A basin flood control panoramic decision-making consultation method and system

Abstract

The present application relates to the technical field of water conservancy dispatching, and provides a basin flood control panoramic decision-making consultation method and system; the method of the present application constructs a physical and digital fusion panoramic consultation environment, integrates multi-sensing hardware and a basin three-dimensional digital twin environment, trains a host model and a multi-field digital person evaluation model, and realizes real-time convergence and intelligent fusion of multi-source heterogeneous data. Through multi-modal interaction such as voice and gesture, real-time flood evolution, parallel comparison and selection of dispatching scheme suggestions, generation of decision minutes and dispatching instructions can be realized; combined with application verification of real historical data and multi-scenario comparison and selection paths, multiple comprehensive and accurate feasible dispatching scheme suggestions are obtained, and the final scheme is obtained through final comparison and selection, which significantly improves the scientificity, accuracy, comprehensiveness and efficiency of major flood emergency dispatching decisions, and forms a reusable decision case library, effectively solving the problems of data island, intelligence deficiency, deduction lag and coordination difficulty in existing basin consultation.

Description

Technical Field

[0001] This invention relates to the field of water conservancy engineering technology, specifically to a panoramic decision-making and consultation method and system for river basin flood control. Background Technology

[0002] In modern water conservancy management and public safety systems, basin flood control scheduling and emergency command are crucial links. Especially against the backdrop of frequent extreme weather events, efficient, scientific, and collaborative consultation and decision-making are essential for mitigating disaster losses and protecting people's lives and property. Currently, basin consultation and decision-making mainly relies on traditional meeting models and information technology-assisted methods, which still have the following significant shortcomings: First, the multi-source data on which consultation and decision-making depend is fragmented and two-dimensional. Heterogeneous data such as hydrological, meteorological, remote sensing, engineering operation, and socio-economic data are scattered in the information systems of different departments, lacking an effective real-time fusion and unified access mechanism. This forces decision-makers to frequently switch between multiple independent systems and two-dimensional charts, making it difficult to quickly and intuitively form a three-dimensional understanding of the overall situation of the basin and its spatiotemporal evolution of floods. Information overload and cognitive burden seriously restrict decision-making efficiency.

[0003] Secondly, the decision-making process relies heavily on the personal experience and on-the-spot judgment of human experts, lacking in-depth support from intelligent auxiliary tools. When faced with complex scheduling scenarios, experts often find it difficult to accurately quantify and compare multiple alternative plans and evaluate their effects within a limited time. Traditional contingency plan databases are outdated and cannot be adaptively adjusted by combining with real-time dynamic data. The decision-making process is highly subjective, and the response speed is insufficient to meet the urgent needs of extreme disasters.

[0004] Furthermore, existing collaborative interaction methods are limited, resulting in weak immersion and participation. Traditional remote consultations are mostly limited to audio and video communication and file sharing, failing to provide participants with a shared, interactive three-dimensional environment for collaborative operations and focused discussions. Remote experts lack a sense of presence and struggle to form an accurate spatial understanding of the situation, severely impacting the efficient collaboration of cross-regional and multidisciplinary teams.

[0005] Furthermore, there is a lack of smooth coordination between key links such as forecasting, early warning, rehearsals, and contingency plans. In the traditional model, the issuance of flood forecasts and early warning information and the subsequent deduction and simulation of dispatching plans and the execution of contingency plans often belong to different stages and systems. There is a lack of a unified, visualized digital twin environment for real-time simulation and dynamic verification. Decision-makers cannot intuitively observe the evolution of floods, changes in inundation range, and risk transfer under different dispatching instructions at the consultation site. As a result, the formulation and adjustment of contingency plans lack accurate data support and intuitive feedback on effects.

[0006] Finally, the efficiency of transforming and consolidating the results of consultations is low. The discussion, decision-making process, and deduction results rely heavily on manual recording and organization, which is time-consuming, labor-intensive, and prone to missing key information and logical chains. The resulting scheduling instructions, meeting minutes, and other documents have low levels of structure, making it difficult to compare and contrast with historical cases and to consolidate knowledge, thus failing to provide effective experience for future decision-making.

[0007] In summary, existing technologies struggle to construct a comprehensive, intelligent, and collaborative intelligent decision-making and consultation environment for watersheds in terms of multi-party data fusion and decision-making scheme generation. This results in the final decision-making schemes not considering all factors, lacking rigor, and being difficult to verify. Summary of the Invention

[0008] This invention aims to solve the problems of low efficiency and insufficient intelligence and comprehensiveness in watershed decision-making consultation in existing technologies, and provides a panoramic decision-making consultation method for watershed flood control that can comprehensively integrate multi-party data to quickly generate decisions.

