Water conservancy multi-model coupling parallel architecture design method

Abstract

According to the water conservancy multi-model coupling parallel architecture design method, the whole water conservancy system is divided into a plurality of basic forecast sections, and the forecast sections are mapped according to the upstream and downstream topological relation of the forecast sections. Meanwhile, a plurality of calculation units are divided by utilizing drainage basin service characteristics and different models, and each unit corresponds to one calculation node except for mapping with a basic forecast section; the computing nodes are mutually connected through a high-speed network to form a distributed computing environment. Corresponding water conservancy model components such as a hydrological model, an engineering scheduling model, a hydrodynamic model and an incoming water forecasting model are deployed on each computing node, and the model components on different nodes can be combined and cooperatively work according to needs so as to realize comprehensive simulation of the whole water conservancy system.

Description

Technical Field

[0001] This invention relates to the field of computer vision technology, and in particular to a method for designing a multi-model coupled parallel architecture for hydraulic engineering. Background Technology

[0002] With the continuous development of water conservancy projects and the increasing demand for refined management of water conservancy systems, traditional stand-alone water conservancy models face problems such as low computational efficiency, difficulty in data sharing, and poor model scalability when simulating and analyzing large-scale and complex water conservancy systems. For example, when simulating complex problems such as flood evolution and water resource allocation in large river basins, stand-alone models often struggle to complete calculations within an acceptable timeframe and cannot fully utilize multi-source data for comprehensive analysis. Therefore, a new architecture and methodology are needed to achieve efficient integration and collaborative operation of water conservancy models, thereby improving the scientific rigor and accuracy of water conservancy project planning, design, and management.

[0003] Chinese invention patent application CN119830225A discloses a model fusion method, apparatus, and computer device based on topology. The method includes: constructing a topology; determining the objects to be fused deployed in each node of the topology, wherein the topology includes multiple nodes; according to a model fusion algorithm, merging and iterating the objects to be fused deployed in multiple nodes within a preset node range in the topology to obtain a local fusion model after each merging iteration of all nodes in the topology; when the coefficients of all local fusion models after multiple merging iterations of all nodes are consistent, determining that the local fusion models corresponding to each node have been merged, resulting in multiple multi-task models in the topology. The drawback of this method is that the model fusion algorithm performs merging iterations within a preset node range, and the determination of the node range lacks an adaptive mechanism. If the range is set unreasonably, it can lead to slow model convergence or even getting trapped in local optima, making it difficult to obtain the globally optimal fusion effect.

[0004] Chinese invention patent application CN119862008A discloses a multi-model scheduling architecture method based on LLM for multimodal input applications, particularly in the field of large-scale AI models. This method processes user input data by identifying and handling multiple modalities. The application implements the application and scheduling of multimodal models through text processing. The main processes include multimodal input identification and processing, text processing and related question output, resource type processing, and interactive experience optimization. Users can interact with the system in various natural ways without changing their behavioral habits, while simultaneously processing multiple types of input, improving interaction efficiency. However, this method has a drawback: by implementing multimodal model application and scheduling solely through text processing, the conversion and processing of non-textual modal information such as images, audio, and video may result in the loss of a significant amount of original information. This can lead to the model being unable to acquire complete and accurate multimodal features, affecting the accuracy and richness of the processing results. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes a multi-model coupled parallel architecture design method for water conservancy. This method divides the entire water conservancy system into multiple basic forecast sections, which are mapped according to their upstream and downstream topological relationships. Simultaneously, multiple computing units are created using the operational characteristics of the watershed and different models. In addition to mapping to the basic forecast sections, each unit corresponds to a computing node. These computing nodes are interconnected via a high-speed network, forming a distributed computing environment. Corresponding water conservancy model components, such as hydrological models, engineering scheduling models, hydrodynamic models, and inflow forecasting models, are deployed on each computing node. Model components on different nodes can be combined and work collaboratively as needed to achieve a comprehensive simulation of the entire water conservancy system.

[0006] The purpose of this invention is to provide a method for designing a coupled parallel architecture for multi-model hydraulic engineering, including basic forecast section collection and mapping, and further including the following steps:

[0007] Step 1: The computing unit collects the mapping;

[0008] Step 2: Mapping the calculation unit to the basic forecast section;

[0009] Step 3: Taking the basic forecast section as the core, configure different models and parameter information required for different sections according to flood control needs or water resources to form a calculation scheme;

[0010] Step 4: Using the constructed calculation scheme, configure various data such as time periods and meteorological monitoring to form different scenario session plans;

[0011] Step 5: Analyze the constructed session plan, allocate various computing units to various nodes, and allocate and call physical resources according to the computing plan of the session;

[0012] Step 6: Visualize the results of the calculated sessions.

[0013] Preferably, the basic forecast section collection mapping includes the following sub-steps:

[0014] Step 01: Collection and analysis of basic watershed data;

[0015] Step 02: Preliminary selection of cross sections based on forecast requirements, including flood forecast cross sections, water resource management cross sections, and ecological flow guarantee cross sections;

[0016] Step 03: Cross-sectional representativeness and feasibility assessment;

[0017] Step 04: Based on the model simulation, optimize and verify the cross-section to determine the final basic cross-section, which includes at least the cross-section location and basic parameters.

[0018] In any of the above solutions, step 1 preferably includes the following sub-steps:

[0019] Step 11: Data collection and preprocessing for computing units;

[0020] Step 12: Calculate the characteristics of the calculation unit;

[0021] Step 13: Constructing the topology of computing units;

[0022] Step 14: Mapping the boundaries of computational units based on topological relationships;

[0023] Step 15: Define the mapping rules.

[0024] In any of the above schemes, step 11 preferably includes obtaining computing unit data from different models, using data cleaning algorithms to remove outliers and missing values, identifying and correcting outliers in the flow data through box plots, and using linear interpolation to fill in missing values ​​in the water level data.

[0025] The computational unit data includes raster cell data from distributed hydrological models, sub-basin cell data from conceptual hydrological models, and channel cross-section or grid cell data from hydrodynamic models.

[0026] In any of the above solutions, step 12 preferably includes the following sub-steps:

[0027] Step 121: Calculate the geometric properties of each computational unit;

[0028] Step 122: Analyze the resolution and coverage of the grid cells to determine their accuracy in describing the spatial characteristics of the watershed;

[0029] Step 123: Use the watershed drainage system extraction algorithm to determine the catchment area and boundary morphology of the sub-watershed unit;

[0030] Step 124: Through spatial overlay analysis, compare the spatial overlap of computational units of different models to identify potential data conflict areas.

[0031] In any of the above solutions, step 13 preferably includes the following sub-steps:

[0032] Step 131: Construct the water system topology. Based on the watershed water system, use the river network tracing algorithm of GIS to construct the upstream and downstream topological relationships of the river channel, assign a unique identifier to each river channel cross-section unit, and establish a topology table to record the connection relationships between the units.

