Drainage basin intelligent flood control scheduling method and system based on digital twinning

Abstract

The invention discloses a drainage basin intelligent flood control scheduling method and system based on digital twinborn, and relates to the technical field of flood control and disaster mitigation, and the method comprises the steps: collecting static data and dynamic data of a drainage basin, building a hydrological and hydrodynamic coupling model based on the static data and the dynamic data, and forming a drainage basin digital twinborn body; inputting the received numerical weather forecast into the digital twin of the watershed for simulation, generating a plurality of flood routing scenes in a future time period, and calculating a dynamic flood risk probability graph; the method comprises the following steps: constructing a simulation training environment by using historical flood data and a high-precision drainage basin digital twinborn body, carrying out offline training on a scheduling strategy network in the simulation training environment based on a reinforcement learning algorithm, and outputting a scheduling instruction according to a real-time drainage basin state to complete training of the scheduling strategy network. According to the method, the core problem that the traditional method is insufficient in decision timeliness and weak in adaptive capacity in an uncertain environment is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of flood control and disaster reduction technology, and in particular to a watershed intelligent flood control scheduling method and system based on digital twins. Background Technology

[0002] With the deep integration of information technology in the water conservancy field, digital twin technology has provided a new paradigm for river basin flood control scheduling. Preliminary smart flood control systems integrating data acquisition, model simulation, and scheduling decision-making functions have emerged. By simulating the evolution of floods through computers, and based on the simulation results, mathematical programming algorithms or pre-set scheduling rules are used to generate scheduling schemes for reservoirs, dams, and other projects. To a certain extent, this has realized the transformation from pure experience-based scheduling to model-driven scheduling, and improved the digitalization level of flood control management.

[0003] However, the aforementioned existing technologies still face challenges in terms of the timeliness and adaptability of their decision-making processes when dealing with sudden and highly uncertain extreme flood events. This is mainly due to two interrelated aspects: First, although the optimization algorithm based on mathematical programming can obtain the theoretical optimal solution, its computational complexity is high. In scenarios involving the joint scheduling of large-scale water conservancy projects and requiring rapid response to weather forecast updates, the solution time may not meet the minute-level or even second-level timeliness requirements of emergency decision-making. Although the system can identify the uncertainty of risks, its data acquisition mode is usually static and uniformly distributed, failing to be deeply coupled with the dynamically evolving scheduling decision-making process. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a watershed intelligent flood control scheduling method based on digital twins to solve the problems of delayed decision-making due to insufficient computational efficiency of optimization algorithms in the prior art, and the limited adaptive optimization capability of the system under uncertain environments due to passive and fixed data acquisition.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a watershed intelligent flood control scheduling method based on digital twins, which includes collecting static and dynamic data of the watershed, and constructing a hydrological and hydrodynamic coupling model based on the static and dynamic data to form a watershed digital twin. The received numerical weather forecasts are input into the watershed digital twin for simulation, generating multiple flood evolution scenarios for future periods and calculating dynamic flood risk probability maps. A simulation training environment is constructed using historical flood data and a high-precision digital twin of the watershed. The scheduling strategy network is then trained offline in the simulation training environment based on a reinforcement learning algorithm. The trained scheduling strategy network outputs scheduling instructions based on the real-time watershed status. The current real-time watershed status is input into the trained scheduling policy network to obtain a preliminary scheduling scheme, and sensitivity factors are identified based on the preliminary scheduling scheme through sensitivity analysis. Based on the scheduling monitoring resources, the initial scheduling scheme is modified by targeted monitoring of the sensitivity factors to obtain an optimized scheduling scheme. The optimized scheduling scheme is executed, and the state of the watershed digital twin is corrected based on real-time monitoring data, and a rolling optimization is performed.

[0007] As a preferred embodiment of the intelligent flood control scheduling method for watersheds based on digital twins as described in this invention, the method includes: collecting static and dynamic data of the watershed, and constructing a hydrological-hydrodynamic coupling model based on the static and dynamic data to form a digital twin of the watershed, comprising the following steps: An integrated air-space-ground monitoring network, consisting of sensors deployed on satellite platforms, airborne platforms, and ground platforms, collects topographic elevation data, river cross-section data, land use type data, geometric model data of water conservancy projects, as well as real-time rainfall data, real-time water level data, and real-time flow data for the watershed. Using static and dynamic data, the parameters of the hydrological model describing the runoff generation and confluence process of the watershed and the hydrodynamic model describing the flood evolution process are calibrated and coupled together, and a hydro-hydrodynamic coupling model with a data exchange interface between the hydrological model and the hydrodynamic model is constructed. The hydrological and hydrodynamic coupling model, after parameter calibration and embedding a real-time driving interface, is defined as a watershed digital twin capable of mapping the physical watershed state.