[0009] To achieve the above objectives, a first aspect of the present invention provides a method for panoramic decision-making and consultation on flood control in a river basin, the method comprising:

[0010] S1: Acquire multidimensional heterogeneous data of the watershed, unify the multidimensional heterogeneous data to a preset geographic coordinate system and time base, and generate a standardized time-series data stream; S2: Construct a three-dimensional digital twin environment of the watershed based on watershed geographic information data, and establish a data-driven mapping relationship between the three-dimensional digital twin environment and meteorological models, hydrological models and hydrodynamic models; S3: Collect flood control documents, historical dispatch cases, emergency plan texts, water conservancy engineering specifications and standards, and expert discussion records in the basin to build a professional corpus; use the professional corpus to adjust the pre-trained Transformer architecture large language model to obtain the host model; S4: Construct a water conservancy knowledge graph, input the water conservancy knowledge graph into multiple large language models for few-shot learning, and train multiple digital human evaluation models for hydrological forecasting, engineering scheduling and risk assessment respectively. S5: Input standardized time-series data streams into the watershed's three-dimensional digital twin environment to drive the hydrological and hydrodynamic models to perform real-time calculations, generating a dynamic projection of floods and droughts for the entire watershed; and combine this with a water resources knowledge graph to form a comprehensive decision-making scenario that includes real-time water and rainfall data, flood forecast curves, and historical disaster cases. S6: The host model analyzes the consultation topic corresponding to the current instruction, controls the virtual camera in the three-dimensional digital twin environment to focus on the target geographical area, and calls the corresponding digital human evaluation model based on the logic of water conservancy knowledge graph, historical case retrieval and physical model simulation results, combined with the overall decision-making situation, to generate quantitative scheduling plan suggestions; S7: Through the three-dimensional digital twin environment of the watershed, the generated quantitative scheduling scheme suggestions are simulated and dynamically extrapolated in real time to generate flood evolution paths, inundation ranges, water level process lines and heat maps of at-risk populations under different schemes; according to the preset optimization objectives, combined with the rule base or optimization algorithm, the scheduling parameters are adjusted to determine the optimal scheduling scheme.

[0011] In one embodiment, S1 specifically includes: The acquired multidimensional heterogeneous data of the watershed is cleaned, verified and formatted; the multidimensional heterogeneous data includes real-time monitoring data of hydrological station network, meteorological forecast data, remote sensing satellite image data, water conservancy project operation data and socio-economic and population distribution data, and is accessed through standard API and Internet of Things protocol; By employing coordinate transformation and time synchronization algorithms, multidimensional heterogeneous data from different sources are unified under a preset geographic coordinate system and time reference, generating a standardized time-series data stream.

[0012] In one embodiment, the watershed geographic information data includes digital elevation models, orthophotos, river system vector data, and building information models of the required water conservancy facilities.

[0013] In one embodiment, establishing the data-driven mapping relationship between the three-dimensional digital twin environment and the meteorological model, hydrological model, and hydrodynamic model specifically includes: The two-dimensional flood evolution model is solved based on the finite volume method or the finite difference method. The input parameter interface, calculation core and result output interface of the solved two-dimensional flood evolution model are encapsulated as RESTful API or gRPC service. The three-dimensional rendering engine calls the RESTful API or gRPC service through timed polling or event triggering mechanism to obtain real-time calculated flow field data and water level data, and converts them into texture data or geometric data that the shader can recognize, so as to realize the dynamic rendering of water flow direction, flow velocity distribution and flood range in the three-dimensional scene.

[0014] In one embodiment, the construction of the water resources knowledge graph specifically includes: Using water conservancy projects, flood control scheduling indicators, and water resource management elements as entities, and the relationships between entities as edges, a water conservancy knowledge graph is obtained by organizing them according to the entity-relationship-attribute structure.

[0015] In one embodiment, the training yields multiple digital human evaluation models for hydrological forecasting, engineering scheduling, and risk assessment, specifically including: By associating key events in the real-time data stream with entities in the water resources knowledge graph, early warning information with semantic tags is generated. The early warning information is overlaid and rendered in a three-dimensional digital twin environment as a visual layer, forming a panoramic view of decision-making that includes real-time water and rainfall data, flood forecast curves, and historical disaster cases; The method of associating key events in the real-time data stream with entities in the water resources knowledge graph includes: The real-time water level data is compared with the corresponding defense standard entities of the dike in the water conservancy knowledge graph. When the water level exceeds the defense standard threshold, the inference engine traverses the protection area, population distribution and important facility entities associated with the dike in the water conservancy knowledge graph. Based on the inference results, natural language early warning information containing the specific impact range, the estimated population to be relocated and the pre-assessment of economic losses are generated, and the spatial anchoring position of the early warning information in the three-dimensional scene is dynamically marked.