[0033] Step 132: Construct the watershed unit topology. Based on the terrain slope and water flow direction, determine the confluence relationship between sub-watershed units. Using ArcGIS's hydrological analysis tools, generate a flow direction raster map of the watershed, then delineate the sub-watershed boundaries and establish a topological network between sub-watersheds. In the distributed raster model, based on the water flow direction of adjacent rasters, construct the local topology of the raster units and clarify the transmission path of water flow between rasters.

[0034] In any of the above solutions, step 2 preferably includes the following sub-steps:

[0035] Step 21: Spatial Overlay Analysis;

[0036] Step 22: Distance calculation and analysis;

[0037] Step 23: Topological relationship analysis.

[0038] Preferably, in any of the above schemes, the construction of the computation scheme includes the following sub-steps:

[0039] Step 31: Create and name the calculation scheme;

[0040] Step 32: By selecting the start and end basic forecast sections, the system will automatically filter and determine the calculation area of ​​the calculation scheme based on their topological relationships;

[0041] Step 33: Within the calculation area, list the basic forecast sections and select a specific basic forecast section;

[0042] Step 34: For this basic forecast section, select various models that are mapped to this section;

[0043] Step 35: After determining the model of the basic forecast section, select the specific model parameters for that model;

[0044] Step 36: In the initial state of the basic forecast section, configure the hydrological model and parameters, select the reservoir model and parameters, and mark it as post-processing;

[0045] Step 37: Configure each basic forecast section in sequence according to the topological relationship, and save the calculation scheme.

[0046] In any of the above solutions, step 32 preferably includes the following sub-steps:

[0047] Step 321: Select the basic forecast section;

[0048] Step 322: Analyze the topological relationships;

[0049] Step 323: Determine the calculation area.

[0050] In any of the above schemes, step 322 preferably includes retrieving pre-constructed watershed channel topology data, storing the upstream and downstream connection relationships of the river in the form of a directed graph, and using breadth-first search in graph theory to search along the topology path from the starting section until the ending section is reached, including all river segments, sub-basins and other areas involved in the path into the calculation area.

[0051] In any of the above solutions, step 5 preferably includes the following sub-steps:

[0052] Step 51: Start the scheduling process;

[0053] Step 52: Execute the node parsing task;

[0054] Step 53: Information Reading and Calculation Preparation;

[0055] Step 54: Planning and Calculation Logic;

[0056] Step 55: Determine boundary information;

[0057] Step 56: Start and execute the computing unit;

[0058] Step 57: Monitor and update the status.

[0059] In any of the above schemes, step 53 preferably includes regional command parsing, rainfall command parsing, and station and model parameter command parsing.

[0060] In any of the above schemes, step 53 preferably includes determining the boundary of the hydrological calculation unit, the boundary of the reservoir scheduling unit, and the boundary of the hydrodynamic calculation unit.

[0061] This invention proposes a design method for a coupled parallel architecture of multiple water conservancy models, which realizes efficient integration and collaborative work of water conservancy models, and improves the scientificity and accuracy of water conservancy project planning, design and management. Attached Figure Description

[0062] Figure 1 This is a flowchart of a preferred embodiment of the hydraulic multi-model coupled parallel architecture design method according to the present invention.

[0063] Figure 2 This is a flowchart illustrating an embodiment of the computational scheme constructed according to the hydraulic multi-model coupled parallel architecture design method of the present invention.

[0064] Figure 3 This is a flowchart of an embodiment of the field-based scheme-based model parallel computation of the hydraulic multi-model coupled parallel architecture design method according to the present invention. Detailed Implementation

[0065] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0066] Example 1

[0067] like Figure 1 As shown, a design method for a multi-model coupled parallel architecture of water conservancy is presented. Step 1000 involves collecting and mapping basic forecast cross sections. Due to the different characteristics of various water conservancy models, the granularity of modeling support varies (some models use small watershed channels and nodes as granularity, while others use regional outlet points as granularity). In order to support the dynamic adjustment and hybrid construction of various hydrological models in different regions within the watershed, a unified basic cross section needs to be constructed.

[0068] The construction of basic cross-sections is divided into four stages: preliminary data collection and analysis, initial selection of cross-sections based on forecast requirements, assessment of cross-section representativeness and feasibility, and optimization and final determination of cross-sections. It includes the following sub-steps:

[0069] Step 1010, watershed basic data collection and analysis, includes the following sub-steps:

[0070] Perform step 1011, watershed basic data collection.

[0071] Topographic data within the watershed, including a high-precision digital elevation model (DEM), will be collected. Geographic information system software such as ArcGIS will be used to delineate watershed boundaries and extract river systems, clarifying basic information such as watershed area and river network distribution. Simultaneously, underlying surface data, including soil type and vegetation cover, will be acquired, as these data will influence precipitation runoff processes.

[0072] Historical hydrological and meteorological data were collected, covering meteorological elements such as precipitation, evaporation, temperature, and wind speed over many years, as well as data on flow, water level, and sediment content from various hydrological stations. Data analysis tools such as Python and R were used to perform statistical analysis on the historical data, and precipitation-runoff and water level-flow curves were plotted to understand the hydrological characteristics and changing patterns of the watershed.

[0073] Step 1012: Review of water conservancy projects and monitoring networks.

[0074] A detailed analysis of the water conservancy facilities within the basin is required, including the location, scale, and operation and scheduling rules of reservoirs, dams, and pumping stations. For reservoirs, key parameters such as their storage capacity curves and flood discharge capacity curves need to be obtained; for dams, their control methods and flow capacity need to be understood.

[0075] Determine the distribution of existing hydrological monitoring stations, including the location, monitoring period, and data quality of rain gauges, water level stations, and flow stations. Analyze the coverage and representativeness of the monitoring stations to determine if there are any monitoring blind spots or weaknesses.

[0076] Step 1020 involves preliminary selection of cross-sections based on forecast requirements. These cross-sections include flood forecasting cross-sections, water resource management cross-sections, and ecological flow guarantee cross-sections.

[0077] 1. Selection of flood forecast sections

[0078] In key flood control areas, such as cities, densely populated areas, and areas surrounding important infrastructure, cross-sections that effectively reflect the flood's evolution should be selected. These are typically river sections with straight channels, regular cross-sections, and no significant backwater or diversion. Forecast cross-sections are usually set up at an appropriate distance upstream of the flood protection zone to provide early warnings. For example, a cross-section might be set up 10-20 kilometers upstream of a city in a stable river section, allowing sufficient response time for urban flood control based on historical flood propagation times.

[0079] Meanwhile, considering the connection with water conservancy projects such as reservoirs and dams, cross-sections are set up at appropriate locations downstream of the reservoir to monitor the impact of reservoir flood discharge on the downstream river channel; cross-sections are set up upstream and downstream of the dam to monitor the changes in water level and flow before and after dam operation.

[0080] 2. Selection of water resource management sections

[0081] To address water resource management needs, forecasting sections are established at key water intakes, areas with concentrated water users, and transboundary rivers. Sections upstream of water intakes forecast inflow volumes to ensure water supply security; sections along transboundary rivers monitor changes in water quantity and quality, providing data support for water resource allocation and coordination. For example, sections are established at the confluence of inter-provincial rivers to monitor water resources in real time and prevent water resource disputes.

