate multiple discrete index values ​​based on the regional interaction influence intensity distribution matrix, the real-time target drainage strategy time series and the target multi-temporal resolution water level field; and obtain the overall synergy score based on the multiple discrete index values ​​through a weighted comprehensive evaluation model.

[0025] The cross-regional flood control and drainage coupled simulation method provided by this invention acquires the target's multi-temporal resolution water level field and extracts complete water level information after strategy implementation, providing data support for evaluation. Furthermore, by calculating multiple discrete index values, the strategy performance can be quantified from dimensions such as safety, efficiency, and coordination, comprehensively evaluating the strategy's effectiveness. Moreover, by calculating the overall coordination score through a weighted comprehensive evaluation model, the overall coordination level of the strategy can be comprehensively reflected, providing a quantitative reference for decision optimization.

[0026] In one alternative implementation, the method further includes: Based on the real-time target drainage strategy time series and the actual operating status of the pumping station units, an optimization problem that satisfies engineering constraints is constructed; based on the optimization problem, a mixed integer programming model is constructed; the mixed integer programming model is solved, and a pumping station control command sequence is generated, which is used to control the operation of the pumping station.

[0027] The cross-regional flood control and drainage coupled simulation method provided by this invention, by constructing an optimization problem that satisfies engineering constraints, can transform theoretical strategies into engineering requirements that fit the actual operation of pumping stations, thus bridging the gap between theory and practice. Furthermore, by constructing a mixed-integer programming model, it can accurately characterize the operational constraints and objectives of pumping station units, providing mathematical model support for instruction generation. Moreover, by generating a sequence of pumping station control instructions, the optimal strategy is transformed into machine-executable operation instructions, realizing the engineering implementation of the strategy.

[0028] In an optional implementation, the method further includes: generating a flood control and drainage impact propagation path map based on the regional interaction influence intensity distribution matrix, through force-directed layout algorithm and visual element encoding processing.

[0029] The cross-regional flood control and drainage coupled simulation method provided by this invention can intuitively display the direction, intensity and time delay of the impact between regions by generating a flood control and drainage impact propagation path map, which helps to improve decision-making transparency and provides a visualization tool for cross-regional collaborative communication. Attached Figure Description

[0030] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0031] Figure 1 This is a schematic diagram of an application scenario according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating the cross-regional flood control and drainage coupled simulation method according to an embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0034] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0035] As an optional application scenario of this invention, the specific application environment architecture or specific hardware architecture on which the cross-regional flood control and drainage coupled simulation method depends is described here. For example... Figure 1 As shown, the architecture system may include at least one terminal device and at least one server. Figure 1The system is illustrated in the example, which includes a computer 101, a mobile terminal 102, and a server 103, and the terminal devices such as the computer 101 and the mobile terminal 102 are connected to the server 103 through a network 110.

[0036] Specifically, the terminal device can be a smartphone, tablet, laptop, PDA, desktop computer, game console, smart TV, smart wearable device, in-vehicle terminal, VR (Virtual Reality) device, AR (Augmented Reality) device, etc. Server 103 can be a standalone physical server, a server cluster, a distributed system, or a cloud server providing cloud services. Network 110 can be a wired or wireless network, examples of which include, but are not limited to, the Internet, corporate intranet, local area network, wide area network, mobile communication network, and combinations thereof.

[0037] Existing collaborative scheduling models generally ignore the time-delayed propagation of hydraulic impacts in the physical world, leading to a break in the causal logic chain of decision-making. Most current optimization models often employ synchronous or quasi-steady-state assumptions during their construction, assuming that within a single decision step, the information of all decision-makers is synchronized, and that the scheduling behavior of one region will instantly manifest in other regions. This flattening of information severely violates the physical law that flood waves require a defined time to propagate in river networks. The actual impact of a drainage decision in an upstream region may take several hours or even longer to reach the downstream. If downstream regions make decisions based on outdated information synchronized with the upstream, they may make unnecessary conservative operations before the impact truly arrives (such as prematurely activating pumping stations and wasting energy), or experience conflicts due to insufficient preparation when the impact arrives (such as the superposition of upstream flood peaks and downstream local flood peaks), triggering beggar-thy-neighbor negative effects.

[0038] Furthermore, traditional hydrodynamic simulation methods inherently present an irreconcilable contradiction between computational accuracy and efficiency, failing to meet the real-time simulation requirements of highly dynamic and localized events. The start-up and shutdown of pumping stations and gates generate shock waves or discontinuities with rapid propagation speeds and dramatic gradients in their vicinity. Accurately capturing these phenomena requires extremely high-resolution computational grids (such as meter-level or even smaller). Using a uniform fine grid across an entire vast watershed or urban area would result in astronomical computational demands, completely failing to meet the real-time or near-real-time requirements of flood control scheduling. Conversely, using a uniform coarse grid to ensure computational efficiency would completely obliterate these crucial local hydraulic details due to numerical dissipation, leading to distorted simulation results and consequently, unreliable scheduling decisions.

[0039] Furthermore, the objective functions (or benefit functions) in existing optimization models are overly simplistic, making it difficult to quantify and weigh the complex power and responsibility relationships in cross-regional decision-making. Most models' objective functions are limited to the decision-maker's own immediate interests, such as minimizing the highest control water level within the region. This approach fails to formally answer complex questions such as: How much cost should I bear for the flood risks already generated upstream and about to reach me? And how much responsibility will my current drainage actions bear for potential future risks downstream? Due to the lack of quantitative expression regarding historical inheritance and anticipated future responsibilities, the decision-making models remain inherently self-interested, unable to guide parties towards genuine collaboration, and their optimization results are often localized and short-sighted, failing to maximize the overall interests of the watershed.

[0040] In summary, existing technologies have several shortcomings: First, most models focus on single-region or weakly coupled analysis, making it difficult to accurately characterize the cross-regional hydrodynamic interaction and its propagation delay characteristics; second, the dynamic characterization of forecast errors and data reliability is insufficient during multi-source data fusion, affecting the reliability of simulation results; third, in terms of scheduling decisions, synchronous or static optimization methods are mostly used, lacking consideration of asynchronous decision-making mechanisms under the constraint of water flow propagation delay; fourth, existing methods lack executability under engineering constraints and lack the ability to quantitatively evaluate and visualize the scheduling coordination effect.

[0041] Based on this, it is necessary to develop a flood control and drainage simulation method that can integrate multi-source data, characterize cross-regional hydrodynamic coupling relationships, and combine propagation delay to achieve asynchronous optimization decision-making. This will improve the scientific, real-time, and collaborative nature of drainage scheduling under complex water system conditions, thereby providing more reliable technical support for cross-regional flood control and disaster reduction.

[0042] According to an embodiment of the present invention, a cross-regional flood control and drainage coupled simulation method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0043] This embodiment provides a cross-regional flood control and drainage coupled simulation method, which can be used on the aforementioned mobile terminals, such as mobile phones and tablets. Figure 2 This is a flowchart of a cross-regional flood control and drainage coupled simulation method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Obtain a standardized input dataset containing real-time hydrological monitoring data and meteorological forecast data.

[0044] In one alternative embodiment, the standardized input dataset represents a preprocessed, uniformly formatted, and quality-controlled dataset that can be directly used as input for subsequent hydrodynamic simulation models.

[0045] In one alternative embodiment, raw data can be collected from various sources (such as telemetry terminals of local hydrological bureaus and meteorological bureaus), such as hourly rainfall observation data, water level data of key river sections at 15-minute intervals, historical operation logs of pumping stations, and gridded numerical weather forecast rainfall data for the next 24 to 72 hours.

[0046] Furthermore, after data acquisition, the acquired multi-source heterogeneous data is aligned with timestamps and standardized with coordinate systems. Methods such as Kriging interpolation are used to interpolate and correct missing or outlier values ​​in the data, thereby forming a spatiotemporally continuous and uniformly formatted dataset.

[0047] Step S202: Based on the standardized input dataset and the preset river network topology model, a cross-regional hydrodynamic interaction simulation is performed, and the regional interaction intensity distribution matrix and flood wave propagation delay matrix are calculated.

[0048] In one optional embodiment, the preset river network topology model is a digital representation of the river network structure, which usually exists in the form of a directed graph. The nodes in the graph represent river intersections, pumping stations, key sections, etc., while the directed edges represent the physical connections and flow directions of the water flow.

[0049] In one alternative embodiment, the regional interaction intensity distribution matrix is ​​used to quantify the cumulative impact of drainage behavior in any region on other regions.

[0050] For example, the regional interaction influence intensity distribution matrix is ​​a spatiotemporal four-dimensional scalar field describing the global influence, denoted as... Its physical meaning is located at the source point. The unit's drainage behavior, for those located at the receiving point In the future The extent of the impact on water levels.

[0051] In an optional embodiment, the flood wave propagation delay matrix is ​​an M×M matrix (M being the number of decision regions), whose elements... Indicates that the flood wave originates from the region Spread to the region The shortest time required.

[0052] In one alternative embodiment, the evolution of floods in the river network is simulated by solving hydrodynamic differential equations (such as the Saint-Venant equations) that describe the movement of water flow, using a river network topology model (directed graph structure) as the spatial and logical basis and a standardized input dataset as the driving and constraint conditions.

[0053] For example, a hydrodynamic propulsion algorithm based on impulse response function and three-layer asynchronous adaptive grid can be used for simulation.

[0054] Meanwhile, the river network topology model can be used to construct a time-varying directed graph, solve for the shortest path, and finally quantify the cumulative impact of regional drainage behavior and generate a regional interaction influence intensity distribution matrix, as well as quantify the shortest time of flood propagation and generate a flood wave propagation delay matrix.

[0055] Furthermore, through cross-regional hydrodynamic interaction simulation, the flood control and drainage decision-making problem between different regions can be transformed from an isolated, mutually independent model into an organic whole based on physical laws and spatiotemporal correlation.

[0056] Furthermore, the regional interaction intensity distribution obtained in this step is a preliminary distribution calculated based on standard test conditions. It is used to characterize the inherent hydraulic propagation characteristics of the river network and can provide physical constraints and information input for the subsequent construction of asynchronous game models.

[0057] In one optional embodiment, when performing cross-regional hydrodynamic interaction simulations, the obtained standardized input dataset is used as the driving term, initial condition, and boundary condition of the model. Specifically, this may include: fused rainfall input sequences as spatiotemporal rainfall forcing; measured water levels at key cross-sections and upstream / downstream boundary flows as initial / boundary fields; and pump station operation logs and facility parameters as local impulse source terms and control inputs. This input data is also used for the calibration and verification of the impulse response function and the nonlinear correction function.