[0016] In one embodiment, the digital human evaluation model in step S6 generates a quantitative scheduling scheme suggestion, specifically including: The system simulates real-time scenarios using a three-dimensional digital twin environment of the watershed; simultaneously, it retrieves constraint rules, historical analogies, and impact assessment indicators related to the current simulation scenario and scheduling issues from the water conservancy knowledge graph; it then compares and performs logical calculations with the retrieved rules, cases, and indicators to generate quantitative analysis conclusions and scheduling scheme recommendations; finally, it uses speech synthesis technology to broadcast the conclusions in natural language and displays them through overlay visualization charts and three-dimensional annotations.

[0017] In one implementation, after the scheduling scheme suggestion is generated, the method further includes: Historical plans are loaded, and the human-computer interaction system preloads a default plan library and provides a variety of historical plans by having experts set scheduling parameters on-site. Real-time scenario simulation and dynamic extrapolation are performed on various combinations of historical scenarios to generate flood evolution paths, inundation ranges, water level process lines, and heat maps of at-risk populations for various combinations of historical scenarios; The differences between the proposed scheduling scheme and various combinations of historical schemes in terms of preset optimization objectives are compared. The preset optimization objectives include flooded area, affected population, economic loss and response time. Based on the real-time scenario simulation and dynamic extrapolation results of the generated quantified scheduling scheme suggestions, the real-time scenario simulation and dynamic extrapolation results of various combinations of historical schemes, and the differences between the scheduling scheme suggestions and various combinations of historical schemes in terms of preset optimization objectives, the system automatically generates adjustment suggestions for scheduling parameters based on rule base, genetic algorithm or particle swarm algorithm, and determines the optimal scheduling scheme.

[0018] In one embodiment, it further includes: The selected target scheduling scheme suggestion is automatically filled into the preset standard instruction template to generate a scheduling command file that is formatted and can be directly applied; all input data, intermediate deduction process, final scheme and expert instructions in the current consultation process are associated and encapsulated to form a decision case package with semantic tags and stored in the system knowledge base.

[0019] A second aspect of the present invention provides a computer system including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the basin flood control panoramic decision-making consultation method as described above.

[0020] Compared with the prior art, the present invention has the following beneficial effects: (1) The method of this invention constructs a panoramic immersive consultation environment that integrates physical and digital elements, integrates multi-sensor hardware and a three-dimensional digital twin environment of the basin, trains a host model and a multi-domain digital human evaluation model, and realizes real-time aggregation and intelligent fusion of multi-source heterogeneous data. It supports multi-modal interaction such as voice, gesture, and touch, and can simulate flood evolution in real time, compare and select scheduling schemes in parallel, generate decision minutes and scheduling instructions. Through application verification combining real historical data and multiple scheme comparison paths, it obtains multiple comprehensive and accurate feasible scheduling scheme suggestions, and obtains the final scheme through final comparison. It significantly improves the scientificity, accuracy, comprehensiveness and efficiency of emergency scheduling decisions for major floods, and forms a reusable decision case library, effectively solving the problems of data silos, lack of intelligence, lagging inference and difficulty in collaboration in existing basin consultations.

[0021] (2) The basin flood control panoramic decision-making consultation method of the present invention reduces the fusion processing delay of multi-source heterogeneous data to the second level through spatiotemporal alignment algorithm and service model interface, realizes data-driven dynamic situation update, and improves the real-time performance of multi-source data fusion. The hydrodynamic simulation output is bound to the three-dimensional scene in real time, which supports direct observation of the flood evolution process under different scheduling scheme suggestions in the three-dimensional environment, improves spatial cognition efficiency, and enhances the visualization and interaction capabilities of the hydrodynamic model. Through real-time synchronization and collaborative annotation of the three-dimensional scene, it ensures that experts in different locations operate in the same spatial coordinate system, eliminates the spatial understanding bias in traditional audio and video consultation, and reduces the spatial cognition error of remote collaboration. Through semantic analysis and structured recording, the decision basis, inference results and scheduling parameters of each consultation are automatically encapsulated into reusable decision cases, providing data support for subsequent plan optimization and model training, and realizing the automated knowledge accumulation of the decision-making process.

[0022] (3) By integrating hydrodynamic models and three-dimensional digital twin scenarios, multiple possible scheduling scheme suggestions are efficiently generated, realizing real-time visualization and simulation of scheduling scheme suggestions and parallel comparative analysis of multiple plans. The forecasting, early warning, simulation and plan are closed-loop operated in a unified environment, enabling decision-makers to intuitively verify the effectiveness of the scheme and quickly iterate and optimize it, greatly improving the scientificity and foresight of flood control scheduling, and realizing the scientification of the decision-making process and results. At the same time, by combining real historical data and application verification of multiple plan comparison paths, the scientificity, accuracy, comprehensiveness and efficiency of emergency scheduling decisions for major floods are significantly improved. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of a panoramic decision-making and consultation method for watershed flood control according to an embodiment of the present invention; Figure 2 This is a diagram illustrating the architecture of a human-computer interaction system according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the hardware layout of the human-computer interaction system according to an embodiment of the present invention; Figure 4 A logic diagram for intelligent roles to work together. Detailed Implementation

[0025] To facilitate understanding of the present invention, the present invention will be described more fully and in detail below with reference to the accompanying drawings and preferred embodiments, but the scope of protection of the present invention is not limited to the following specific embodiments.