[0082] 3. Selection of ecological flow protection sections

[0083] In ecologically sensitive areas, such as wetlands, fish spawning grounds, and habitats of rare aquatic organisms, cross-sections that reflect ecological water needs should be selected. Combining ecological protection goals and ecological flow standards, cross-sections should be established in key river sections to ensure that the river's ecological base flow meets requirements. For example, a cross-section could be established upstream of a wetland reserve; by forecasting flow, water resources could be rationally allocated to maintain the stability of the wetland ecosystem.

[0084] Step 1030, the assessment of cross-sectional representativeness and feasibility, includes the following sub-steps:

[0085] Step 1031: Hydrological Representativeness Assessment

[0086] Using collected historical hydrological data, mathematical statistical methods, such as correlation analysis and frequency analysis, are employed to evaluate the correlation between the initially selected cross-section and upstream and downstream hydrological station data. The correlation coefficients between the cross-section and the flow and water levels of surrounding hydrological stations are calculated; a higher correlation coefficient indicates better hydrological representativeness of the cross-section. If the correlation coefficient is below 0.7, the cross-section location needs to be reconsidered, or measures such as data interpolation and extension should be taken to improve representativeness.

[0087] Analyze the flow pattern at the cross-section to determine whether it is a stable and uniform flow. For river sections with bends, branching, or dramatic changes in scouring and deposition, careful selection is required, or adverse effects should be eliminated through model correction or other means.

[0088] Perform step 1032, project feasibility assessment.

[0089] Consider the engineering conditions for setting up the cross-section, including the terrain, accessibility, and construction difficulty. Prioritize locations with flat terrain, convenient transportation, and ease of construction and maintenance of monitoring facilities. Avoid setting up cross-sections in areas with complex geological conditions, high construction difficulty, and high maintenance costs.

[0090] Assess the interaction between the cross-section and existing water conservancy projects and monitoring facilities to ensure that the cross-section setting will not interfere with the operation of water conservancy projects and the collection of monitoring data. For example, the cross-section should not be set in an area of ​​severe scouring downstream of a reservoir dam, so as not to affect the stability of monitoring equipment and the accuracy of data.

[0091] Step 1040 is executed to optimize and verify the cross-section based on model simulation, and to determine the final basic cross-section, which includes at least the cross-section location and basic parameters.

[0092] 1. Cross-section optimization based on model simulation

[0093] Hydrological forecasting models, such as the Xin'anjiang model, API model, and spatiotemporal variable source model, were used to simulate and analyze the initially selected cross sections. By adjusting the cross section location and parameters, the degree of fit between the forecast results and measured data under different schemes was compared, and the cross section location and parameter combination with the best fit was selected.

[0094] Sensitivity analysis was conducted using a model to determine the impact of changes in the cross-sectional location on the forecast results. For cross-sections with high sensitivity, their locations were further optimized to ensure the reliability and stability of the forecast results.

[0095] 2. Determine the final cross-section through multi-party collaboration.

[0096] Experts in hydrology, water conservancy engineering, and ecological environment were organized to conduct discussions and to comprehensively evaluate the cross-section setting scheme based on their opinions and suggestions from different disciplines.

[0097] After multiple rounds of demonstration and optimization, the final basic cross-section was determined, including the cross-section location, basic parameters, etc., to provide a basis for subsequent hydrological forecasting work.

[0098] Based on the unified basic cross-sections constructed, it is necessary to spatially overlay them with the rivers using geographic information system software such as ArcGIS, and match and set the topological relationships of various basic cross-sections according to the hierarchical topological relationships of the river channels within the basin.

[0099] Step 1100 involves collecting and mapping computational units from various models. Based on the characteristics and topological relationships of the computational units in different models, the boundaries between these units are mapped and managed. This includes the following sub-steps:

[0100] Step 1110 involves data collection and preprocessing within the computing unit, including:

[0101] 1. Multi-source data acquisition

[0102] Computational unit data is obtained from various models, including raster cell data from distributed hydrological models (such as spatiotemporal variable sources), sub-basin cell data from conceptual hydrological models (such as the Xin'anjiang model), and channel cross-section or grid cell data from hydrodynamic models. Data is collected through model output interfaces and database reading. For example, the Python pandas library is used to read sub-basin area, topographic slope, and other attribute data from CSV format output files, and the GDAL library is used to read spatial data such as raster cell elevation and soil type.

[0103] 2. Data cleaning and standardization

[0104] Data cleaning algorithms were employed to remove outliers and missing values. Outliers in flow data were identified and corrected using box plots, and missing values ​​in water level data were filled using linear interpolation. Data formats were standardized by converting various computational unit data into standard formats supported by Geographic Information Systems (GIS) (such as Shapefile and GeoJSON), and attribute field naming was standardized to ensure compatibility between different model data.

[0105] Execute step 1120 to calculate the unit characteristic analysis, which includes the following sub-steps:

[0106] Perform step 1121, using GIS spatial analysis tools to calculate the geometric attributes of each calculation unit, such as area, perimeter, centroid coordinates, etc.

[0107] Perform step 1122 to analyze the resolution and coverage of the grid cells and determine their accuracy in describing the spatial characteristics of the watershed.

[0108] Execute step 1123, for each sub-basin unit, use the watershed drainage system extraction algorithm to determine its catchment area and boundary morphology;

[0109] Step 1124 is executed to identify potential data conflict areas by comparing the spatial overlap of computational units in different models through spatial overlay analysis.

[0110] Execute step 1130 to construct the topology relationship of the computational unit, which includes the following sub-steps:

[0111] Step 131: Constructing the water system topology

[0112] Based on the watershed system, the river network tracing algorithm (D8 algorithm) in GIS is used to construct the upstream and downstream topological relationships of the river channels. A unique identifier is assigned to each river cross-section unit, and a topological table is established to record the connections between units, for example, by identifying upstream and downstream relationships through start and end node numbers. For areas with bifurcated channels, a directed acyclic graph (DAG) structure from graph theory is used to describe complex water flow paths.

[0113] Step 132: Constructing the watershed unit topology

[0114] For sub-basin units, the confluence relationships between units are determined based on topographic slope and water flow direction. Using ArcGIS's hydrological analysis tools, a flow direction raster map of the basin is generated, which is then used to delineate sub-basin boundaries and establish a topological network between sub-basins. In the distributed raster model, the local topology of the raster units is constructed based on the water flow direction of adjacent rasters, clarifying the transmission paths of water flow between rasters.

[0115] Step 1140 involves mapping the boundaries of computational units based on topological relationships. A direct mapping relationship is established between the boundaries of computational units based on the constructed topological network. Between upstream and downstream sections of the river channel, the calculated results of flow rate and water level at the upstream section are directly used as the input boundary conditions for the downstream section. For the connection between a sub-basin and the river channel, the outlet flow of the sub-basin is mapped to the corresponding river channel section unit. Efficient data transfer between computational units is achieved through the connection information in the topology table.