[0058] Furthermore, the river network topology model serves both as an index table for numerical discretization and physical parameters in simulations, and as the graph structure basis for propagation paths and decision-making time series. Specifically, the river network topology model can be used to: (1) map the spatial relationship between the computational grid and the decision region and provide hydraulic parameters such as cross-section and roughness; (2) construct a directed graph with reciprocal velocity as the weight under a time-varying propagation velocity field to find the shortest propagation path and calculate the time delay matrix; (3) locate the pulse source and response node of the pump station to solve the impulse response function matrix; (4) divide the adaptive grid according to the influence intensity index derived from the impulse response and establish a cross-layer nested boundary mapping; (5) abstract it into a decision-making directed graph to generate asynchronous decision-making time series.

[0059] Step S203: Based on the regional interaction influence intensity distribution matrix and the flood wave propagation delay matrix, the strategy is optimized using a time-delay-based asynchronous game model, and a real-time target drainage strategy time series is generated.

[0060] In one alternative embodiment, the time-delay-based asynchronous game model is a special game model whose core feature is that the decision-makers (i.e., each flood control area) do not act simultaneously, but follow an asynchronous decision sequence with a sequence determined by the propagation delay of the flood wave.

[0061] In one optional embodiment, the target drainage strategy time series is the optimal drainage strategy time series, which is a time-indexed set of decision vectors, for example... ,in Each element of the vector Representative area exist The optimal drainage flow rate that should be executed at any given time.

[0062] In one optional embodiment, a time-varying payoff function is first established for each decision-making region, reflecting its own security, its assumption of historical impacts from upstream, and its responsibility for future impacts from downstream. Then, following an asynchronous decision-making sequence determined by the propagation delay, each region aims to maximize its own payoff function and determines its optimal strategy by iteratively solving for Nash equilibrium, thus forming a corresponding real-time target drainage strategy time series.

[0063] Furthermore, by introducing an asynchronous game framework based on physical time delay, the problem of beggar-thy-neighbor or decision-making conflicts that are common in cross-regional flood control and drainage is fundamentally solved. This allows downstream regions to fully wait for and take into account the actual impact of upstream decisions, thereby generating a coordinated and efficient optimal strategy sequence on a global spatiotemporal scale, which ultimately helps to improve the overall flood control and drainage benefits of the entire basin.

[0064] The cross-regional flood control and drainage coupled simulation method provided in this embodiment ensures the uniformity and reliability of input data by acquiring a standardized input dataset. Furthermore, by conducting cross-regional hydrodynamic interaction simulations and calculating correlation matrices, the degree of hydraulic influence and propagation time between regions are quantified, establishing spatiotemporal correlations between regions and breaking down decision-making silos. Moreover, by utilizing an asynchronous game model to optimize strategies and generate time series, the physical propagation laws can be integrated into decision-making, resolving decision conflicts caused by traditional synchronous assumptions. This achieves a shift in flood control and drainage decision-making from local optimization to global coordination, improving the overall decision-making efficiency and scientific rigor of the watershed.

[0065] In some optional implementations, step S201 above includes: Step S2011: Obtain the real-time hydrological monitoring dataset and the meteorological forecast dataset.

[0066] In one optional embodiment, the real-time hydrological monitoring dataset represents a collection of rainfall data actually measured by equipment such as ground rain gauges, for example... .

[0067] In one alternative embodiment, the weather forecast dataset represents a collection of future rainfall forecast data generated by a numerical weather prediction model, such as... .

[0068] Step S2012: Determine the forecast error index value based on the real-time hydrological monitoring dataset and the meteorological forecast dataset.

[0069] In one optional embodiment, by comparing measured and forecast data from historical periods and calculating the degree of deviation between forecast and measured data, the reliability of the current forecast can be clearly determined and the recent performance of the forecast model can be quantified.

[0070] In one alternative embodiment, a retrospective historical time window can be set, for example, the most recent 6 hours. Subsequently, the measured rainfall sequence within these 6 hours can be extracted, and the forecast rainfall sequence for the same time period can be found from historical forecast data. Furthermore, by comparing these two sequences, an error index, i.e., the forecast error index value, can be calculated to quantify the performance of the recent forecast model.

[0071] For example, the forecast error index can be represented by the root mean square error (RMSE), as shown in the following equation (1): (1) In the formula: Indicates time The corresponding root mean square error of the forecast characterizes the degree of bias in the recent forecast model; Represents discrete moments within a historical time window; Indicates the historical time window Real-time weather forecasts and rainfall data; Indicates the historical time window Real-time hydrological monitoring (measured) rainfall data at any given moment; This indicates the number of sample points within the window.

[0072] Furthermore, since the accuracy of numerical weather prediction models fluctuates with changes in meteorological conditions, this embodiment calculates a short-term, dynamically rolling forecast error index to assess the reliability of forecast data in the current meteorological context in real time.

[0073] Step S2013: Calculate the dynamic fusion weight value that characterizes the reliability of the forecast data using the forecast error index value.

[0074] In an optional embodiment, the dynamic fusion weight value is a dimensionless coefficient that varies over time. Its value ranges from 0 to 1 and is used to allocate the proportion of measured data and forecast data in subsequent weighted fusion.

[0075] In an optional embodiment, the weight values ​​are dynamically fused. The calculation method should be aligned with the recent forecast error index. They are inversely proportional; that is, the larger the error, the lower the reliability of the forecast data, and the smaller the weight it should be assigned.

[0076] For example, the dynamic fusion weight value is calculated using the following relation (2). : (2) In the formula: This represents the weights assigned to the measured data, i.e., the dynamic fusion weight values; Represents the natural exponential function; A sensitivity coefficient representing a positive value is used to adjust the degree to which the weight changes drastically with error, and its value can be calibrated using historical data.

[0077] Furthermore, when When it approaches 0 (the forecast is extremely accurate), Approaching 0.5 (the formula here is designed as a shift of the sigmoid function; a more reasonable expression would be...) Approaching 0, for example .in, This represents the baseline error. (At this point, the forecast data weights...) Larger.

[0078] Furthermore, when When it becomes very large, When the value approaches 1, the weight of the measured data is at its maximum.

[0079] Furthermore, the above mechanism allows the weight allocation to dynamically adapt to changes in forecast quality.

[0080] Step S2014: Based on dynamic fusion weights, the real-time hydrological monitoring dataset and the meteorological forecast dataset are weighted and fused to obtain the fused rainfall input sequence.

[0081] In one optional embodiment, by dynamically weighting the measured and forecast data, the accuracy of the measured data and the foresight of the forecast data can be taken into account, thereby generating a more reliable and higher-quality rainfall sequence, i.e., a fused rainfall input sequence.

[0082] In an alternative embodiment, the fused rainfall input sequence is calculated using the following relation (3): (3) In the formula: Indicates time The fused rainfall input sequence.

[0083] Furthermore, regarding history and the present moment, If it exists, the above relation (3) can be directly applied; for future moments, If it does not exist, its value can be considered as 0 or not involved in the calculation. Mainly composed of The decision, but its weight They have already been punished or motivated by their recent historical performance.

[0084] Furthermore, through the above process, a high-quality fused rainfall input sequence can be generated that includes both past realities and future trend predictions, with the ratio of the two dynamically adjusted according to the forecast reliability.

[0085] Step S2015: Generate a standardized input dataset based on the fused rainfall input sequence.

[0086] In one optional embodiment, the obtained fused rainfall input sequence is combined with other hydrological data (such as upstream and downstream boundary conditions) to form the final standardized input dataset, which can significantly improve the input data foundation of the entire simulation method, thereby improving the accuracy of subsequent hydrodynamic simulation and game decision-making.

[0087] In an alternative embodiment, the flow process at the upstream control section of the watershed can be screened from the post-quality control hydrological dataset. and the water level process at the downstream tidal gauge station By using cubic spline interpolation, irregular time interval data are unified into 15-minute intervals, generating upstream inflow boundary sequences and downstream tidal level boundary sequences.

[0088] Furthermore, multiple ensemble members (typically 20-50) of the numerical weather prediction rainfall data are read, and the rainfall forecast mean for each grid point is calculated. and standard deviation A probabilistic rainfall forecast field is constructed by combining the Bayesian method with historical forecast error statistics. This includes 50%, 75%, and 95% confidence intervals.

[0089] Furthermore, based on the error analysis of the measured rainfall and the corresponding time period forecast in the hydrological dataset after the quality control of the last 6 hours, the time-varying fusion weight is calculated using the above relation (2), and the probabilistic rainfall forecast field and the measured data are weighted and fused using the above relation (3) to generate the fused rainfall input sequence, which together with the boundary sequence constitutes the complete input dataset.

[0090] In some optional implementations, step S202 above includes: Step S2021: Obtain the impulse response function matrix.

[0091] In one optional embodiment, the impulse response function matrix is ​​an N×N matrix that quantifies the spatiotemporal impact of drainage behavior in any region (pumping station) of the entire river network on other regions, and is used to quantify the spatiotemporal interaction between drainage behaviors in different regions. Here, the matrix elements represent the hydraulic response triggered by a unit drainage impulse from a certain pumping station in other regions; N represents the number of computational nodes in the river network.

[0092] Specifically, step S2021 above includes: Step a1: Add a pulse source term to the standard Saint-Venant hydrodynamic equations to obtain the pulse-embedded hydrodynamic equations.

[0093] In an alternative embodiment, the pulse source term is used to simulate the instantaneous start-up behavior of the pumping station.

[0094] In an alternative embodiment, by embedding a pulse source term simulating the instantaneous activation of a pumping station into the Saint-Venant hydrodynamic equations describing water flow motion, the local operation of the pumping station can be transformed into an excitation term that can be calculated by the hydrodynamic model, thereby enabling the quantification of its impact on the global hydraulic state.

[0095] For example, in the standard Saint-Venant equations Based on this, a pump station pulse source term is introduced. This yields the corresponding pulse-embedded hydrodynamic equations.

[0096] in, Indicates the cross-sectional area of ​​the water passage. Rate of change over time; Indicates flow rate Length along the river channel The rate of change; This indicates lateral inflow (the amount of lateral inflow per unit length of the river channel). Dirac The function represents a function located in space at the pumping station. Location The instantaneous effect; Indicates pumping station The drainage flow process, i.e., the pumping station Drainage flow rate over time The process of change; This represents a time envelope function used to describe the actual physical process of the pump station's response after startup, rather than an idealized instantaneous step. Its preferred form is: , For Heaviside step function, This is a characteristic response time determined based on the characteristics of the pumping station and the river channel.