[0026] Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by those skilled in the art. The technical terms used herein are for the purpose of describing particular embodiments only and are not intended to limit the scope of the invention.

[0027] Unless otherwise specified, all raw materials, reagents, instruments and equipment used in this invention can be purchased from the market or prepared by existing methods.

[0028] In this embodiment, a human-computer interaction system required for the consultation is pre-deployed at the consultation location as hardware support to implement a panoramic decision-making consultation method for watershed flood control in this embodiment. Specifically, it establishes a physical space supporting immersive decision-making and a three-dimensional digital twin environment with real-time simulation and computation capabilities, providing a foundational platform for all subsequent functions. Figure 3 The hardware deployment scheme is shown below, and includes the following: A panoramic immersive display wall, consisting of multiple high-lumen laser projectors, an edge blending server, and a geometric correction system, is deployed at the consultation location to create a surround-view visualization environment. A distributed microphone array, depth-sensing cameras, motion-sensing interaction devices, and a multi-touch console are simultaneously configured to construct a multi-sensory hardware interaction system integrating voice, gesture, touch, facial recognition, and positioning. This system can also store the models required for the basin flood control panoramic decision-making consultation method. The stored models, in conjunction with the aforementioned hardware devices, are invoked to fully implement the basin flood control panoramic decision-making consultation method.

[0029] The human-computer interaction system in this embodiment also transmits 3D scene data streams in real time to remote terminals via 5G or dedicated network. These remote terminals include desktop clients, video conferencing terminals, and VR headsets, with rendering precision adaptively adjusted based on terminal performance. This allows the system to receive operation commands from remote experts (such as rotating the viewpoint, clicking to query, and spatial annotation), which are then synchronized in real time to the main venue server via a signaling channel. The server broadcasts these commands to all online terminals, ensuring all participants see the same scene. Spatial annotation supports drawing lines and adding text labels in the 3D scene; the annotation information is bound to geographic coordinates and can rotate and scale with the scene.

[0030] The core implementation process of the basin flood control panoramic decision-making and consultation method is as follows: Figure 1 As shown, it includes: S1: Data input and preprocessing, acquiring multidimensional heterogeneous data of the watershed, unifying the multidimensional heterogeneous data to a preset geographic coordinate system and time base, and generating a standardized time-series data stream, specifically including: The acquired multidimensional heterogeneous data of the watershed is cleaned, verified and formatted; the multidimensional heterogeneous data includes real-time monitoring data of hydrological station network, meteorological forecast data, remote sensing satellite image data, water conservancy project operation data and socio-economic and population distribution data, and is accessed through standard API and Internet of Things protocol; By employing coordinate transformation and time synchronization algorithms, multidimensional heterogeneous data from different sources are unified under a preset geographic coordinate system and time reference, generating a standardized time-series data stream. Spatiotemporal alignment uses a combination of linear interpolation and nearest neighbor matching to ensure that data with different sampling frequencies are aligned on the time axis.

[0031] S2: Construction of a three-dimensional digital twin base, constructing a three-dimensional digital twin environment of the watershed based on watershed geographic information data, establishing a data-driven mapping relationship between the three-dimensional digital twin environment and meteorological models, hydrological models and hydrodynamic models, so that the three-dimensional digital twin environment has the ability to perform real-time dynamic simulation based on physical laws; The watershed geographic information data includes digital elevation models, orthophotos, river system vector data, and building information models of the required water conservancy facilities.

[0032] The establishment of a data-driven mapping relationship between the three-dimensional digital twin environment and meteorological, hydrological, and hydrodynamic models specifically includes: The two-dimensional flood evolution model is solved based on the finite volume method or the finite difference method. The input parameter interface, calculation core and result output interface of the solved two-dimensional flood evolution model are encapsulated as RESTful API or gRPC service. The three-dimensional rendering engine calls the RESTful API or gRPC service through timed polling or event triggering mechanism to obtain real-time calculated flow field data and water level data, and converts them into texture data or geometric data that the shader can recognize, so as to realize the dynamic rendering of water flow direction, flow velocity distribution and flood range in the three-dimensional scene.