[0116] Perform step 1150 to formulate mapping rules; formulate detailed boundary mapping rules to clarify the data conversion and transfer methods between different types of computing units. For example, stipulate that when a sub-basin overlaps with a river cross-section, the hydraulic parameters of the river cross-section shall prevail; for cross-boundary computing units, data conflicts shall be handled according to the weight allocation principle. Compile the mapping rules into a configuration file for easy modification and maintenance later.

[0117] Step 1200 involves mapping the computational unit to the basic forecast transect. This mapping associates the computational unit with the basic forecast transect based on the physical spatial location of the transect. This includes the following sub-steps:

[0118] Perform step 1210, Spatial Overlay Analysis; using the spatial overlay function of GIS, overlay the calculation unit layer with the basic forecast section layer. Taking a raster calculation unit as an example, when the basic forecast section is located within a raster, the attribute information of that raster can be directly obtained through overlay analysis; for polygonal sub-basin calculation units, if the basic forecast section falls within the sub-basin area, the spatial intersection relationship between the two can be determined using the intersection analysis tool. In ArcGIS, the "Intersect" tool can be used to achieve layer intersection and obtain the detailed attributes of the calculation unit where the section is located.

[0119] Step 1220 involves distance calculation and analysis. When the basic forecast section cannot directly fall within the calculation unit, a distance calculation method is used to determine its correlation. The distance between the basic forecast section and the centroid (or boundary) of each calculation unit is calculated using the Euclidean distance formula in planar coordinates. Based on the principle of closest proximity, the basic forecast section is correlated with the nearest calculation unit. In Python, the Geopy library can be used to conveniently calculate the spherical distance between two points, thus implementing distance analysis functionality.

[0120] Perform step 1230, topology analysis; construct the topology relationship between the calculation unit and the basic forecast cross-section. For river-type calculation units and forecast cross-sections, determine their topological connections based on the upstream and downstream relationships of the river. Using the network analysis function of GIS, treat the river calculation units and forecast cross-sections as network nodes. By tracing the water flow direction, clarify the position of each cross-section in the calculation unit network and establish upstream and downstream topological connections. By setting river nodes and connection relationships, construct a complete river topology network to achieve topological mapping between cross-sections and calculation units.

[0121] Step 1300 involves using the basic forecast cross-section as the core, and configuring different models and parameter information for different cross-sections based on flood control needs or water resources to form a calculation scheme. For example... Figure 2 As shown, it includes the following sub-steps:

[0122] Execute step 1310 to create and name a calculation scheme; call the scheme creation function module through the user interface or API interface. The system backend uses Unique Identifier (UUID) generation technology to assign a globally unique ID to each newly created scheme, ensuring the scheme's uniqueness and identifiability in the database. Users enter the scheme name in the interactive interface; the name must follow specific naming conventions, such as the format "basin name + business type + creation time" (e.g., "Yongding River Basin Flood Control Dispatch Scheme_20250516"), facilitating scheme classification, management, and retrieval. After the scheme is created, the system establishes a corresponding scheme data table in the database, initializing basic information fields such as scheme ID, name, creator, and creation time.

[0123] Step 1320, Selecting the Region: By selecting the start and end basic forecast sections, the system will automatically filter and determine the calculation region of the calculation scheme based on their topological relationships; this includes the following sub-steps:

[0124] Perform step 1321 to select the basic forecast section; in the Geographic Information System (GIS) visualization interface, the user selects the start and end of the basic forecast section by clicking with the mouse or entering the section number. The system utilizes the spatial query function of GIS to accurately locate the selected section in the watershed drainage layer based on the section's coordinate information.

[0125] Step 1322 involves analyzing the topological relationships. The system retrieves pre-constructed watershed channel topological relationship data, generated based on the D8 algorithm and stored in the form of a directed graph to show upstream and downstream connections. Using breadth-first search (BFS) in graph theory, the system starts from the initial cross-section and searches along the topological path until the final cross-section is reached, incorporating all river segments and sub-basins involved in the path into the computational domain. During this process, the branching and merging of the river are identified through node connection information in the topology table, ensuring that the computational domain fully covers the required area.

[0126] Perform step 1323 to determine the calculation area; utilize the spatial analysis functions of GIS to perform geometric operations on the searched area, generating polygon vector data for the calculation area. Visualize the area's extent and perform boundary checks using its geometric attributes such as area and perimeter. If omissions or errors are found, users can manually adjust the cross-section selection and re-determine the calculation area.

[0127] In step 1330, within the calculation area, the basic forecast sections are listed, and a specific basic forecast section is selected. Based on the scope of the calculation area, the system uses a spatial query to filter all section data located within the basic forecast section database. This section data is displayed on the user interface in a list or tree structure, with each section entry containing basic information such as section name, number, geographical location, and watershed. The user selects a basic forecast section from the list using mouse clicks, keyboard selection, or the search function. The system then obtains the section's unique identifier (ID) for subsequent operations.

[0128] Execute step 1340, for the basic forecast section, select various models mapped to the section, such as hydrological models, water resource models, scheduling models, etc.

[0129] 1. Model library call

[0130] The system connects to a model library, which stores metadata for various hydraulic models, including model name, functional description, applicable scenarios, and input / output data formats. Through API interfaces or database queries, a list of models suitable for flood control operations can be retrieved, such as the Xin'anjiang hydrological model and reservoir scheduling rule models.

[0131] 2. Model Recommendation and Selection

[0132] A rule-based recommendation algorithm is used, combining information such as topography, climate conditions, and watershed characteristics of the cross-section area to recommend suitable models to users. For example, for cross-sections in humid regions, the Xin'anjiang model is recommended first. Users can select the model to be mapped to the cross-section from the model list based on their needs and the recommended information, using interactive methods such as checking boxes and dragging. The system records the ID and related configuration information of the selected model.

[0133] Execute step 1350, after determining the model for the basic forecast section, select the specific model parameters, including:

[0134] 1. Loading parameter templates

[0135] The system loads the corresponding parameter template from the parameter template library based on the selected model. The parameter template contains information such as the name, type, unit, value range, and default value of all parameters in the model. For example, the parameter template for the Xin'anjiang model includes parameters such as the watershed water storage capacity curve (B value) and evapotranspiration calculation parameters.

[0136] 2. Parameter selection and adjustment

[0137] Users can select and adjust parameters in the following ways:

[0138] Empirical values ​​were obtained by referencing existing research findings under similar watershed conditions, topography, and climate. The recommended values ​​in the database were compiled through literature reviews and summaries of historical projects.

[0139] Automatic calibration: Parameters are automatically calibrated using historical measured data and optimization algorithms (such as genetic algorithms and particle swarm optimization algorithms). The objective function is to minimize the error between measured flow and water level data and model simulation results (such as root mean square error RMSE and Nash efficiency coefficient NSE). The optimal parameter combination is automatically searched through iterative algorithm calculations.