[0008] As a preferred embodiment of the watershed intelligent flood control scheduling method based on digital twins described in this invention, the method includes the following steps: inputting received numerical weather forecasts into the watershed digital twin for simulation, generating multiple flood evolution scenarios for future periods, and calculating a dynamic flood risk probability map. The system receives numerical weather prediction products from multiple forecast members, and sequentially inputs the rainfall forecast data of each forecast member into the watershed digital twin for parallel simulation, generating multiple flood evolution scenarios corresponding to the number of forecast members. Based on various flood evolution scenarios, the probability that the water depth at each location within the basin exceeds the preset disaster-causing water depth at different times is statistically analyzed to generate a dynamic flood risk probability map.

[0009] As a preferred embodiment of the intelligent flood control scheduling method for watersheds based on digital twins described in this invention, the method includes: constructing a simulation training environment using historical flood data and a high-precision digital twin of the watershed; offline training of the scheduling strategy network in the simulation training environment based on a reinforcement learning algorithm; and outputting the trained scheduling strategy network with scheduling instructions based on the real-time watershed status. The method comprises the following steps: By using historical flood data to drive a calibrated, high-precision watershed digital twin, a massive flood scenario covering different rainfall patterns and initial conditions is generated, thus constructing a simulation training environment. The real-time watershed status includes network-wide water level data, areal rainfall data, and soil moisture data; Based on reinforcement learning algorithms, the scheduling policy network is trained offline in a constructed simulation training environment. The training objective is to maximize a long-term reward function that comprehensively considers flood control safety, water resource utilization, and scheduling stability. After offline training is completed, the training-completed scheduling policy network is generated by instantaneously outputting scheduling instructions based on the real-time watershed status.

[0010] As a preferred embodiment of the intelligent flood control scheduling method for watersheds based on digital twins described in this invention, the method includes the following steps: inputting the current real-time watershed status into the trained scheduling strategy network to obtain a preliminary scheduling scheme, and identifying sensitivity factors based on the preliminary scheduling scheme through sensitivity analysis. The current real-time watershed status of the entire network's water level data, areal rainfall data, and soil moisture data is input into the trained scheduling strategy network to obtain a preliminary scheduling scheme output by the trained scheduling strategy network. Based on the preliminary scheduling scheme, sensitivity analysis was conducted by changing the input parameters to identify the sensitive factors affecting the final flood control effect of the preliminary scheduling scheme.

[0011] As a preferred embodiment of the watershed intelligent flood control scheduling method based on digital twins described in this invention, the following steps are included: Based on scheduling monitoring resources, the preliminary scheduling scheme is modified by targeted monitoring of the sensitive factors to obtain an optimized scheduling scheme: Use or move monitoring stations to monitor resources, enhance monitoring of the spatial locations corresponding to the identified sensitive factors, and obtain measured data of the sensitive factors; Using the measured data of the sensitivity factor, the corresponding data items in the current real-time watershed state are corrected to form the corrected real-time watershed state; The corrected real-time watershed state is then input back into the trained scheduling policy network to obtain the optimized scheduling scheme output by the trained scheduling policy network.

[0012] As a preferred embodiment of the intelligent flood control scheduling method for watersheds based on digital twins as described in this invention, the following steps are included: executing the optimized scheduling scheme, correcting the state of the watershed digital twin based on real-time monitoring data, and performing rolling optimization: The optimized scheduling plan will be distributed to the reservoir and sluice gate water conservancy project control units within the basin for execution. Real-time monitoring data of the watershed is collected after the water conservancy project control unit executes the optimized scheduling scheme, and the internal state variables of the watershed digital twin are corrected by data assimilation algorithm using the real-time monitoring data. Based on the corrected digital twin of the watershed, after waiting for a fixed time interval, the process restarts from generating multiple flood evolution scenarios for future periods, forming a closed-loop rolling optimization.

[0013] Secondly, the present invention provides a watershed intelligent flood control scheduling system based on digital twins, including a data fusion module, which collects static and dynamic data of the watershed, and constructs a hydrological and hydrodynamic coupling model based on the static and dynamic data to form a watershed digital twin. The simulation engine module inputs the received numerical weather forecasts into the watershed digital twin for simulation, generates multiple flood evolution scenarios for future periods, and calculates dynamic flood risk probability maps. The decision-making module uses historical flood data and a high-precision digital twin of the watershed to construct a simulation training environment, and performs offline training on the scheduling strategy network in the simulation training environment based on reinforcement learning algorithms. Based on the real-time watershed status, it outputs the training completed scheduling strategy network with scheduling instructions. The monitoring and optimization module inputs the current real-time watershed status into the trained scheduling strategy network to obtain a preliminary scheduling scheme, and identifies sensitive factors based on the preliminary scheduling scheme through sensitivity analysis. The execution control module performs targeted monitoring and correction of the preliminary scheduling scheme based on the scheduling monitoring resources to obtain an optimized scheduling scheme; The optimization control module executes the optimized scheduling scheme, corrects the state of the watershed digital twin based on real-time monitoring data, and performs rolling optimization.