[0097] Furthermore, the improved source term representation method described above can more realistically reflect the smooth transition characteristics of pump station startup in the time dimension, thereby improving the physical fidelity of the model.

[0098] Step a2: Solve the impulse embedded hydrodynamic equations using the Green's function method and generate the impulse response function matrix.

[0099] In one optional embodiment, the Green's function method represents a mathematical method for quantifying the spatiotemporal response of local disturbances (such as pumping station drainage) to the global system (river network) by solving linear partial differential equations under unit impulse excitation. It is mainly used for river sections with small disturbances and slow cross-sectional changes. In local areas such as pumping station outlets, the model needs to be modified.

[0100] In one optional embodiment, the linearized pulse embedding hydrodynamic equations are solved by the Green's function method to obtain the hydraulic response solution under unit pulse excitation. Then, a nonlinear correction function is combined to adapt to the actual river characteristics, and finally, the regional interaction influence quantification matrix, i.e., the pulse response function matrix, which reflects the real river characteristics, can be obtained.

[0101] In an alternative embodiment, the pulse-embedded hydrodynamic equations are linearized to obtain a basic linear Green's function kernel.

[0102] For example, its fundamental solution can be obtained through mathematical methods such as Fourier-Laplace transform, as shown in the following relation (4): (4) In the formula: This represents the linear Green's function solution (hydraulic response under unit impulse excitation). Indicates the location of the point of influence; Indicates the moment of impact; Indicates the location of the source point (i.e., the pumping station); Indicates the time when the pulse occurs; Indicates the water diffusion coefficient; This indicates the reference wave velocity.

[0103] Furthermore, the impulse response function matrix can be further determined using the aforementioned fundamental solution.

[0104] In some optional implementations, step S2021 above further includes: Step a3: Correct the impulse response function matrix using a nonlinear correction function.

[0105] In one optional embodiment, the nonlinear correction function is used to reflect the nonlinear effect of the actual cross-sectional shape of the river channel, thereby correcting the deviation between the linear Green's function solution and the actual cross-sectional shape of the river channel (nonlinear flow effect) and improving the physical reality of the impulse response function.

[0106] In an alternative embodiment, due to the above linear solution The method is derived under the idealized assumptions of a rectangular, wide, and shallow river channel. However, the actual cross-sectional shape of a river channel is complex and variable, and the water flow exhibits significant nonlinear effects. Therefore, this embodiment introduces a correction function that reflects the nonlinear effects of the actual cross-sectional shape of the river channel.

[0107] For example, based on pre-stored historical flood data and a river cross-section parameter database, a model corresponding to water depth is calibrated and generated through numerical simulation or data fitting. Related nonlinear correction function The following relation (5) is shown: (5) In the formula: Indicates real-time water depth; Indicates the reference water depth; and These represent dimensionless coefficients obtained through calibration using historical flood data; they collectively reflect the nonlinear characteristics of the river channel cross-sectional morphology.

[0108] Furthermore, applying the aforementioned correction function, the linear impulse response solution obtained through the Green's function method is... By performing point-by-point corrections, we can obtain a more physically accurate impulse response function that has undergone nonlinear correction. .

[0109] Furthermore, the river network is discretized into N computational nodes, and the unit impulse response is calculated by numerical integration for any pumping station i and influence point j. and all When assembled, they form an N×N impulse response function matrix H. This matrix H is a physical law library that quantifies the spatiotemporal interaction between any two points in the entire river network.

[0110] In an optional embodiment, in the above relation (5) and The specific calibration method is as follows: Several typical historical flood events within the watershed with complete hydrological data (including upstream inflow processes, inter-regional rainfall processes, water level processes at various cross-sections along the river, and pump station operation records) were selected for study.

[0111] Furthermore, for each historical flood event, without including nonlinear corrections (i.e. , Under the condition of ), the hydrodynamic model (a model that can efficiently and accurately simulate the global and spatiotemporal dynamic effects caused by local drainage behavior) is run to obtain a set of benchmark simulated water level processes for each section.

[0112] Furthermore, constructing a system based on and The objective function for the independent variable is to minimize the error between the simulated water level and the historical measured water level.

[0113] One typical objective function It can be the sum of the root mean square error (RMSE) of all key sections at all times, as shown in the following relationship (6): (6) In the formula: This indicates the model output including nonlinear corrections; Indicates historical measured values; Represents a set of historical flood events; This represents the set of key river sections.

[0114] Furthermore, nonlinear optimization algorithms, such as the Levenberg-Marquardt algorithm or a genetic algorithm, are used to iteratively solve the above objective function to find the solution that makes the objective function more efficient. Reaching the minimum value and These optimal solutions are the parameters obtained from the final calibration.

[0115] Step S2022: Based on the impulse response function matrix, generate a three-layer asynchronous adaptive mesh containing a global coarse mesh, an influence region fine mesh, and an impulse region ultrafine mesh.

[0116] In one alternative embodiment, based on the spatial distribution characteristics of the influence intensity revealed by the impulse response function matrix, a three-layer asynchronous adaptive grid can be generated, comprising a global coarse grid, an influence region fine grid, and an impulse region ultrafine grid.

[0117] In one alternative embodiment, since the impact of the pumping station on flood drainage is spatially uneven, with drastic and rapid gradient changes in the area near the pumping station (pulse zone) and the main water flow channel (affected zone), a high-resolution grid is required to accurately capture the impact. In contrast, a coarse grid can be used in the vast, less affected areas (normal zone) to save computational resources.

[0118] For example, based on the impulse response function matrix H, a spatial influence intensity index characterizing the cumulative influence intensity at each location in the river network is calculated. ,For example .

[0119] Furthermore, this spatial impact intensity index Reflects location The intensity of the maximum combined impact that may be generated by all pumping stations.

[0120] Furthermore, the spatial second derivative of this spatial influence intensity index is evaluated. And by combining the second derivative with a set of preset thresholds and ( By comparing these, the pulse region, the affected region, and the normal region can be identified and divided.

[0121] For example, when At that time, the region was identified as the pulse region with the most dramatic gradient changes; when At that time, the region was identified as the area with significant gradient changes; the remaining regions were considered normal regions.

[0122] Furthermore, mesh elements with ultra-fine, fine, and coarse spatial resolutions are configured for the divided pulse region, affected region, and normal region, respectively, to construct a three-layer asynchronous adaptive mesh. For example, the mesh size of the pulse region can be configured. Area of ​​impact Regular area .

[0123] In an optional embodiment, in order to achieve effective information transfer between different grid levels, an overlapping area can be set at the boundary of grid layers with different spatial resolutions.

[0124] Furthermore, within this overlapping region, numerical interpolation relationships are established between the boundary nodes of the fine mesh and the internal nodes of the coarse mesh to form a mesh nesting boundary mapping table for cross-scale information transfer. For example, the water level values ​​of the boundary nodes of the fine mesh. It can be determined by the water level value of the internal nodes of the coarse grid cell in which it is located. Bilinear interpolation yields: ,in These are the weights for bilinear interpolation.

[0125] Step S2023: In the three-layer asynchronous adaptive grid, hydrodynamic calculations are performed using a standardized input dataset, and the calculation results are obtained at multiple levels.

[0126] In one alternative embodiment, for different levels of mesh, a corresponding numerical solution format and asynchronous time step are used to perform layered independent calculations.

[0127] In one optional embodiment, hierarchical independent calculations are performed using appropriate numerical solution schemes, specifically including: On a global coarse grid, the coverage is the widest but the physical quantities change gradually. To ensure long-term stability and efficiency of the calculation, an implicit difference scheme is preferred for basis current calculation, prioritizing stability and computational efficiency. For example, the robust and computationally efficient Preissmann four-point implicit scheme can be used.

[0128] On the fine grid of the affected area, the primary function is to capture the propagation and evolution of flood waves. To effectively capture the nonlinear characteristics of flood waves, such as shock waves, an explicit prediction-correction scheme is preferred. For example, the MacCormack explicit scheme, which can handle convection terms well, can be used.

[0129] In the ultrafine grid of the pulse region, the water flow state will be severely interrupted due to the start and stop of the pump station. In order to accurately handle the flow interruption caused by the start and stop of the pump station, a high-precision weighted essentially oscillatory-free (WENO) scheme is preferred. For example, a fifth-order WENO scheme that can effectively suppress numerical oscillations can be used.

[0130] In an optional embodiment, a cross-scale flux conservation correction can also be periodically performed at the boundaries of different grid levels to ensure the physical authenticity and numerical continuity of cross-level computation.

[0131] For example, at each preset synchronization time point The flux at the interface between the coarse and fine meshes needs to be checked. The total flux flowing through the coarse mesh boundary within one time step needs to be calculated. Simultaneously calculate the sum of flux flowing through all corresponding fine mesh boundaries. Ideally, the two should be equal; if a difference exists, a flux correction factor should be calculated. This factor is used to correct the flux or water level values ​​at the fine grid boundary, thereby forcibly ensuring the conservation of mass and momentum at the cross-scale boundary.

[0132] In an optional embodiment, the hierarchical independent propulsion calculation also includes a dynamic activation mechanism. Furthermore, the motivation for this dynamic activation mechanism is that high-resolution fine and ultra-fine grids are only necessary to be calculated during periods affected by pumping stations, and can remain silent at other times to conserve resources. Specifically, this includes: real-time monitoring of a pumping station control command sequence derived from the optimal drainage strategy time series; when the command sequence indicates that a pumping station is about to start, querying and determining the fine grids of the affected area and the ultra-fine grids of the pulse area associated with the pumping station based on a pre-established grid nesting boundary mapping table or grid mapping index table; setting the queried and determined fine grids of the affected area and ultra-fine grids of the pulse area to an active state to trigger high-resolution propulsion calculations on these grids. Once the pumping station's influence subsides, these grids are restored to an inactive state.

[0133] Furthermore, the aforementioned dynamic activation mechanism enables on-demand computing services, thereby achieving efficient asynchronous computing.

[0134] For example, the dynamic activation mechanism is the core of realizing on-demand allocation of computing resources, and its complete lifecycle process is as follows: 1. Continuously monitor the pump station control command sequence generated in Example 11.