[0140] Manual adjustment: Users can manually modify parameter values ​​based on their professional knowledge and practical experience. The system provides a parameter sensitivity analysis tool, which displays the changes in model output results in real time by changing the value of a single parameter, helping users determine the impact of parameter adjustments on simulation results and assisting them in parameter optimization.

[0141] 3. Parameter Verification

[0142] The adjusted parameters are verified using the measured data from the reserved verification period. The error index between the simulation results and the measured data is calculated. If the error is within an acceptable range (e.g., NSE ≥ 0.7), the parameter selection is confirmed; otherwise, the parameters are readjusted until the accuracy requirements are met.

[0143] In step 1360, multiple models may be coupled at a certain cross-section. For example, if the flow result of the hydrological model at this cross-section is used as the inflow of the reservoir scheduling model, it means that in addition to setting the model parameters normally, post-processing configuration is also required for this cross-section. That is, in the initial state of this cross-section, the hydrological model and parameters are configured first, and then the reservoir model and parameters are selected and marked as post-processing.

[0144] 1. Coupling Relationship Identification

[0145] When a user selects multiple models, the system automatically identifies model coupling relationships based on the logical relationships between models and the data flow direction. Using a predefined model coupling rule base, it determines whether the output data of the hydrological model (such as flow rate) can be used as input data for the reservoir scheduling model. If the conditions are met, a coupling relationship is determined between the two.

[0146] 2. Post-processing configuration

[0147] Following the coupling sequence, the parameters of the hydrological model are configured and validated first. Then, the reservoir scheduling model is selected and its parameters are set. In the parameter configuration interface, a dedicated post-processing marker field is set to mark the reservoir scheduling model as a post-processing model, and a data transfer mapping relationship is established, clarifying the correspondence between the flow data fields output by the hydrological model and the inflow parameters of the reservoir scheduling model. The system records the execution order and data transfer rules of the coupled models to ensure that data interaction and model calculations are performed according to the correct process during the computation.

[0148] Execute step 1370, configure each basic forecast section sequentially according to the topological relationship, and save the calculation scheme.

[0149] The system traverses all basic forecast sections within the calculation area, checking the completeness and correctness of the model configuration, parameter settings, and coupling relationships for each section. Data validation algorithms verify that parameter values ​​are within reasonable ranges and that model input and output data formats match. If any sections are found to be incompletely configured or incorrect, the system provides prompts to guide the user in making corrections. Once all sections are configured correctly, the system packages all information about the calculation scheme (including basic scheme information, calculation area data, model configurations for each section, parameter values, coupling relationships, etc.) and stores it in the database in formats such as JSON and XML. A scheme version number is also generated for subsequent scheme management, retrieval, and version tracking.

[0150] Step 1400 involves using the constructed calculation scheme to configure various data such as time periods and meteorological monitoring to form different scenario-based session plans.

[0151] Initialization settings: Create a session task, enter the forecast start time, end time and time step, and complete the time dimension configuration.

[0152] Rainfall data loading: Retrieve the compiled hourly rainfall data from the data center, and distribute the areal rainfall data to each computing unit according to the watershed grid or sub-watershed division.

[0153] Calculation scheme invocation: Based on the watershed characteristics and flood control requirements, select a calculation scheme from the calculation scheme library and load the corresponding hydrological model, hydrodynamic model and parameter configuration file.

[0154] Step 1500 involves parsing the constructed session plan, allocating various computing units to different types of nodes, and then allocating and calling physical resources according to the session's computing plan. Figure 3 As shown, it includes the following sub-steps:

[0155] Execute step 1510 to initiate the scheduling job process. The system receives computation scheduling instructions via the RabbitMQ message queue or API interface. The instruction format follows a predefined protocol standard (such as JSON or XML format). When an instruction arrives, the system triggers the scheduling job process, first validating the instruction by matching key fields in the instruction (such as scheme ID, computation time range, job type, etc.) using regular expressions. If the instruction format is incorrect or key information is missing, an error message is returned. If the validation passes, the instruction is stored in the task queue, and a unique job ID is generated for subsequent process tracking and management.

[0156] Execute step 1520 to perform the node parsing task.

[0157] 1. Regional Command Parsing

[0158] Using the spatial query function of Geographic Information System (GIS), based on the topological relationship of the basic forecast cross sections and the start and end basic cross sections of the scheme area, the specific geographic boundary information is retrieved from the regional database by using SQL statements combined with spatial functions (such as ST_Contains).

[0159] 2. Analysis of Rainfall Commands

[0160] For rainfall commands, regular expressions are used to extract key parameters such as rainfall time range and rainfall intensity. If the command includes a rainfall data file path (e.g., "hourly rainfall data in .csv format"), the data is loaded using a file reading function (e.g., the `read_csv` function in the pandas library of Python), and format conversion and quality checks are performed to remove outliers (e.g., data with negative rainfall). For complex rainfall commands (e.g., "72-hour rainfall forecast based on numerical weather prediction"), a meteorological data interface is called to obtain the rainfall grid data output by the numerical model. Interpolation algorithms (e.g., inverse distance weighted interpolation, Kriging interpolation) are used to convert the grid data into areal rainfall data applicable to each computational unit within the watershed.

[0161] 3. Site and Model Parameter Command Parsing

[0162] The site instruction parsing process combines string matching with database queries to extract the site number or name from the instruction and then retrieves the corresponding latitude and longitude coordinates, monitoring equipment type, data acquisition frequency, and other information from the site information database. For model and parameter instructions, a hierarchical parsing strategy is employed: first, the model type (e.g., Xin'anjiang hydrological model, hydrodynamic model) is identified, and then its corresponding parameter list is parsed. During parameter parsing, parameter values ​​are compared with predefined parameter templates to verify the correctness of the parameter type and value range. If errors are found, they are marked and recorded for subsequent correction.

[0163] Execute step 1530, information reading and calculation preparation

[0164] 1. Regional Information Acquisition

[0165] Based on the analyzed regional spatial extent, GIS spatial analysis tools (such as ArcGIS's SpatialAnalyst module) are used to acquire basic geographic information such as topographic data (Digital Elevation Model, DEM), soil type data, and land use data within the region. Topological relationship data of rivers and sub-basins within the region are retrieved from the watershed topology database, including upstream and downstream connections of rivers and confluence paths of sub-basins, providing spatial data support for subsequent calculations.

[0166] 2. Acquisition of rainfall event information

[0167] Based on the rainfall command parsing results, rainfall data for the corresponding time range is retrieved from the meteorological database. If it is historical rainfall data, the stored rainfall process file is read directly; if it is predicted rainfall data, the meteorological forecast model interface is called to obtain the latest rainfall forecast results. Data cleaning algorithms (such as outlier removal and missing value imputation) are used to preprocess the rainfall data to ensure data integrity and accuracy.