[0033] S3: Intelligent role model construction, collecting flood control documents, historical dispatch cases, emergency plan texts, water conservancy engineering specifications and standards and expert discussion records in the basin to build a professional corpus; using the professional corpus to adjust the pre-trained Transformer architecture large language model to obtain a host model with water conservancy professional semantic understanding and consultation process management capabilities; The presenter model in this embodiment controls the perspective switching and model calculation startup of the 3D digital twin scene based on the preset agenda and semantic parsing results; when a keyword of a specific topic is detected, the relevant data layer is automatically retrieved and the visualization content is updated, realizing intelligent agenda control and scene linkage.

[0034] S4: Construct a water conservancy knowledge graph, input the water conservancy knowledge graph into multiple large language models for few-sample learning, and train multiple digital human evaluation models that are respectively specialized in multiple fields such as hydrological forecasting, engineering scheduling and risk assessment. The collaborative logic of the intelligent role constructed in this embodiment is as follows: Figure 4 As shown, the role and steps of the moderator model and digital human evaluation model in the consultation process are demonstrated. Based on the pre-constructed consultation environment, unified collaborative decision-making by participating experts, digital human experts, etc. is achieved, and the efficiency and accuracy of decision-making are greatly improved.

[0035] The construction of the water conservancy knowledge graph specifically includes: Using water conservancy projects, flood control scheduling indicators, and water resource management elements as entities, and the affiliation, influence, and association relationships between entities as edges, a knowledge graph for flood control in the water conservancy field is obtained by organizing the data according to the entity-relationship-attribute structure.

[0036] The expert experience rule base is a database formed by summarizing and generalizing the case studies of the actions and deployments made by experts in previous scheduling processes or other meetings.

[0037] S5: Panoramic Situation Fusion and Generation. Standardized time-series data streams are input into the watershed's 3D digital twin environment. The spatiotemporally aligned data serves as boundary conditions, driving the hydrological and hydrodynamic models to perform real-time calculations and rapid simulations, generating a dynamic projection of flood and drought conditions for the entire watershed (including water level changes, flow velocity distribution, and inundation range). Combined with a water resources knowledge graph, a decision-making panoramic situation is formed, including real-time water and rainfall data, flood forecast curves, and historical disaster cases. The training yielded several digital human evaluation models for hydrological forecasting, engineering scheduling, and risk assessment, specifically including: By associating key events in the real-time data stream with entities in the water resources knowledge graph, early warning information with semantic tags is generated. The early warning information is overlaid and rendered in a three-dimensional digital twin environment as a visual layer, forming a panoramic view of decision-making that includes real-time water and rainfall data, flood forecast curves, and historical disaster cases; The method of associating key events in the real-time data stream with entities in the water resources knowledge graph includes: The real-time water level data is compared with the corresponding defense standard entities of the dike in the water conservancy knowledge graph. When the water level exceeds the defense standard threshold, the inference engine traverses the protection area, population distribution and important facility entities associated with the dike in the water conservancy knowledge graph. Based on the inference results, natural language early warning information containing the specific impact range, the estimated population to be relocated and the pre-assessment of economic losses are generated, and the spatial anchoring position of the early warning information in the three-dimensional scene is dynamically marked.

[0038] S6: Interactive consultation and decision-making. It receives voice or text commands from participating experts, analyzes the consultation topic corresponding to the current command through a moderator model, and automatically controls the human-computer interaction system to present the overall situation of the watershed. Based on the analyzed topic, the moderator model automatically controls the virtual camera in the 3D digital twin environment to focus on the target geographical area and calls the corresponding digital human evaluation model. Based on the logic of the water resources knowledge graph, historical case retrieval, and physical model simulation results, combined with the overall decision-making situation, it generates quantitative scheduling plan suggestions. When commands such as "assess the impact of a certain plan on the downstream" are detected, the system retrieves relevant domain rules (such as dike design standards and flood storage area activation conditions) and historical similar cases from the water resources knowledge graph. Combining real-time hydrological data and the model extrapolation results of the current plan, it generates quantitative analysis conclusions, such as "the peak water level at downstream station X is expected to rise by Y meters, and the warning time will be Z hours earlier," and overlays them onto the 3D scene in the form of heat maps and water level curves.

[0039] The proposed scheduling scheme for the digital human evaluation model in S6 includes the following: The system simulates real-time scenarios using a three-dimensional digital twin environment of the watershed; simultaneously, it retrieves constraint rules, historical analogies, and impact assessment indicators related to the current simulation scenario and scheduling issues from the water conservancy knowledge graph; it then compares and performs logical calculations with the retrieved rules, cases, and indicators to generate quantitative analysis conclusions and scheduling scheme recommendations; finally, it uses speech synthesis technology to broadcast the conclusions in natural language and displays them through overlay visualization charts and three-dimensional annotations.