[0135] 2. When it is detected that a pumping station p is about to [experience a sudden change in power] within a short time window (e.g., within the next 30 minutes), [the following information is provided]. When the system starts, a pre-activation process is triggered.

[0136] 3. Query the pre-established grid mapping index table to find the IDs of all fine grids in the influence zone and ultrafine grids in the pulse zone associated with the geographical location of pump station p. Set the status of these grids from inactive to active.

[0137] 4. Boundary condition initialization (using overlapping regions): In At time 1, for the newly activated fine mesh, the initial boundary conditions required for its computation are obtained by interpolating the internal node values ​​of its parent coarse mesh at the same time using the mesh nesting boundary mapping table described in step S2022. This step is crucial for achieving seamless computation startup.

[0138] 5. Asynchronous computation: Activated grids begin computation according to their actual time steps ( or (To perform high-resolution calculations.)

[0139] 6. Information Synchronization and Flux Correction: At each global coarse grid time step At the end point, cross-scale information synchronization is performed. At this time, the active fine grid summarizes its boundary flux, compares it with the flux of the corresponding boundary of the coarse grid, and corrects it according to the flux correction factor to ensure physical conservation.

[0140] 7. Deactivation: When the pump station stops operating and its effects have physically decayed sufficiently (which can be determined by the impulse response function), the previously activated grid state is reset to inactive, its high-resolution calculation is stopped, and it returns to a state covered only by a coarse grid, thereby freeing up computing resources.

[0141] Step S2024: Prioritize and fuse the calculation results of multiple levels, and generate a multi-temporal and spatial resolution water level field.

[0142] In one optional embodiment, by utilizing the calculation results of different levels of mesh (ultra-fine for the pulse region, fine for the influence region, and coarse for the regular region), and based on the priority of the regional influence intensity, high-precision mesh results are used in the critical region (where the influence is severe) and low-precision mesh results are used in the regular region. This achieves the fusion of high-resolution water level fields in the critical region and low-resolution water level fields in the non-critical region, taking into account both simulation accuracy and computational efficiency. In the end, it is possible to generate a multi-temporal resolution water level field that retains high spatiotemporal details in the critical regions such as near the pumping station and water flow channels, while simplifying the calculation in the regions with weak influence.

[0143] In one alternative embodiment, firstly for any spatial location and time Set the priority of calculation results according to region type: if If the result is located within an ultrafine mesh in an active pulse region, the final result uses the calculation value from the ultrafine mesh; otherwise, if... If the value is located within a fine mesh in an active influence zone, the calculation value of the fine mesh is used; in all other cases, the calculation value of the global coarse mesh is used.

[0144] Furthermore, by filtering and integrating the calculation results of each grid according to the aforementioned priority rules, a multi-temporal resolution water level field with high spatiotemporal resolution in key regions and lower resolution in other regions can be obtained. .

[0145] Step S2025: Based on the multi-temporal resolution water level field and the preset river network topology model, determine the regional interaction influence intensity distribution matrix and the flood wave propagation delay matrix.

[0146] Specifically, step S2025 above includes: Step b1: Obtain the real-time water level dataset based on the multi-temporal and spatial resolution water level field.

[0147] In an alternative embodiment, water level fields with multi-temporal and spatial resolution can be used. Extract each calculation section within the river network at any time. The real-time water level data is collected and a real-time water level dataset is generated.

[0148] Step b2: Generate a time-varying propagation velocity field based on the real-time water level dataset and the preset river cross-section parameter library.

[0149] In one optional embodiment, the preset river cross-section parameter library represents a pre-stored dataset that describes the geometric morphological characteristics of each calculated cross-section within the river network. It may include parameter information such as the cross-sectional area, water surface width, and cross-sectional shape (e.g., rectangle, trapezoid, natural irregular shape) corresponding to different water levels.

[0150] In one alternative embodiment, the time-varying propagation velocity field is used to reflect real-time hydrological conditions and river geometry.

[0151] In one optional embodiment, dynamic wave velocity (including river bend correction) is calculated by combining real-time water level and river cross-sectional geometric parameters. The wave velocity data of the entire river network is integrated to form a time-varying field, so that the wave velocity estimation fits the real-time water conditions and river morphology. In this way, a wave velocity field that changes dynamically with time and space can be generated, which improves the accuracy of flood propagation time calculation.

[0152] For example, for each calculation section in the river network, at any given time... First from The real-time water level of the cross-section is extracted. Then, combined with a pre-stored library of river cross-section parameters describing the geometry of the cross-section, the real-time hydraulic parameters, which may include the cross-sectional area of ​​the water passage, are calculated. and water surface width .

[0153] Furthermore, the dynamic wave velocity of the cross section is calculated using the following relationship (7). : (7) In the formula: It represents the acceleration due to gravity.

[0154] In an optional embodiment, to improve the accuracy of wave velocity calculation, the influence of river channel bends on wave velocity can be considered, in which case a bend correction factor is introduced. ;in: For position The dimensionless correction factor at the location; Let be the radius of curvature of the center of mass of the water flow; Let be the radius of curvature of the river channel centerline. Finally, the corrected effective wave velocity is obtained. .

[0155] Furthermore, by integrating the effective wave velocities of all sections of the entire river network, a time-varying propagation velocity field can be formed that reflects real-time hydrological conditions and river channel geometry. .

[0156] Furthermore, the dynamic calculation method described above can eliminate the simplistic approach of using fixed or empirical wave velocities, making the wave velocity estimation highly consistent with physical reality, which helps to accurately calculate the propagation delay.

[0157] Step b3: Generate a decision-oriented graph based on the preset river network topology model.

[0158] In one optional embodiment, the flood wave propagation delay matrix uses the decision region as nodes and the direction of water flow influence as directed edges.

[0159] In one alternative embodiment, the river network topology model is mapped to a directed decision graph with decision regions as nodes and the direction of water flow influence as directed edges. For example, if the water flow in region A affects region B, then there exists a directed edge in the graph pointing from A to B.

[0160] For example, the river network topology model can be viewed as a directed graph G=(N,E), where N is the set of nodes and E is the set of edges. For each edge in the graph (representing a river segment), its weight is dynamically set to the average flood wave transit time of that river segment, i.e., the path length ds divided by the effective wave velocity of that river segment. That is, the path weight is .

[0161] Step b4: Using the reciprocal of the time-varying propagation velocity field as the path weight, the decision directed graph is processed by the shortest path search algorithm to obtain the set of target propagation paths.

[0162] In one optional embodiment, the reciprocal of the wave speed (corresponding to the propagation time per unit distance) is used as the path weight. The time-varying shortest path algorithm is used to search for the fastest propagation path between any two nodes, thus finding the actual channel with the shortest flood propagation time between any two decision areas in the river network. This ensures that the path conforms to the flood propagation law under real-time water conditions and avoids time delay errors caused by path misjudgment.

[0163] For example, the time-varying Dijkstra algorithm can be used to search for and determine any two decision region nodes in a river network. and The shortest propagation time path between .

[0164] Furthermore, "shortest" refers to the shortest time, not the shortest geographical distance.

[0165] Furthermore, the above process can accurately identify the fastest propagation path that flood impacts will actually follow in complex river networks, avoiding propagation time calculation errors caused by incorrect path selection.

[0166] Step b5: Construct the flood wave propagation delay matrix based on the target propagation path set.

[0167] In one optional embodiment, for each shortest propagation path, the propagation time between the two regions is calculated by path integral, and all time results are assembled into a matrix to generate a propagation time matrix with the decision region as the dimension, which quantifies the spatiotemporal correlation of flood propagation between regions.

[0168] For example, for each path in the obtained set of optimal propagation paths The final propagation delay can be calculated by integrating the path with respect to the reciprocal of the effective wave speed. The following relation (8) is shown: (8) Furthermore, all decision-making region nodes are paired Propagation delay between Through calculation and assembly, an M×M flood wave propagation delay matrix (where M is the total number of decision-making areas) is finally constructed, which can comprehensively reflect the dynamic connectivity of the entire river network. .

[0169] Step b6 involves convolving and integrating the impulse response function matrix and the historical target drainage strategy time series to determine the regional interaction intensity distribution matrix.

[0170] In one optional embodiment, the spatiotemporal impact of unit drainage behavior described by the impulse response function matrix is ​​utilized. Through convolutional integrals, the actual drainage process of pumping station scheduling is decomposed into the superposition of multiple impulses, enabling the quantification of the time-varying cumulative impact of drainage behavior in one area on other areas. Furthermore, by introducing a nonlinear correction term, the nonlinear hydraulic effects of multiple pumping stations operating simultaneously can be adapted, improving the physical accuracy of the quantification results.

[0171] In an optional embodiment, an impulse response function matrix that characterizes the propagation characteristics of the pumping station's influence is convolved with the actual pumping station scheduling scheme corresponding to the time series of the optimal drainage strategy to quantify the time-varying cumulative impact of drainage behavior in any region on other regions, thereby determining the distribution of regional interaction intensity.

[0172] An example of an actual pump station scheduling scheme This refers to the time series of the optimal drainage strategy determined in the previous stage, which is also the time series of the historical target drainage strategy.

[0173] Furthermore, the region The drainage activities have an impact on the region The time-varying cumulative effect of water level, i.e., the intensity of the effect. It can be calculated using the convolution integral shown in the following relation (9): (9) In the formula: Indicates the current moment; This represents the integral variable, which represents any point in the past. Represents the impulse response function, indicating Time zone Unit pulse drainage, after Time after the region The resulting impact on water levels; express Time zone The actual drainage flow.

[0174] Furthermore, the physical meaning of the convolution integral is to transform the region... The impacts of drainage activities at all past moments, according to their respective propagation laws and attenuation characteristics, are linearly superimposed onto the current moment. Thus, its influence on the region is obtained. The total cumulative effect. Applying this to all regions. Influence intensity By performing calculations and assembling the data, a time-varying matrix of regional interaction intensity can be obtained. .

[0175] In an alternative embodiment, nonlinear superposition analysis can be introduced to more accurately simulate the complex hydraulic effects of multiple pumping stations operating simultaneously.

[0176] For example, when multiple upstream pumping stations discharge water into the same river section at the same time, the resulting downstream water level rise (swelling) effect is often not a simple linear sum of the individual effects of each pumping station, but rather exhibits a nonlinear enhancement effect where 1+1>2.

[0177] Furthermore, to capture this effect, in the computational region Total impact In this case, a nonlinear superposition term can be added to the linear superposition. ,Right now .