[0168] 3. Obtaining Site Monitoring Information

[0169] Real-time monitoring data such as water level, flow rate, and water quality from various monitoring stations are acquired through an IoT data acquisition platform or database interface. Data compression technologies (such as Gzip and Zlib) are used to reduce data transmission volume and improve data reading efficiency. Real-time quality control is performed on the acquired data, and an alarm mechanism with thresholds (such as triggering an alarm when the water level exceeds the warning level) is set to promptly detect and handle abnormal data.

[0170] 4. Scheduling model and parameter information acquisition

[0171] The system loads the model code and configuration files involved in scheduling from the model library and enables rapid model deployment via JAR / Python packages. Based on the parsed parameter instructions, it retrieves the corresponding model parameter values ​​from the parameter database. If the parameter values ​​are not fully specified, it uses default parameters or generates initial parameter values ​​through a parameter recommendation algorithm (such as a parameter analogy algorithm based on similar watersheds).

[0172] Execute step 1540, planning and calculation logic

[0173] 1. Correlation of forecast cross-section calculation units

[0174] Based on the geographical location of the forecast section, the computational units associated with the section are determined using GIS spatial overlay analysis. For distributed hydrological models, the mapping relationship between the section and the raster unit is established by determining whether the section coordinates are located within a certain raster unit. For lumped parameter models, the sub-basin where the section is located and its corresponding computational unit are determined based on sub-basin division and river topology. A data association table is established to record the correspondence between the forecast section and the computational unit, as well as the data transmission direction and format requirements.

[0175] 2. Post-processing association and logical planning

[0176] For forecast sections with model coupling, the execution order and data transfer relationships of each model are determined according to pre-configured post-processing rules. A directed acyclic graph (DAG) data structure is used to describe the computational logic, with nodes representing computational tasks (such as hydrological model calculations and reservoir scheduling model calculations), and edges representing data flow and task dependencies. For example, the output of the hydrological model calculation node serves as the input of the reservoir scheduling model calculation node. The execution sequence of computational tasks is generated through a topological sorting algorithm of the DAG, ensuring that each model is calculated in the correct logical order.

[0177] Execute step 1550 to determine boundary information.

[0178] 1. Determination of the boundary of the hydrological calculation unit

[0179] Upstream boundary conditions are typically derived from measured flow data or the output of upstream computational units, transmitted via data interfaces (such as Web Services or APIs). Downstream boundary conditions, depending on the actual river conditions, can be set to a fixed water level, a flow-water level curve, or boundary information fed back from downstream computational units. Utilizing hydrological principles (such as continuity equations and energy equations) and combining them with river cross-sectional geometric parameters (such as cross-sectional area and roughness), hydraulic elements at the boundary, such as water level and flow velocity, are calculated to provide accurate boundary conditions for hydrological calculations.

[0180] 2. Determination of reservoir operation boundaries

[0181] The input boundaries for reservoir scheduling include inflow (provided by the hydrological calculation unit), initial reservoir water level (obtained from reservoir monitoring data), and downstream river safe discharge (determined according to flood control standards). The output boundary is the reservoir outflow, which must satisfy the constraints of downstream river flood carrying capacity and flood control scheduling rules. By establishing a reservoir scheduling model (such as a rule-based scheduling model or an optimized scheduling model), the outflow process is calculated based on the input boundary conditions and scheduling rules, serving as the input boundary for the downstream calculation unit.

[0182] 3. Determination of the boundary of the hydrodynamic calculation unit

[0183] In a one- or two-dimensional hydrodynamic model, the upstream boundary can be set as a flow process line or a water level process line, while the downstream boundary is usually a water level boundary condition. When discretizing and solving the hydrodynamic equations using the method of characteristics or the finite difference method, appropriate numerical boundary conditions need to be set according to the boundary type. By interacting with data from surrounding computing units, the boundary information is updated in real time to ensure the accuracy and stability of the hydrodynamic calculations.

[0184] Step 1560: Start and execute the computing unit.

[0185] A multi-threaded or distributed computing framework is employed to launch computing units in parallel. Before starting, each computing unit checks the completeness of its input boundary data via a status query interface. If the input boundary data is not ready, it enters a waiting state; if the data is complete, the computation task is triggered. During computation, numerical methods (such as the finite element method and the finite volume method) are used to solve the model equations, and the computation progress and intermediate results are recorded in real time. After computation is completed, the output results are stored in a designated data storage location (such as a database or file system), and the boundary status flag of the computing unit is updated to "completed" for subsequent computing units to query and access.

[0186] Perform step 1570 to monitor and update the status.

[0187] A real-time monitoring system is deployed to acquire the status information of each computing unit and forecast section in real time through message queues or WebSocket technology. For computing units, the system monitors indicators such as computing progress, resource usage (CPU utilization, memory usage), and accuracy of calculation results. For forecast sections, the system monitors changes in key elements such as water level and flow rate. When a computing unit completes its calculation or the status of a forecast section changes, the system automatically updates the boundary state and forecast section status databases and, through an event-driven mechanism, notifies relevant computing units and forecast sections to proceed with the next calculation or data update, ensuring the continuity and accuracy of the entire computing scheduling process.

[0188] Execute step 1600 to visualize the results of the calculated sessions.

[0189] Example 2

[0190] This invention proposes a multi-model coupled parallel architecture design method for water conservancy systems. This method divides the entire water conservancy system into multiple basic forecast sections, which are mapped according to their upstream and downstream topological relationships. Simultaneously, multiple computing units are created using watershed operational characteristics and different models. In addition to mapping to the basic forecast sections, each unit corresponds to a computing node. These computing nodes are interconnected through a high-speed network, forming a distributed computing environment. Corresponding water conservancy model components, such as hydrological models, engineering scheduling models, hydrodynamic models, and inflow forecasting models, are deployed on each computing node. Model components on different nodes can be combined and work collaboratively as needed to achieve a comprehensive simulation of the entire water conservancy system.

[0191] (I) Principle of Basic Forecast Section Mapping

[0192] Based on an in-depth analysis of key elements of the water conservancy system, such as water flow, water resource distribution, and engineering regulation, this paper innovatively introduces the core concept of basic forecast sections to finely divide the vast and complex water conservancy system. Basic forecast sections are a unified and unique set of forecast sections formed based on hydrological and water resource models. The division of forecast sections is rigorously selected based on factors such as the natural hydrological network of the water conservancy system, river course, and the layout of key water conservancy facilities.

[0193] A rigorous mapping mechanism is established between basic forecast sections based on the natural topological relationships between upstream and downstream sections. This topological mapping is like an invisible yet robust network, tightly connecting various isolated sections and clearly outlining the dynamic transmission path of water flow throughout the entire water conservancy system. For example, the topological network enables real-time linkage of water conditions from the river source to the river mouth.