[0040] S7: Scheme Comparison and Output. Through the three-dimensional digital twin environment of the watershed, the generated scheduling scheme suggestions are simulated and dynamically extrapolated in real time, generating flood evolution paths, inundation ranges, water level process lines, and heat maps of at-risk populations under different schemes; multiple scheduling scheme suggestions and their dynamic extrapolation results are visualized and compared in parallel on the same comparison dashboard, showing the differences between the schemes in key indicators; based on the preset optimization objectives, combined with the rule base or optimization algorithm, suggestions for adjusting scheduling parameters are generated, forming optimized scheduling scheme suggestions.

[0041] In one embodiment, S7 specifically includes: S7.1: Historical scheme loading: The human-computer interaction system preloads the default scheme library and provides a variety of historical schemes by having experts set scheduling parameters on-site (such as reservoir discharge and gate opening combinations). S7.2: Real-time simulation and deduction. Using a 3D digital twin environment of the watershed, based on real-time hydrological data and historical scenarios, the simulation depicts the evolution of future floods and visualizes it using WebGL dynamic inundation animations, water level process graphs, and / or impact range heat maps. The simulation results for each scenario include: dynamic inundation range animations (with different colors indicating water depth), key section water level process graphs, risk area statistics, and a list of affected populations and important facilities. These results are displayed side-by-side in a comparison dashboard, allowing users to switch timelines to observe differences at different times.

[0042] S7.3: Parallel comparison, through a multi-view comparison dashboard, simultaneously displays the differences between the suggested scheduling schemes and historical schemes in terms of flooded area, affected population, economic losses, and response time, to assist in quantitative comparison and selection; S7.4: Optimization Generation. Based on the deduction results and preset optimization objectives, and using a rule base, genetic algorithm, or particle swarm optimization algorithm, the system automatically generates suggestions for adjusting scheduling parameters to form a new optimal scheduling scheme.

[0043] S8: At the end of the meeting, the moderator model automatically extracts key conclusions and decision points, and generates a structured draft meeting minutes; based on the finalized scheduling plan, the system automatically matches a standard template and fills in the parameters to generate a standardized and directly applicable scheduling command file. Furthermore, the human-computer interaction system records voice and text commands, 3D operation trajectories, and deduction results in real time during the consultation process; it performs real-time speech transcription of the entire consultation process using the host model, and generates a timestamped text stream using a Transformer-based speech recognition model; it extracts key sentences using the TextRank algorithm and constructs a decision event sequence by combining it with operation logs (model call records, parameter modification records); and it encapsulates the finalized scheduling parameters, deduction result snapshots, and key decision texts into structured data in JSON-LD format, adds semantic tags (such as event type, geographical location, and time range), and stores it in the Elasticsearch index to support subsequent semantic retrieval and case reuse.

[0044] In one embodiment, a computer system is provided, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the basin flood control panoramic decision-making consultation method as described above.

[0045] Through the hardware deployment, data integration, and model training of the consultation environment in the above embodiments, a complete human-computer interaction system is obtained, and the system architecture is as follows: Figure 2 As shown, this creates a perfect closed loop between each link.

[0046] Taking the scheduling decision of Reservoir A during a flood in a certain water area as an example, the implementation process of the present invention will be explained.

[0047] Example Background: Two consecutive flood events upstream of this waterway have had a severe cumulative impact on its middle and lower reaches. Reservoir A faces a continuous massive inflow of floodwater, and the middle and lower reaches of the main stream, especially section B, are under immense flood control pressure. The core challenge in decision-making lies in how to coordinate the impoundment and discharge of water from Reservoir A to minimize the flood control pressure on the downstream section B while ensuring the safety of the reservoir itself, and to reduce the social impact of activating the flood storage and detention area as much as possible. Initially, a decision-making consultation hall integrating "immersive visualization" and "intelligent multimodal interaction" has been built. The conference hall utilizes a seamless panoramic projection technology with eight high-definition projectors to build a "Watershed-wide Major Flood Virtual-Real Twin Scenario Simulation System Platform." This platform, based on Cesium, employs 3D digital twin technology and real-time hydrodynamic simulation methods to achieve a dynamic, three-dimensional presentation of the watershed flood evolution process. It also deploys an intelligent voice interaction system and Kinect motion-sensing devices, supporting natural interaction functions with both voice and gestures, achieving a closed-loop response of "command-scenario-control." Through a unified data interface, it integrates multi-source heterogeneous information, providing an integrated support environment for command and decision-making, encompassing fusion perception, intelligent scheduling, and collaborative analysis. Simultaneously, it trains a digital human evaluation model library in fields such as meteorology, hydrology, scheduling, and emergency management.

[0048] After the consultation was launched, the host model first gave a voice report on the current severe situation, while the panoramic screen simultaneously highlighted the river section exceeding the warning level, the real-time reservoir capacity of Reservoir A, and the animation of the movement of upstream rain clouds, enabling decision-makers to quickly form a unified panoramic understanding.