[0178] Furthermore, the nonlinear superposition term can be modeled as the following relation (10): (10) In the formula: This represents a dimensionless nonlinear interaction parameter obtained through calibration using historical measured data or a refined hydrodynamic model.

[0179] Furthermore, by introducing this nonlinear correction, the comprehensive impact under extreme conditions (such as multi-regional coordinated strong placement) can be quantified more accurately, providing more reliable and safer data support for risk assessment and game theory decision-making, thereby improving the robustness of the entire simulation method.

[0180] In an optional embodiment, to reflect the nonlinear and time-varying propagation characteristics of the river channel, a water level field can be used in practical applications. Corrected effective impulse response matrix With based The calculated time delay matrix Perform convolution, that is The final determination involves two layers of relationships: linear convolution terms. (For preliminary estimation) and nonlinear correction terms based on multi-temporal and spatial resolution water level fields (For online calibration), thus forming .

[0181] Furthermore, an indirect but necessary dependency exists between the regional interaction intensity distribution and the multi-temporal and spatial resolution water level field: Used to determine dynamic wave velocity and propagation delay; calibrate and correct the impulse response function online; provide spatial weighting and thresholding criteria for convolution results. Real-time multi-temporal resolution water level field can be temporarily omitted only during offline calibration or initial estimation stages.

[0182] In an optional embodiment, the regional interaction intensity matrix (Its elements are) This is a discretized representation of the continuous distribution across discrete decision region nodes. Its more rigorous continuous mathematical form is: ;in: As the source In the past moment Unit drainage flow rate; This is a constructed, nonlinearly modified continuous Green's function (i.e., impulse response function). Therefore, this embodiment can be achieved through discretization. It performs rapid calculations and game theory, but behind it lies a continuous process. It provides theoretical support and physical explanation.

[0183] In some optional implementations, step S203 above includes: Step S2031: Use the flood wave propagation delay matrix to determine the asynchronous decision-making time series.

[0184] In one alternative embodiment, asynchronous decision time series is used to ensure that the upstream decision influence of each decision region can physically propagate to the downstream before downstream decision-making is carried out.

[0185] In an optional embodiment, a topological sorting algorithm, such as breadth-first search (BFS), is performed on the decision-directed graph generated in step b3 to obtain a basic decision priority sequence that clearly defines the order of decisions for each region. This sequence ensures that any region makes a decision only after all its upstream regions have made their decisions.

[0186] Furthermore, following this decision priority sequence and based on the flood wave propagation delay matrix... Assign specific decision-making moments to each decision-making region.

[0187] For example, for the most upstream source region Its decision-making time can be set as the initial time. For any region downstream of it. Its decision-making moment It is then set to Among them, the one here Indicates the area The set of all directly upstream regions, Indicates in Flood waves from the region Spread to the region The required time. In this way, an asynchronous decision-making time series that ensures physical causality can be constructed. .

[0188] Furthermore, by embedding the hydraulic propagation delay in the physical world into the game's rule layer through the aforementioned asynchronous decision-making mechanism, invalid or conflicting decisions caused by information asynchrony are avoided, ensuring the scientific and forward-looking nature of downstream decisions.

[0189] Step S2032: Obtain the time-varying revenue function set.

[0190] In an optional embodiment, the time-varying revenue function set is used to integrate the current term of the security status of each region, the upstream historical impact term quantified by the flood wave propagation delay matrix, and the future expected responsibility term for the downstream region.

[0191] For example, for each decision region At its decision-making moment Constructed profit function Its specific form is shown in the following relation (11): (11) In the formula: Indicates the area exist The amount of time required for flood drainage decisions; , , This represents the time-varying weighting coefficient that is dynamically adjusted. , , Let represent the utility functions for the current item, the historical impact item, and the future responsibility item, respectively.

[0192] Specifically, step S2032 includes: Step c1: Calculate the current item that reflects the security status of each decision-making region.

[0193] Step c2: Calculate the upstream historical impact term using the flood wave propagation delay matrix.

[0194] Step c3: Based on the impulse response function matrix, determine the expected future liability items.

[0195] Step c4: Based on the regional interaction influence intensity distribution matrix, dynamically configure time-varying weight coefficients for the current item, upstream historical influence items, and future expected responsibility items, and generate a set of time-varying revenue functions.

[0196] In one alternative embodiment, the current item is first determined: it can be determined from a multi-temporal resolution water level field. In the middle, directly extract the decision region At this moment of decision-making water level .

[0197] Furthermore, the utility function of the current term It is typically used to represent the flood control safety benefits of a region itself; for example, it can be set as a quadratic function related to the safe water level. ,in For the region The safe target water level.

[0198] Secondly, calculate the historical impact term from upstream. Specifically, this can be done based on the flood wave propagation delay matrix. Extracting upstream regions from this multi-temporal and spatial resolution water level field In the corresponding propagation delay Previous water level data Furthermore, the upstream historical influence term is calculated by weighted summation, as shown in the following relation (12): (12) Among them, weight Further consideration could be given to factors such as distance attenuation and watershed area ratio, for example... ,in, For the region arrive The distance; The distance attenuation constant; and These represent the catchment areas of the two regions.

[0199] Furthermore, the utility function of the historical influence term This is used to punish the risk of exceeding limits caused by upstream water flow, for example... ,in For the region The warning water level.

[0200] Finally, the expected future responsibilities are determined. Specifically, this is based on an impulse response function matrix H characterizing the propagation characteristics of the pumping station's influence and a predetermined test drainage volume. Predict and quantify the impact of current decisions on downstream regions. Future water level impact The specific calculation is shown in the following relation (13): (13) In the formula: Indicates the area The unit pulse drainage has a propagation delay Then for downstream areas The resulting impact on water levels The importance weight of downstream regions.

[0201] Furthermore, the utility function of expected future liabilities This is used to measure the negative impact on downstream industries, for example... .

[0202] Furthermore, it can be based on the distribution of regional interaction intensity. And the current water situation extracted from the multi-temporal resolution water level field. A set of time-varying weighting coefficients is dynamically assigned to the current item, the upstream historical impact item, and the future expected responsibility item. , , .

[0203] For example, the following dynamic adjustment mechanism can be designed: 1. This indicates that when the water level in the area is higher, the decision-making process should focus more on its own safety, meaning the weight of the current item is greater. 2. This indicates that the weight of historical items decreases as the regional interaction influence weakens. for Time zone The overall intensity of the impact; 3. This indicates that the closer the connection with downstream entities, the higher the weight of future liability items should be. At the same time, it is necessary to ensure... .in, This represents the average propagation delay to the downstream.

[0204] Furthermore, by dynamically adjusting the relative importance of each influencing factor in the decision optimization process, the game-theoretic decision can intelligently adapt to different working conditions, such as being more self-interested when in crisis and more cooperative when there are close upstream and downstream connections.

[0205] In one example, the construction of the time-varying revenue function described above is illustrated through a specific and reproducible numerical calculation process.

[0206] 1. Assume a simplified river network containing three decision-making units: upstream region A, midstream region B, and downstream region C. The objective of this example is to calculate the decision time of midstream region B. Regarding its drainage decision-making volume payoff function .

[0207] 2. Decision-making time and water level: The decision-making time for area B is... At this moment, the water level in region B is measured from the multi-temporal resolution water level field. The safe target water level for area B. Warning water level The safe target water level for downstream area C. .

[0208] 3. Propagation Delay and Historical Data: The propagation delay from A to B is obtained from the flood wave propagation delay matrix T. .exist At any time, a query is required. Historical data of hours were measured .

[0209] 4. Impulse Response and Future Prediction: From the impulse response function matrix H, the influence function of B on C is obtained during the propagation delay. The value at that location is This means that for every increase in B... The drainage volume will cause the water level at point C to rise by 0.005m after 3 hours. Importance weight of downstream area C. Set to 1.2.

[0210] 5. Geometric and Influencing Parameters: Catchment area of ​​upstream region A The catchment area of ​​region B The distance of the river channel from A to B Distance attenuation constant .exist At what time, the total influence intensity index of region B Average propagation delay from B to downstream .

[0211] 6. The drainage flow rate to be decided for Region B is: The unit is m³ / s.

[0212] Furthermore, the step-by-step calculation process includes: The first step is to calculate the current item. The utility value. This item reflects the security status of region B itself. According to the utility function defined in step S2032 above, its value is: .

[0213] The second step is to calculate the upstream historical impact item. The utility value. First, calculate the weight of the influence of upstream region A on B. : .

[0214] Then, calculate the upstream historical impact water level. : .

[0215] Finally, calculate the utility value of this item: .

[0216] This result indicates that the current impact of upstream water flow has not yet exceeded the warning level of B.

[0217] The third step is to calculate the expected future liabilities. The utility function. First, predict the impact of decision Q_B on the future water level of downstream C. : .

[0218] Then, calculate the utility function of this term, which is a function of... Functions: .in, This should be understood as the baseline water level without the influence of B. For simplicity, we assume here that the baseline water level is the safe water level, but the actual calculation will be more complicated.

[0219] Furthermore, to make the formula more meaningful, it should be: To simplify the calculation in this example, we assume... ,but .

[0220] The fourth step is to calculate the dynamic time-varying weighting coefficients. , , The calculation is performed based on the formula in step S2032 and the parameter values ​​set in this example: ; ; .

[0221] Furthermore, to ensure that the sum of the weights is 1, normalization is performed: .further, ; ; .

[0222] The fifth step is to integrate the results to obtain the final profit function U_B. By integrating the above components and their weights, the final profit function for region B is obtained, as shown in the following equation (14).

[0223] (14) Furthermore, the above calculations show that the goal of region B in making decisions is to solve... By solving for this decision variable... The optimal drainage flow rate can be obtained by finding a quadratic function. (In this example, it is 0, because) (The coefficient of the term is negative and there are no positive benefit terms).

[0224] Furthermore, the above calculation process fully demonstrates how the present invention transforms complex multidimensional factors into a clear and optimizable mathematical objective, fully proving the logical rigor and feasibility of its technology.

[0225] In an optional embodiment, the time-varying revenue function set can also be obtained in the following manner.

[0226] For example, for the decision-making region At any moment From multi-temporal and spatial resolution water level fields The upstream historical influence is extracted as shown in the following relationship (15): (15) Among them, weight Considering distance attenuation and watershed area ratio, a historical influence vector is generated. .

[0227] Furthermore, the region is predicted based on the impulse response function matrix H. The impact of the decision on downstream industries: ,in To test the drainage capacity, Assigning importance weights to downstream components through sensitivity analysis. Determine and generate future impact prediction vectors. .