[0194] (II) Principle of Computational Unit Mapping

[0195] The numerous computational units derived from the hydraulic model do not exist in isolation. They are interconnected through tightly defined boundaries, forming a hierarchical topological relationship that enables seamless integration of data and computational processes. This boundary coupling mechanism acts like an invisible yet resilient bond, linking the various local units and enabling them to work collaboratively to jointly depict the overall operational picture of the hydraulic system. Specifically, after a computational unit completes its predetermined computational process, the key output, typically the outlet data at its boundary—such as dynamic parameters like flow rate, velocity, and water level when simulating water flowing to the unit boundary—is immediately transformed into the inlet input data for adjacent, downstream computational units, becoming the crucial initial conditions for driving the next unit's computational process to start and operate accurately. Similarly, along the water flow direction or business logic chain of the water conservancy system, each computing unit works closely together in sequence. The output of the preceding unit continuously injects power into the subsequent unit, enabling the entire model system to operate accurately and efficiently like a precisely meshed mechanical clock. This gradually realizes the leap from local unit simulation to comprehensive simulation of the entire watershed and the entire process of the water conservancy system, providing rock-solid technical support for the scientific decision-making of water conservancy projects, the sustainable management of water resources, and the effective protection of the ecological environment.

[0196] (III) Mapping Principle between Calculation Units and Basic Forecast Sections

[0197] In the architecture of water conservancy models, there exists a specific mapping logic relationship between basic forecast sections and the computational units constructed by different water conservancy models. This relationship is the key link to achieve accurate simulation and efficient regulation of water conservancy systems.

[0198] From the perspective of mapping from basic forecast sections to computational units, a basic forecast section can be associated with specific computational units under various different hydraulic model frameworks. This is because different hydraulic models focus on different characteristics and process analyses of the hydraulic system. For example, a basic forecast section located at a point of abrupt change in the slope of a river's main stream, within a hydrodynamic model system, will have its monitored flow velocity and water level change data mapped to a computational unit responsible for simulating the energy conversion of river flow and the dynamic adjustment of river morphology, used to accurately predict the trajectory of water flow under this special terrain. Within a hydrological model, the precipitation and runoff monitoring values ​​of this section will be associated with a computational unit focused on analyzing the runoff generation and concentration mechanisms of the watershed, helping to analyze the efficiency and process details of precipitation conversion into runoff in this region. This multi-dimensional mapping model fully leverages the multi-dimensional value of basic forecast section data, enabling it to comprehensively serve the core computational needs of different hydraulic models.

[0199] Conversely, from the perspective of computational units targeting basic forecast sections, some individual computational units are not directly associated with these sections. This is mainly because the functional positioning of some computational units does not directly rely on real-time monitoring section data, but rather focuses on in-depth exploration of the inherent laws of the water conservancy system, comprehensive analysis of historical data, or prediction and forecasting of future trends. These computational units, through complex calculations on various non-section data, provide high-level intellectual support for the long-term planning and macro-level decision-making of the water conservancy system. They complement those computational units closely driven by basic forecast section data, together weaving a rigorous and efficient computational network for the water conservancy model system.

[0200] (iv) The principle of adaptation between computing units and nodes

[0201] A close one-to-one relationship is established between computing units and computing nodes to ensure the precise allocation and efficient utilization of computing resources. Each computing node is carefully equipped with hardware resources and software environment that complement the characteristics of the corresponding computing unit. For example, for computing units undertaking flood evolution simulation tasks, their corresponding computing nodes are equipped with high-performance multi-core processors and large-capacity memory to meet the stringent requirements of hydrodynamic models to perform complex numerical calculations on massive grid data in a short period of time; while for computing units focusing on long-term water resource allocation planning, their computing nodes focus on optimizing the storage system to ensure that water resource supply and demand data spanning decades can be properly stored and efficiently accessed, working in conjunction with engineering scheduling models to achieve optimal water resource allocation.

[0202] (V) Principles of Distributed Collaborative Computing Tasks in Models

[0203] The complex computational process of hydraulic models is decomposed into multiple relatively independent subtasks, which can be executed simultaneously on different computation nodes. Taking a flood evolution model as an example, the entire flood evolution process can be decomposed into subtasks such as flood generation, river flood propagation, and floodplain inundation calculation. Each subtask has clear inputs and outputs and independent computational logic, and can begin execution without waiting for intermediate results from other subtasks, as long as its required initial data is ready.

[0204] The parallel framework is responsible for scheduling and coordinating these decomposed tasks, ensuring they are executed on appropriate computing nodes in the correct order and according to their dependencies. Based on the dependencies between tasks (e.g., reservoir water release decisions need to be based on upstream inflow predictions) and the availability of computing resources, the parallel framework rationally arranges the start and execution order of tasks, dynamically allocating computing resources to each task. This allows multiple tasks to proceed in parallel on different nodes, maximizing the utilization of computing resources and shortening the overall model's computation time.

[0205] Example 3

[0206] Taking the Daqing River as an example, it is necessary to integrate the hydrological and reservoir regulations models of the Xin'an River, as well as one- and two-dimensional hydrodynamic models.

[0207] (1) Basic forecast section collection and mapping

[0208] Within this basin, there are 23 hydrological model forecast sections for the Xin'an River, 7 reservoir forecast sections, and 1 hydrodynamic section.

[0209] After being compiled and organized, 23 basic forecast sections were formed, and upstream and downstream relationships were mapped.

[0210] (2) The computing unit collects the mapping

[0211] Within this watershed, there are 13 computational units for the Xin'anjiang hydrological model, 7 computational units for the reservoir, and 1 computational unit for the hydrodynamic model. Boundary mapping is also performed on the computational units for each type of model.

[0212] (3) Mapping of calculation units and basic forecast sections

[0213] Generate the mapping between the forecast sections and the basic sections of the Xin'anjiang hydrological model, reservoir regulation and scheduling model, and hydrodynamic model within the basin.

[0214] (4) Construction of calculation scheme

[0215] Select the basic forecast sections for the entire basin (Gongnongbing Gate, Ziya New River Tide Gate (newly built), and Erdao Gate (Haihe Main Stream) as the final basic forecast sections) to construct an integrated flood control forecasting and dispatching business for the basin and form a calculation scheme (named: Daqinghe Hydrological and Hydrodynamic Reservoir Demonstration Scheme).

[0216] (5) Setting up sessions

[0217] Based on the constructed calculation scheme (named: Daqinghe Hydrological and Hydrodynamic Reservoir Demonstration Scheme), the forecast time was selected as 8:00 on July 29, 2023, with a 24-hour warm-up period, a 306-hour lead time, and a calculation range of 1 hour. The rainfall information selected was the actual measured rainfall throughout the entire period, forming a specific measured + forecast rainfall event scheme (named: Daqinghe Hydrological Dispatch and Hydrodynamic Coupled Event).

[0218] (6) Parallel computing

[0219] Based on the constructed measured and forecasted rainfall scenario scheme (named: Daqing River Hydrological Scheduling Hydrodynamic Coupling Scenario), the various calculation units are allocated to various nodes, and the physical resources are allocated and called through the calculation scheme of this scenario (named: Daqing River Hydrological Hydrodynamic Reservoir Demonstration Scheme).

[0220] Based on the calculation scheme (named: Daqinghe Hydrological and Hydrodynamic Reservoir Demonstration Scheme), the basic cross-section, model, parameters, and topological information of the scheme were obtained.