[0049] The system accesses real-time data, including the National Meteorological Center's grid rainfall forecast (5km resolution), real-time data from 156 hydrological stations in the area, high-resolution satellite remote sensing images (to extract the inundation range), and operational data from Reservoir A (inflow, water level, and status of flood discharge facilities).

[0050] An initial situation is constructed, and based on the aforementioned data, a hydrodynamic model is driven to generate a 24-hour flood evolution simulation, displaying an animation of water level changes from Reservoir A to River section B in a 3D scene. The system automatically identifies that the water level at station B will exceed the warning line and highlights the high-risk river section in the scene.

[0051] Multiple contingency plans are simulated in parallel, generating three sets of plans based on the scheduling rule base: Plan A (optimized compensation scheduling), Plan B (activation of one flood storage and detention area), and Plan C (activation of multiple flood storage and detention areas). The system starts three model instances in parallel and completes a 24-hour simulation within 5 minutes.

[0052] Through expert interaction and analysis, the commander-in-chief gave the voice command: "Assess the impact of Plan A on Station B." After parsing the command, the system retrieved the simulation results of Plan A, generated a comparison chart of the water level process line at Station B with Plans B and C, and simultaneously retrieved the scheduling effects of similar historical flood events from the water conservancy knowledge graph for reference.

[0053] Through remote collaboration, experts at branch venues accessed the platform using VR headsets to mark vulnerable sections of a levee in a 3D scene. The marked information was synchronized in real time to the main venue's large screen. Experts at the main venue could click on the marked sections to view the levee's design standards and historical incident records.

[0054] The system records and archives the entire meeting process, generating structured minutes that include: the final adopted plan A, the model parameters used (the outflow from Reservoir A was adjusted from 35,000 m³ / s to 32,000 m³ / s), the decision-making basis (the analysis conclusions of the digital human evaluation model), and an animated deduction result. This information is packaged into a "Reservoir A scheduling case on a certain day of a certain month of a certain year" and stored in the knowledge base.

[0055] Experimental data show that after adopting this invention, the multi-source data fusion time is shortened from an average of 15 minutes to 2 minutes, the efficiency of contingency plan simulation is increased by 3 times, and the accuracy of remote experts' spatial cognition of the flood range is increased from 68% to 92%.

[0056] The above are merely preferred embodiments of the present invention. It should be noted that the present invention is not limited to the above embodiments. For those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A panoramic decision-making and consultation method for watershed flood control, characterized in that, The method includes: S1: Acquire multidimensional heterogeneous data of the watershed, unify the multidimensional heterogeneous data to a preset geographic coordinate system and time base, and generate a standardized time-series data stream; S2: Construct a three-dimensional digital twin environment of the watershed based on watershed geographic information data, and establish a data-driven mapping relationship between the three-dimensional digital twin environment and meteorological models, hydrological models and hydrodynamic models; S3: Collect flood control documents, historical dispatch cases, emergency plan texts, water conservancy engineering specifications and standards, and expert discussion records in the basin to build a professional corpus; use the professional corpus to adjust the pre-trained Transformer architecture large language model to obtain the host model; S4: Construct a water conservancy knowledge graph, input the water conservancy knowledge graph into multiple large language models for few-shot learning, and train multiple digital human evaluation models for hydrological forecasting, engineering scheduling and risk assessment respectively. S5: Input standardized time-series data streams into the watershed's three-dimensional digital twin environment to drive the hydrological and hydrodynamic models to perform real-time calculations, generating a dynamic projection of floods and droughts for the entire watershed; and combine this with a water resources knowledge graph to form a comprehensive decision-making scenario that includes real-time water and rainfall data, flood forecast curves, and historical disaster cases. S6: The host model analyzes the consultation topic corresponding to the current instruction, controls the virtual camera in the three-dimensional digital twin environment to focus on the target geographical area, and calls the corresponding digital human evaluation model based on the logic of water conservancy knowledge graph, historical case retrieval and physical model simulation results, combined with the overall decision-making situation, to generate quantitative scheduling plan suggestions; S7: Through the three-dimensional digital twin environment of the watershed, the generated quantitative scheduling scheme suggestions are simulated and dynamically extrapolated in real time to generate flood evolution paths, inundation ranges, water level process lines and heat maps of at-risk populations under different schemes; according to the preset optimization objectives, combined with the rule base or optimization algorithm, the scheduling parameters are adjusted to determine the optimal scheduling scheme.

2. The basin flood control panoramic decision-making and consultation method according to claim 1, characterized in that, S1 specifically includes: The acquired multidimensional heterogeneous data of the watershed is cleaned, verified and formatted; the multidimensional heterogeneous data includes real-time monitoring data of hydrological station network, meteorological forecast data, remote sensing satellite image data, water conservancy project operation data and socio-economic and population distribution data, and is accessed through standard API and Internet of Things protocol; By employing coordinate transformation and time synchronization algorithms, multidimensional heterogeneous data from different sources are unified under a preset geographic coordinate system and time reference, generating a standardized time-series data stream.