[0228] Furthermore, based on the regional interaction intensity matrix Based on the current water situation, the weights are dynamically adjusted: Used for the current item. Used for historical items, Used for future items, to ensure Generate a time-varying weight coefficient set .

[0229] Furthermore, the three temporal terms are integrated to construct a complete payoff function: , in For safety and profit; As punishment for history; For future responsibility, output a set of time-varying revenue functions .

[0230] Step S2033: Based on the asynchronous decision time series, with the time-varying payoff function set as the optimization objective, the Nash equilibrium is solved iteratively to generate the real-time target drainage strategy time series.

[0231] In one optional embodiment, using asynchronous decision-making time series as the time-series constraint, each region, at its decision time, takes its own time-varying benefit function as the optimization objective and iteratively solves the Nash equilibrium: each region adjusts its own drainage strategy to maximize benefits based on the historical decisions made by other regions; numerical algorithms such as gradient ascent are used to iteratively update the strategy until the decision quantities of all regions converge (the change in decision quantities between adjacent iterations is less than a preset threshold); and then, by solving, the optimal drainage strategy time series that satisfies the physical propagation time sequence and global coordination can be finally formed, namely the real-time target drainage strategy time series.

[0232] For example, according to the established asynchronous decision time series Each region At its decision-making moment The requirement is to solve an optimization problem: ,in This indicates historical decisions that have been made in other regions.

[0233] Furthermore, numerical optimization algorithms such as gradient ascent can be used to iteratively solve the above optimization problem: .in, Indicates the number of iterations; This represents the learning rate. Furthermore, the iterative process continues until the decision values ​​for all regions converge, i.e., the convergence condition is met. ,in This is a preset minimum threshold. Furthermore, the converged policy set... This is the preliminary optimal drainage strategy, i.e., the real-time target drainage strategy time series.

[0234] In an optional embodiment, since standard asynchronous iterative algorithms may exhibit slow convergence or oscillations around the optimal solution when dealing with complex systems, this embodiment introduces mechanisms such as state management, convergence monitoring, and momentum acceleration to ensure stable and efficient convergence of the algorithm. Specifically, this includes: The first step is asynchronous decision state initialization and management. In a preferred implementation of this embodiment, the system first constructs and maintains an asynchronous decision state matrix S. This matrix has dimensions n×T, where n is the number of decision regions and T is the number of discrete-time steps within the decision period. Matrix elements... Indicates the area At any moment The decision state. This state can be defined as: 1. State 0 (No Decision): Decision-making has not yet begun in this region at this moment; 2. State 1 (Decision Completed): The decision calculation for this region has been completed and converged at this moment; 3. State 2 (In Decision-Making): This region is currently undergoing iterative computation of its decision.

[0235] Furthermore, before the iteration begins, based on the determined asynchronous decision time series, the state of all regions at their respective decision times is initialized to undecided, and an initial strategy is set, such as the normal drainage volume for each region. .

[0236] Furthermore, the above steps manage the complex asynchronous decision-making process through a formalized state machine, ensuring that decisions in each region are made strictly in a predetermined order, at the right time, and based on the correct input information (i.e., all relevant upstream regions are already in a decided state), thus avoiding data races and logical confusion.

[0237] The second step is distributed gradient calculation and asynchronous policy update.

[0238] Specifically, following the order of the asynchronous decision-making time series, when the region At its decision-making moment When the state transitions to decision-making, based on its real-time payoff function... Calculate the gradient of the decision variables .

[0239] Furthermore, the region Update the strategy based on this gradient: .in, Indicates the number of iterations; This represents a learning rate that decreases with the number of iterations, for example... ,in, and Indicates preset parameters; This represents the projected gradient that takes into account constraints such as the upper and lower limits of the pump station's capacity.

[0240] Furthermore, after the update is complete, the region They will immediately implement their new strategy. The impact is calculated and updated by propagating the delay matrix T to prepare for subsequent decisions in the downstream regions.

[0241] The third step, convergence monitoring and momentum acceleration: After each iteration update, the system performs convergence monitoring on all regions in the decision-making state. Specifically, it calculates the convergence of each decision region. relative change rate of strategy .in, This represents a very small positive number that prevents the denominator from being zero.

[0242] Furthermore, when all active areas All are less than a preset convergence threshold When these regions converge, their states can be updated to "decided".

[0243] Furthermore, to address potential oscillations or convergence stalls, this embodiment introduces a momentum acceleration mechanism.

[0244] Specifically, the system monitors the improvement of each iteration. For example, if the total improvement of the reward function is less than a minimum value in three consecutive iterations, momentum acceleration is activated. The policy update formula is then modified as follows: .in, This represents a momentum coefficient, typically ranging from 0.8 to 0.99.

[0245] Furthermore, the momentum term The physical meaning is that an additional inertial component related to the previous update direction is added in the current update direction.

[0246] Furthermore, the aforementioned momentum acceleration mechanism can help the iterative process overcome small saddle points or flat regions on the surface of the reward function and effectively suppress high-frequency oscillations caused by asynchronous updates, thereby significantly improving the convergence speed and stability of the algorithm.

[0247] In an optional embodiment, the momentum coefficient Instead of using a fixed value, adaptive adjustments can be made. The specific adaptive adjustment mechanism is as follows: when the iteration direction is stable (i.e., the gradient direction changes little), the probability is increased to encourage it to continue accelerating in this direction; when the iteration direction changes drastically (potential overshoot or oscillation may occur), the probability is decreased. This serves to brake and suppress oscillations.

[0248] For example, in each iteration update Then, calculate the current gradient. gradient with the previous step The dot product.

[0249] Furthermore, if This indicates that the gradient directions are basically the same, which can increase the probability. ;like This indicates a reversal of the gradient direction, which may cause oscillations and should be reduced. : .in, and express The upper and lower limits (e.g., 0.99 and 0.5); and This indicates adjusting the step size.

[0250] The fourth step is the final verification of the asynchronous equilibrium solution. Specifically, after all regions have reached a decided state, i.e., the entire system has converged, the system will perform a final verification of the Nash equilibrium conditions. That is, for the final policy set... Verify whether this applies to all regions. All satisfied ,in It is a region Any other feasible strategy.

[0251] Furthermore, if the verification passes, the initial set of drainage strategies is output. If it does not meet the requirements, it may be necessary to adjust hyperparameters such as the learning rate or momentum coefficient and iterate again.

[0252] Furthermore, by introducing refined state management, convergence monitoring, momentum acceleration, and final verification, a complete and robust engineering implementation scheme is provided for the solution process of asynchronous games, ensuring the feasibility and efficiency of the algorithm in complex practical applications.

[0253] In some optional implementations, step S203 above includes: Step S2034: Using the real-time target drainage strategy time series as the historical target drainage strategy time series, re-perform the cross-regional hydrodynamic interaction simulation and generate the resulting water level field.

[0254] In one optional embodiment, the currently obtained real-time target drainage strategy is used as input and substituted into the hydrodynamic simulation model. The actual water level field under the strategy is calculated by the model, which can verify the water level response of the drainage strategy under real hydraulic laws.

[0255] In an optional embodiment, the preliminary drainage strategy set is reloaded into the cross-regional hydrodynamic interaction simulation model, that is, the real-time target drainage strategy time series is used as the historical target drainage strategy time series, and the cross-regional hydrodynamic interaction simulation is performed again in step S202 to verify its physical feasibility and generate a result water level field.

[0256] For example, the preliminary optimal drainage strategy set obtained in step S2033 is... That is, the real-time target drainage strategy time series is used as input to the hydrodynamic model (the model described in step a3) to perform a rapid verification simulation calculation.

[0257] Step S2035: Verify the resulting water level field and the preset physical constraint set, and identify whether there are any violations in the resulting water level field.

[0258] In one optional embodiment, the simulated water level field is compared with preset engineering safety constraints to check whether the strategy violates physical limitations such as water level, flow rate, and hydraulic gradient, thereby identifying potential safety risks.

[0259] In one optional embodiment, the preset physical constraint set includes at least: 1. Water level constraints: Check the water levels at key sections in each region obtained from the simulation. Whether it exceeds its guaranteed water level or dike elevation at any time .

[0260] 2. Flow Constraints: Check the drainage flow rate of each area. Does it exceed the maximum installed drainage capacity of its pumping station? .

[0261] 3. Hydraulic gradient constraint: Check the hydraulic gradient in the river channel. Does it exceed the preset threshold? This is to avoid abnormal backflow or turbulent water flow.

[0262] Furthermore, during the verification process, all spatiotemporal points that violate the above constraints will be recorded, forming a constraint violation record table.

[0263] Step S2036: When a violation occurs, the real-time target drainage strategy time series is iteratively corrected using a penalty function until no violation occurs, and the real-time target drainage strategy time series is determined.

[0264] In one optional embodiment, when a violation is identified, an augmented objective function is constructed by adding a penalty term to the original reward function. The drainage strategy is then adjusted iteratively to satisfy all physical constraints. Furthermore, the penalty mechanism guides the iterative correction of the strategy, transforming the mathematically optimal solution obtained through game theory optimization into an engineering-safe and feasible scheduling scheme, thereby enhancing the robustness and reliability of the strategy.

[0265] For example, the correction procedure is initiated when the constraint violation record table is not empty. Furthermore, the core of this procedure is to construct an augmented objective function that adds a penalty term related to the severity of the violation to the original reward function. .

[0266] Furthermore, this penalty function It can be designed as the sum of various violation penalties. As shown in (16): (16) In the formula: , , This represents the penalty coefficient for each item. Its value can be adaptively adjusted according to the severity of the violation. For example, the greater the number of violations, the larger the coefficient value.

[0267] Furthermore, the correction process is a new optimization solution process, the goal of which is to maximize... This problem can be solved iteratively using methods such as the Lagrange multiplier method or the alternating direction multiplier method (ADMM) to generate a corrected policy vector. .

[0268] For example, the objective function to be optimized can be written in the form of an augmented Lagrangian function: .in, The original profit function, This is a penalty term for physical constraints. For equality constraints that need to be satisfied (e.g., a certain coupling relationship between different regional strategies). For Lagrange multipliers, This is the penalty parameter.

[0269] Furthermore, the ADMM algorithm updates alternately. and To solve, in the first In the next iteration: 1. -Update step: Fixed multipliers Seeking answers regarding The minimization problem (or maximization for the payoff function): This step is the core computational step, and it is usually solved using methods such as gradient descent.