[0221] First, each basic forecast section in the region is taken as the core calculation point (23 in total), and the corresponding calculation units (13 calculation units) are associated; the reservoir scheduling post-processing (7 reservoir calculation units) and hydrodynamic post-processing (1 hydrodynamic calculation unit) of the basic forecast section are associated; and the logical structure of the calculation is planned.

[0222] Then, the calculations of the 21 calculation units (including hydrology, reservoir scheduling, and hydrodynamics) are initiated. Next, the system checks whether the input boundaries of the unit are complete. If the input boundaries are complete, the calculation for that unit begins. After the calculation is completed, the boundaries of the calculation unit are updated, indicating that the calculation for that unit is finished. Other calculation units proceed with their calculations according to this logical judgment.

[0223] While the computational unit is performing calculations, boundary state monitoring and forecast section state monitoring are also conducted simultaneously. These are used to update the boundary state and forecast section state to ensure that the next computational unit and the next forecast section can perform calculations.

[0224] (7) Results presentation

[0225] The calculation results of the field scheme (named: Daqing River Hydrological Dispatch Hydrodynamic Coupled Field) will be divided into hydrological, reservoir and hydrodynamic results.

[0226] To better understand this invention, specific embodiments have been described in detail above, but these are not intended to limit the invention. Any simple modifications made to the above embodiments based on the technical essence of this invention still fall within the scope of this invention. Each embodiment in this specification focuses on its differences from other embodiments; similar or identical parts between embodiments can be referred to mutually. For system embodiments, since they basically correspond to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

Claims

1. A method for designing a coupled parallel architecture for multi-model hydraulic engineering, comprising basic forecast section collection and mapping, characterized in that, It also includes the following steps: Step 1: The computing unit collects the mapping; Step 2: Mapping the calculation unit to the basic forecast section; Step 3: Taking the basic forecast section as the core, and based on flood control needs or water resources, configure different models and parameter information required for different sections to form a calculation scheme. The construction of the calculation scheme includes the following sub-steps: Step 31: Create and name the calculation scheme; Step 32: By selecting the start and end basic forecast sections, the system will automatically filter and determine the calculation area of ​​the calculation scheme based on their topological relationships; Step 33: Within the calculation area, list the basic forecast sections and select a specific basic forecast section; Step 34: For this basic forecast section, select various models that are mapped to this section; Step 35: After determining the model of the basic forecast section, select the specific model parameters for that model; Step 36: In the initial state of the basic forecast section, configure the hydrological model and parameters, select the reservoir model and parameters, and mark it as post-processing; Step 37: Configure each basic forecast section sequentially according to the topological relationship, and save the calculation scheme; Step 4: Using the constructed calculation scheme, configure various data such as time periods and meteorological monitoring to form different scenario session plans; Step 5: Analyze the constructed session plan, allocate various computing units to various nodes, and allocate and call physical resources according to the computing plan of the session; Step 6: Visualize the results of the calculated sessions.

2. The hydraulic multi-model coupled parallel architecture design method as described in claim 1, characterized in that, The basic forecast section collection mapping includes the following sub-steps: Step 01: Collection and analysis of basic watershed data; Step 02: Preliminary selection of cross sections based on forecast requirements, including flood forecast cross sections, water resource management cross sections, and ecological flow guarantee cross sections; Step 03: Cross-sectional representativeness and feasibility assessment; Step 04: Based on the model simulation, optimize and verify the cross-section to determine the final basic cross-section, which includes at least the cross-section location and basic parameters.

3. The hydraulic multi-model coupled parallel architecture design method as described in claim 2, characterized in that, Step 1 includes the following sub-steps: Step 11: Data collection and preprocessing for computing units; Step 12: Calculate the characteristics of the calculation unit; Step 13: Constructing the topology of computing units; Step 14: Mapping the boundaries of computational units based on topological relationships; Step 15: Define the mapping rules.

4. The hydraulic multi-model coupled parallel architecture design method as described in claim 3, characterized in that, Step 11 includes obtaining computational unit data from different models, using data cleaning algorithms to remove outliers and missing values, identifying and correcting outliers in flow data through box plots, and using linear interpolation to fill in missing values ​​in water level data. The computational unit data includes raster cell data from distributed hydrological models, sub-basin cell data from conceptual hydrological models, and channel cross-section or grid cell data from hydrodynamic models.

5. The hydraulic multi-model coupled parallel architecture design method as described in claim 4, characterized in that, Step 12 includes the following sub-steps: Step 121: Calculate the geometric properties of each computational unit; Step 122: Analyze the resolution and coverage of the grid cells to determine their accuracy in describing the spatial characteristics of the watershed; Step 123: Use the watershed drainage system extraction algorithm to determine the catchment area and boundary morphology of the sub-watershed unit; Step 124: Through spatial overlay analysis, compare the spatial overlap of computational units of different models to identify potential data conflict areas.

6. The hydraulic multi-model coupled parallel architecture design method as described in claim 5, characterized in that, Step 13 includes the following sub-steps: Step 131: Construct the water system topology. Based on the watershed water system, use the river network tracing algorithm of GIS to construct the upstream and downstream topological relationships of the river channel, assign a unique identifier to each river channel cross-section unit, and establish a topology table to record the connection relationships between the units. Step 132: Construct the watershed unit topology. Based on the terrain slope and water flow direction, determine the confluence relationship between sub-watershed units. Using ArcGIS's hydrological analysis tools, generate a flow direction raster map of the watershed, then delineate the sub-watershed boundaries and establish a topological network between sub-watersheds. In the distributed raster model, based on the water flow direction of adjacent rasters, construct the local topology of the raster units and clarify the transmission path of water flow between rasters.

7. The hydraulic multi-model coupled parallel architecture design method as described in claim 6, characterized in that, Step 2 includes the following sub-steps: Step 21: Spatial Overlay Analysis; Step 22: Distance calculation and analysis; Step 23: Topological relationship analysis.

8. The hydraulic multi-model coupled parallel architecture design method as described in claim 7, characterized in that, Step 32 includes the following sub-steps: Step 321: Select the basic forecast section; Step 322: Analyze the topological relationships; Step 323: Determine the calculation area.

9. The hydraulic multi-model coupled parallel architecture design method as described in claim 8, characterized in that, Step 322 includes retrieving pre-constructed watershed channel topology data, storing the upstream and downstream connections of the river in the form of a directed graph, and using breadth-first search in graph theory to search along the topology path from the starting section until the ending section is reached, including all river segments, sub-basins and other areas involved in the path into the calculation area.

10. The hydraulic multi-model coupled parallel architecture design method as described in claim 9, characterized in that, Step 5 includes the following sub-steps: Step 51: Start the scheduling process; Step 52: Execute the node parsing task; Step 53: Information Reading and Calculation Preparation; Step 54: Planning and Calculation Logic; Step 55: Determine boundary information; Step 56: Start and execute the computing unit; Step 57: Monitor and update the status.

Images