3. The basin flood control panoramic decision-making and consultation method according to claim 1, characterized in that, The watershed geographic information data includes digital elevation models, orthophotos, river system vector data, and building information models of the required water conservancy facilities.

4. The basin flood control panoramic decision-making and consultation method according to claim 3, characterized in that, The establishment of a data-driven mapping relationship between the three-dimensional digital twin environment and meteorological, hydrological, and hydrodynamic models specifically includes: The two-dimensional flood evolution model is solved based on the finite volume method or the finite difference method. The input parameter interface, calculation core and result output interface of the solved two-dimensional flood evolution model are encapsulated as RESTful API or gRPC service. The three-dimensional rendering engine calls the RESTful API or gRPC service through timed polling or event triggering mechanism to obtain real-time calculated flow field data and water level data, and converts them into texture data or geometric data that the shader can recognize, so as to realize the dynamic rendering of water flow direction, flow velocity distribution and flood range in the three-dimensional scene.

5. The basin flood control panoramic decision-making and consultation method according to claim 1, characterized in that, The construction of the water conservancy knowledge graph specifically includes: Using water conservancy projects, flood control scheduling indicators, and water resource management elements as entities, and the relationships between entities as edges, a water conservancy knowledge graph is obtained by organizing them according to the entity-relationship-attribute structure.

6. The basin flood control panoramic decision-making and consultation method according to claim 1, characterized in that, The training yielded several digital human evaluation models for hydrological forecasting, engineering scheduling, and risk assessment, specifically including: By associating key events in the real-time data stream with entities in the water resources knowledge graph, early warning information with semantic tags is generated. The early warning information is overlaid and rendered in a three-dimensional digital twin environment as a visual layer, forming a panoramic view of decision-making that includes real-time water and rainfall data, flood forecast curves, and historical disaster cases; The method of associating key events in the real-time data stream with entities in the water resources knowledge graph includes: The real-time water level data is compared with the corresponding defense standard entities of the dike in the water conservancy knowledge graph. When the water level exceeds the defense standard threshold, the inference engine traverses the protection area, population distribution and important facility entities associated with the dike in the water conservancy knowledge graph. Based on the inference results, natural language early warning information containing the specific impact range, the estimated population to be relocated and the pre-assessment of economic losses are generated, and the spatial anchoring position of the early warning information in the three-dimensional scene is dynamically marked.

7. The basin flood control panoramic decision-making and consultation method according to claim 1, characterized in that, The proposed scheduling scheme for the digital human evaluation model in S6 includes the following: The system simulates real-time scenarios using a three-dimensional digital twin environment of the watershed; simultaneously, it retrieves constraint rules, historical analogies, and impact assessment indicators related to the current simulation scenario and scheduling issues from the water conservancy knowledge graph; it then compares and performs logical calculations with the retrieved rules, cases, and indicators to generate quantitative analysis conclusions and scheduling scheme recommendations; finally, it uses speech synthesis technology to broadcast the conclusions in natural language and displays them through overlay visualization charts and three-dimensional annotations.

8. The basin flood control panoramic decision-making and consultation method according to claim 1, characterized in that, After the scheduling scheme suggestion is generated, it also includes: Historical plans are loaded, and the human-computer interaction system preloads a default plan library and provides a variety of historical plans by having experts set scheduling parameters on-site. Real-time scenario simulation and dynamic extrapolation are performed on various combinations of historical scenarios to generate flood evolution paths, inundation ranges, water level process lines, and heat maps of at-risk populations for various combinations of historical scenarios; The differences between the proposed scheduling scheme and various combinations of historical schemes in terms of preset optimization objectives are compared. The preset optimization objectives include flooded area, affected population, economic loss and response time. Based on the real-time scenario simulation and dynamic extrapolation results of the generated quantified scheduling scheme suggestions, the real-time scenario simulation and dynamic extrapolation results of various combinations of historical schemes, and the differences between the scheduling scheme suggestions and various combinations of historical schemes in terms of preset optimization objectives, the system automatically generates adjustment suggestions for scheduling parameters based on rule base, genetic algorithm or particle swarm algorithm, and determines the optimal scheduling scheme.

9. The basin flood control panoramic decision-making and consultation method according to claim 1, characterized in that, Also includes: The selected target scheduling scheme suggestions are automatically filled into the preset standard instruction template to generate a standardized and directly applicable scheduling command file. All input data, intermediate deduction processes, final solutions, and expert instructions in the current consultation process are associated and encapsulated to form a decision case package with semantic tags, which is then stored in the system knowledge base.

10. A computer system, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the basin flood control panoramic decision-making consultation method as described in any one of claims 1-9.

Images