[0270] 2. -Update step (multiplier update): Fixed Update the Lagrange multipliers: Repeat the above two steps until the convergence condition is met, such as the original residual. and dual residuals All are less than the preset minimum threshold.

[0271] Furthermore, the corrected strategy Qcorrected is substituted back into steps S2034 and S2035 for verification. This simulation-verification-correction cycle continues until a verification result fully meets all physical constraints. At this point, the strategy set is determined as the final, physically feasible, and coordinated optimal drainage strategy time series, i.e., the real-time target drainage strategy time series.

[0272] Furthermore, through the aforementioned closed-loop mechanism of verification and correction, the mathematically optimal solution optimized by game theory can be successfully transformed into a safe and reliable scheduling scheme in engineering, thereby enhancing the robustness and credibility of this invention in practical applications.

[0273] In some alternative implementations, after step S203 above, the method includes: Step d1: Obtain the target multi-temporal resolution water level field corresponding to the real-time target drainage strategy time series.

[0274] Step d2 involves calculating multiple discrete index values ​​based on the regional interaction influence intensity distribution matrix, the real-time target drainage strategy time series, and the target multi-temporal resolution water level field.

[0275] Step d3: Based on multiple discrete index values, the overall synergy score is obtained through a weighted comprehensive evaluation model.

[0276] In one alternative embodiment, comprehensive decision support can be provided by evaluating the overall performance of the strategy in a multi-dimensional and quantitative manner and presenting it to decision-makers in an intuitive and visual form.

[0277] In an optional embodiment, the optimal drainage strategy is obtained. and the resulting multi-temporal resolution water level field Then, calculate at least three types of core indicators: 1. Flood control safety indicators: such as calculating the highest water level at each key section, the cumulative duration of water levels exceeding the warning line, and the comprehensive regional flood control safety index. wait.

[0278] 2. Drainage efficiency indicators: such as calculating the total drainage volume of the watershed, the average drainage intensity of the region, and the energy consumption per unit of drainage volume, to form a set of drainage efficiency evaluation indicators. .

[0279] 3. Regional synergy indicators: for example, by analyzing the regional interaction intensity matrix. The study quantifies the timing and intensity of positive collaboration (such as upstream companies staggering peak flows for downstream companies to alleviate downstream pressure) and negative conflict (such as upstream companies rushing to discharge waste downstream, increasing the burden on downstream companies), and calculates the collaboration index. .

[0280] Furthermore, a weighted comprehensive evaluation model, such as... Calculate an overall synergy score. .

[0281] For example, a specific implementation process is provided, including: 1. Calculation of flood control safety indicators: (1) Highest water level Extract the highest water level values ​​at each key control section throughout the entire flood process; (2) Duration of exceeding the warning level : Calculate the cumulative duration for which the water level at each cross-section is higher than its warning level; (3) Flood Risk Index : By integrating the water level and time of the section exceeding the warning level, for example The summation covers all periods when water levels exceed warning levels.

[0282] 2. Evaluation of drainage efficiency indicators: (1) Total drainage volume : Perform time integration on the flow curve of the optimal strategy. .

[0283] (2) Energy consumption indicators Based on the power curve of the pumping station, calculate the estimated energy consumption of the entire drainage process. ,in For pump station power, This refers to the runtime.

[0284] 3. Regional Cooperative Quantitative Analysis: Systematically analyzes and calculates the regional interaction intensity matrix. .

[0285] (1) Positive collaborative events: Identify For periods with significantly negative values, this represents the region. Flood control measures (such as pre-emptive drainage to free up reservoir capacity) effectively mitigated the impact on downstream areas. The pressure of flood control.

[0286] (2) Negative conflict events: Identify For periods with significantly positive values, this represents the region. The drainage work exacerbated the flooding in the downstream areas. The burden.

[0287] (3) Coordination index Quantification is achieved by statistically analyzing the ratio of the total duration of positive collaborative events to the total duration of their impact, for example... .

[0288] Furthermore, the overall synergy score can be calculated using a weighted comprehensive evaluation model. .For example, ,in, , , The weights of each sub-indicator; , , The scores are for the normalized security, efficiency, and coordination.

[0289] Furthermore, all strategies, simulation results, and evaluation indicators can be integrated to generate a comprehensive visual evaluation report that includes water level process animation, risk distribution heat map, strategy comparison Gantt chart, and specific decision-making recommendations, for decision-makers to use.

[0290] For example, a visual comprehensive evaluation report should include at least: 1. Water level process animation: Dynamically displays the process of water level field evolution in the entire river network over time and space under the optimal strategy.

[0291] 2. Risk Heat Map: On the river network geographic information map, the light and dark colors represent the flood risk level or the degree of exceeding the warning level in different areas at different times.

[0292] 3. Strategy Gantt Chart: Clearly displays the start-up and shutdown status and operating power of each pump station at different time periods.

[0293] 4. Impact Propagation Path Diagram: The regional interaction impact intensity matrix is ​​visualized as a network diagram, with nodes representing decision-making regions. The thickness, color, or arrows of the edges indicate the intensity, direction, and time delay of the impact, intuitively revealing the interaction relationships between regions.

[0294] Furthermore, through the aforementioned evaluation and output mechanism, not only can an optimal strategy be provided, but also a clear and comprehensive answer can be given regarding the advantages of the strategy, its risks, costs, and benefits, thereby greatly improving the transparency and scientific nature of decision-making.

[0295] In some alternative implementations, after step S203 above, the method includes: Step e1: Based on the real-time target drainage strategy time series and the actual operating status of the pumping station units, construct an optimization problem that satisfies engineering constraints.

[0296] Step e2: Based on the optimization problem, construct a mixed integer programming model.

[0297] Step e3: Solve the mixed integer programming model and generate a sequence of pump station control commands.

[0298] In one alternative embodiment, the pump station control command sequence is used to control the operation of the pump station.

[0299] In one alternative embodiment, the optimal drainage strategy In other words, the real-time target drainage strategy time series is a descriptive region. exist There should always be a function of the drainage flow rate, but actual pumping station units usually only have a limited number of operating states (such as: shutdown, No. 1 pumping unit on, No. 2 pumping unit on, No. 1+2 pumping units on, etc.), and have engineering constraints such as minimum continuous operating time and minimum shutdown interval.

[0300] Therefore, the problem is transformed into: finding a practical pumping station operation scheme that makes the total drainage flow process similar to the theoretically optimal flow process. It is the closest and satisfies all engineering constraints.

[0301] Furthermore, the optimization problem satisfying the engineering constraints described above is modeled as a mixed-integer programming (MIP) problem, i.e., the mixed-integer programming model: 1. Decision variables: for each pumping station Each unit In each discrete time period (For example, every 15 minutes), define a binary decision variable. .further, This indicates that the unit was started during that time period; This indicates that the window is closed.

[0302] 2. Objective Function: The objective is to minimize the deviation between the actual drainage flow and the target flow. A typical objective function is to minimize the sum of squared errors, as shown in the following equation (17): (17) In the formula: Indicates pumping station unit The rated drainage capacity.

[0303] 3. Constraints: (1) Minimum operating time constraint: If a unit is in If it starts at any time, then... Decision variables over a time period It must be 1, and can be expressed by introducing additional auxiliary variables and linear inequalities; (2) Minimum downtime constraint: If a unit is in If the machine is constantly shut down, then... Decision variables over a time period It must be 0; (3) Other electrical or hydraulic constraints.

[0304] Furthermore, mature open-source optimization solvers (such as Gurobi, CPLEX, or CBC) can be used to solve the mixed-integer programming model. The solver will then provide a set of optimal binary decision variables. The value of is then calculated. Finally, this optimal 0-1 matrix is ​​post-processed to convert it into a sequence of human-readable and machine-executable control instructions. For example, this can be achieved by scanning the time series of each unit. It identifies the start and end times of segments where the value is 1 consecutively.

[0305] Furthermore, the final output sequence of pump station control commands is in the format of a set of four tuples: {pump station ID, unit ID, start time, stop time}.

[0306] Furthermore, through the above processing, the complex game theory optimization results at the upper level can be successfully translated into specific and executable engineering operation instructions at the lower level, thus opening up a complete link from theoretical analysis to practical application and giving it high practical application value.

[0307] Furthermore, the generated pump station control command sequence can be directly sent to the pump station automation control system (such as SCADA system) for execution.

[0308] In some alternative implementations, after step S202 above, the method includes: Step f1: Based on the regional interaction influence intensity distribution matrix, a flood control and drainage influence propagation path map is generated through force-directed layout algorithm and visual element encoding.

[0309] In an optional embodiment, the regional interaction intensity distribution matrix is ​​used. Transform into a directed weighted graph with edge weights as The force-directed layout algorithm is used to generate node positions, and the thickness of the arrows represents the influence intensity and the color represents the propagation delay to construct an influence propagation path graph.

[0310] For example, based on the regional interaction intensity distribution matrix For each directed edge in a directed weighted graph Calculate a static weight that represents its strongest influence. .

[0311] Furthermore, a force-directed layout algorithm is used to determine the position of each region node in the directed weighted graph on the two-dimensional plane.

[0312] Specifically, the algorithm simulates a directed weighted graph as a physical system: nodes are treated as mutually repelling charges, while edges are considered as springs connecting the nodes. The spring constant can be related to its weights. Proportional. Furthermore, through iterative simulation, the algorithm eventually reaches a state of mechanical equilibrium, forming the node position distribution. This layout naturally clusters nodes with strong mutual influence together, while distant nodes are kept apart.

[0313] Further, visual element encoding is performed: 1. Edge thickness: related to weight Proportional, which intuitively represents the intensity of the influence.

[0314] 2. Edge color: Propagation delay can be represented using color gradations. For example, a gradual change from blue (short latency) to red (latency).

[0315] 3. Arrow: Indicates the direction of influence, from the area where the influence is applied to the area where it is affected.

[0316] In some alternative implementations, a hierarchical, pointer-based, or indexed sparse data structure is used instead of a large, dense uniform array in order to efficiently store and access data with multiple different spatiotemporal resolutions.

[0317] Specifically, the data structure can adopt a quadtree (for two-dimensional river networks) or a hierarchical grid structure.

[0318] Furthermore, the base layer is a low-resolution global coarse grid data array covering the entire study area, storing the coarse grid calculation results at all times.

[0319] Furthermore, each cell in this coarse grid data structure contains a pointer or index. By default, this pointer is null.

[0320] Furthermore, when the physical region corresponding to a coarse grid cell is refined (becoming an influence region or impulse region), its pointer will point to a fine grid object. This object is an independent data structure containing its own spatial extent, a higher-resolution data array (storing the calculation results of the fine or ultra-fine grid), and its corresponding time step and other metadata.

[0321] Furthermore, when this data structure needs to be persistently stored, a format that supports complex and hierarchical data models can be used, such as HDF5 (Hierarchical Data Format 5). HDF5 can store the aforementioned hierarchical relationships and metadata information together in a single file, facilitating efficient data archiving, exchange, and retrieval.

[0322] The cross-regional flood control and drainage coupled simulation method provided in this embodiment has the following beneficial effects: 1. By constructing an asynchronous game-theoretic decision-making mechanism based on physical propagation delay, the problem of physical causal time-series distortion caused by the use of synchronous or quasi-steady-state assumptions in the background technology is solved. Specifically, this mechanism constructs a unique asynchronous decision-making time sequence for each decision-making region, ensuring physical causality, through the calculated flood wave propagation delay matrix. This means that the decision-making behavior of the downstream region is forcibly placed after the physical arrival time of the impact of the decision-making behavior of its upstream region. This method of binding abstract game theory rounds with specific hydrodynamic propagation times makes the entire collaborative decision-making process completely synchronized with the real laws of the physical world. Its beneficial effects are that it can effectively avoid decision-making conflicts caused by information lag. For example, it can prevent downstream pumping stations from starting prematurely due to incorrect predictions before the actual arrival of the upstream flood peak, thereby saving a lot of energy. More importantly, it eliminates the dangerous situation of the downstream region passively accepting the flood peak when there is insufficient preparation for upstream flood discharge, and wins a valuable and accurate physical time window for the downstream to take targeted countermeasures (such as pre-discharge capacity), greatly improving the scientificity and safety of cross-regional collaborative scheduling.

[0323] 2. By constructing a spatiotemporally adaptive hydrodynamic fine simulation engine, the inherent contradiction between simulation computational efficiency and accuracy in the background technology is resolved. This is achieved through the synergistic effect of a three-layer asynchronous adaptive grid and a dynamic activation mechanism. First, the system uses the impulse response function to pre-identify the pulse region and influence region that are severely affected by local behaviors such as pump station start-up and shutdown, and deploys high-resolution grids only in these regions. Second, the dynamic activation mechanism ensures that these computationally expensive fine grids are only activated and used for computation when they are needed to accurately capture shock waves or discontinuities. Compared with the traditional unified grid method, this method of intelligently and dynamically focusing computational resources on the most critical spatiotemporal regions of physical phenomena can reduce the overall computational load of the entire watershed by several orders of magnitude while ensuring accurate capture of key local hydraulic phenomena such as pump station outlet shock waves (meeting accuracy requirements). Its specific advantages in flood control and drainage scenarios are that it makes large-scale watershed-local fine-grained real-time coupled simulation, which was previously impossible due to excessive computational load, a reality, providing accurate and timely simulation basis for emergency dispatch decisions.

[0324] 3. By constructing a three-temporal dynamic benefit function that quantifies spatiotemporal responsibilities, the problem of beggar-thy-neighbor suboptimal decision-making caused by limitations in the setting of decision-making objectives in the background technology is solved. This is achieved through a complex benefit function structure that includes a current term, an upstream historical impact term, and a downstream future expected responsibility term. The historical impact term uses a propagation delay matrix to quantify the cost that the region needs to pay for the upcoming upstream flood; the future responsibility term uses an impulse response function to quantify the future flood control risks that the current drainage decision may cause to the downstream. This design places the behavior of each decision-making region in a spatiotemporal responsibility chain spanning the past, present, and future. Its advantage in cross-regional flood control scenarios is that it provides clear and calculable intrinsic incentives for cooperative and collaborative behavior. For example, if an upstream region chooses to forcibly drain water downstream, the future responsibility term in its benefit function will generate a huge negative utility, thereby inhibiting such self-interested behavior. Conversely, if it chooses to sacrifice some of its own interests for downstream peak shifting, its benefit function can also be compensated from the global benefits. This enables the system to automatically and intrinsically find a globally coordinated Nash equilibrium strategy, fundamentally improving the overall flood control safety redundancy and asset utilization efficiency of the basin.

[0325] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A cross-regional flood control and drainage coupled simulation method, characterized in that, The method includes: Obtain a standardized input dataset containing real-time hydrological monitoring data and meteorological forecast data; Based on the standardized input dataset and the preset river network topology model, a simulation of cross-regional hydrodynamic interaction effects is performed, and the regional interaction effect intensity distribution matrix and flood wave propagation delay matrix are calculated. Based on the regional interaction influence intensity distribution matrix and the flood wave propagation delay matrix, a delay-based asynchronous game model is used for strategy optimization, and a real-time target drainage strategy time series is generated.

2. The method according to claim 1, characterized in that, Obtain a standardized input dataset containing real-time hydrological monitoring data and meteorological forecast data, including: Acquire real-time hydrological monitoring datasets and meteorological forecast datasets; The forecast error index value is determined based on the real-time hydrological monitoring dataset and the meteorological forecast dataset; Using the forecast error index value, a dynamic fusion weight value characterizing the reliability of the forecast data is calculated; Based on the dynamic fusion weights, the real-time hydrological monitoring dataset and the meteorological forecast dataset are weighted and fused to obtain a fused rainfall input sequence; The standardized input dataset is generated based on the fused rainfall input sequence.

3. The method according to claim 1, characterized in that, Based on the standardized input dataset and the preset river network topology model, a cross-regional hydrodynamic interaction simulation is performed, and the regional interaction intensity distribution matrix and flood wave propagation delay matrix are calculated, including: Obtain the impulse response function matrix; Based on the impulse response function matrix, a three-layer asynchronous adaptive mesh is generated, comprising a global coarse mesh, an influence region fine mesh, and an impulse region ultrafine mesh. In the three-layer asynchronous adaptive grid, hydrodynamic calculations are performed using the standardized input dataset, resulting in calculations at multiple levels; The calculation results from the multiple levels are prioritized and fused to generate a multi-temporal and spatial resolution water level field; Based on the multi-temporal resolution water level field and the preset river network topology model, the regional interaction influence intensity distribution matrix and the flood wave propagation delay matrix are determined.

4. The method according to claim 3, characterized in that, Obtain the impulse response function matrix, including: By adding a pulse source term to the standard Saint-Venant hydrodynamic equations, a pulse-embedded hydrodynamic equations are obtained, where the pulse source term is used to simulate the instantaneous start-up behavior of a pumping station. The Green's function method is used to solve the impulsive embedded hydrodynamic equations and generate the impulsive response function matrix.

5. The method according to claim 4, characterized in that, The method further includes: The impulse response function matrix is ​​corrected using a nonlinear correction function, which is used to reflect the nonlinear effects of the actual cross-sectional morphology of the river channel.

6. The method according to claim 3, characterized in that, Based on the multi-temporal resolution water level field and the preset river network topology model, the regional interaction influence intensity distribution matrix and the flood wave propagation delay matrix are determined, including: Based on the multi-temporal resolution water level field, obtain the real-time water level dataset; Based on the real-time water level dataset and the preset river cross-section parameter library, a time-varying propagation velocity field is generated, which is used to reflect the real-time water conditions and river geometry. Based on the preset river network topology model, a decision-oriented directed graph is generated, wherein the flood wave propagation delay matrix has decision regions as nodes and the direction of water flow influence as directed edges; Using the reciprocal of the time-varying propagation velocity field as the path weight, the decision directed graph is processed by the shortest path search algorithm to obtain the target propagation path set; Construct the flood wave propagation delay matrix based on the target propagation path set; The impulse response function matrix and the historical target drainage strategy time series are convolved and integrated to determine the regional interaction intensity distribution matrix.

7. The method according to claim 1, characterized in that, Based on the regional interaction influence intensity distribution matrix and the flood wave propagation delay matrix, a delay-based asynchronous game model is used for strategy optimization, and a real-time target drainage strategy time series is generated, including: Using the flood wave propagation delay matrix, an asynchronous decision-making time series is determined. This asynchronous decision-making time series is used to ensure that the upstream decision-making influence of each decision-making region can physically propagate to the downstream before downstream decision-making is carried out. Obtain the set of time-varying revenue functions; Based on the asynchronous decision time series, and with the time-varying payoff function set as the optimization objective, the real-time target drainage strategy time series is generated by iteratively solving the Nash equilibrium.

8. The method according to claim 7, characterized in that, Obtain the set of time-varying revenue functions, including: Calculate the current item reflecting the security status of each decision-making area; Using the aforementioned flood wave propagation delay matrix, the upstream historical impact term is calculated; Based on the impulse response function matrix, future expected responsibilities are determined; Based on the regional interaction influence intensity distribution matrix, time-varying weight coefficients are dynamically configured for the current item, the upstream historical influence item, and the future expected responsibility item, and the time-varying revenue function set is generated.

9. The method according to claim 7, characterized in that, The method further includes: Using the real-time target drainage strategy time series as the historical target drainage strategy time series, a cross-regional hydrodynamic interaction simulation was performed again, and the resulting water level field was generated. The resulting water level field and the preset physical constraint set are checked, and it is identified whether there are any violations in the resulting water level field. When violations are found, the time series of the real-time target drainage strategy is iteratively corrected using a penalty function until no violations are found, and the time series of the real-time target drainage strategy is determined.

10. The method according to claim 9, characterized in that, The method further includes: Obtain the target multi-temporal resolution water level field corresponding to the time series of the real-time target drainage strategy; Based on the regional interaction influence intensity distribution matrix, the real-time target drainage strategy time series, and the target multi-temporal resolution water level field, multiple discrete index values ​​are calculated. Based on the aforementioned discrete index values, the overall synergy score is obtained through a weighted comprehensive evaluation model.

11. The method according to claim 7, characterized in that, The method further includes: Based on the time series of the real-time target drainage strategy and the actual operating status of the pumping station units, an optimization problem that satisfies engineering constraints is constructed. Based on the aforementioned optimization problem, a mixed-integer programming model is constructed; The mixed integer programming model is solved, and a pump station control command sequence is generated, which is used to control the operation of the pump station.

12. The method according to claim 1, characterized in that, The method further includes: Based on the regional interaction influence intensity distribution matrix, a flood control and drainage influence propagation path map is generated through force-directed layout algorithm and visual element encoding processing.

Images