[0014] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the watershed intelligent flood control scheduling method based on digital twins as described in the first aspect of the present invention.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the watershed smart flood control scheduling method based on digital twins as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: by constructing an integrated air-space-ground monitoring network and a hydrological and hydrodynamic coupling model to form a high-fidelity digital twin of the watershed, and using numerical weather prediction to generate multi-scenario flood evolution risk probability maps; innovatively adopting a reinforcement learning offline training scheduling strategy network to achieve second-level decision response, and combining sensitivity analysis to dynamically identify sensitive factors for targeted monitoring and optimization, forming a closed-loop intelligent scheduling system of monitoring-simulation-decision-verification-rolling optimization, effectively solving the core problems of insufficient decision-making timeliness and weak adaptive ability under uncertain environments in traditional methods. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a watershed intelligent flood control scheduling method based on digital twins.

[0019] Figure 2 This is a schematic diagram of a watershed intelligent flood control scheduling system based on digital twins. Detailed Implementation

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0023] Reference Figure 1 and Figure 2 As one embodiment of the present invention, this embodiment provides a watershed intelligent flood control scheduling method based on digital twins, comprising the following steps: S1. Collect static and dynamic data of the watershed, and construct a hydrological and hydrodynamic coupling model based on the static and dynamic data to form a digital twin of the watershed.

[0024] S1.1. Through an integrated air-space-ground monitoring network consisting of sensors deployed on satellite platforms, air platforms, and ground platforms, topographic elevation data, river cross-section data, land use type data, geometric model data of water conservancy projects, as well as real-time rainfall data, real-time water level data, and real-time flow data of the watershed are collected.

[0025] Furthermore, optical and radar sensors deployed on satellite platforms are responsible for collecting large-scale digital elevation models and land use type data; UAV lidar deployed on aerial platforms are responsible for collecting high-precision river cross-section data and hydraulic engineering geometric model data; water level gauges, rain gauges, and flow meters deployed on ground platforms are responsible for continuously collecting real-time rainfall data, real-time water level data, and real-time flow data; the data obtained from the integrated air-space-ground monitoring network together constitute the static and dynamic data of the watershed.

[0026] S1.2 Using static and dynamic data, the parameters of the hydrological model describing the runoff generation and confluence process of the watershed and the hydrodynamic model describing the flood evolution process are calibrated and coupled together to construct a hydro-hydrodynamic coupling model with a data exchange interface between the hydrological model and the hydrodynamic model.

[0027] Furthermore, an automatic optimization algorithm is used to calibrate the runoff generation and confluence parameters of the hydrological model and the channel roughness parameters of the hydrodynamic model. After calibration, the flood hydrograph at the channel inlet section calculated by the hydrological model is used as the upstream boundary condition of the hydrodynamic model, while the backwater level calculated by the hydrodynamic model is used as the feedback of the downstream boundary condition of the hydrological model. In this way, a hydro-hydrodynamic coupling model with a two-way data exchange interface between the hydrological model and the hydrodynamic model is constructed.

[0028] S1.3 Define the hydrological and hydrodynamic coupling model, which has been calibrated with parameters and embedded in the real-time driving interface, as a watershed digital twin that can map the physical watershed state.

[0029] Furthermore, the hydrological and hydrodynamic coupling model, which has been parameter-calibrated and configured with an interface capable of receiving real-time rainfall data and real-time water level data, is instantiated as a watershed digital twin. This watershed digital twin is regarded as a virtual mapping of the physical watershed state.

[0030] S2. Input the received numerical weather forecast into the watershed digital twin for simulation, generate multiple flood evolution scenarios for future periods, and calculate a dynamic flood risk probability map.

[0031] S2.1 Receive numerical weather prediction products from multiple forecast members, and input the rainfall forecast data of each forecast member in the numerical weather prediction products into the watershed digital twin in sequence for parallel simulation to generate multiple flood evolution scenarios corresponding to the number of forecast members.

[0032] Furthermore, numerical weather prediction products containing multiple forecast members are received from the meteorological data server, and grid precipitation forecast data for each forecast member in a specific future time period are extracted. The grid precipitation forecast data of each forecast member is used as input to drive the parallel simulation of the watershed digital twin. Each simulation operation corresponds to one forecast member, and the watershed digital twin outputs the flood evolution process under the scenario of that member, including the water depth and flow rate changes over time at various locations in the watershed. After the simulation of all forecast members is completed, multiple flood evolution scenarios are obtained, equal to the number of forecast members.

[0033] S2.2 Based on multiple flood evolution scenarios, the probability that the water depth at each location in the basin exceeds the preset disaster-causing water depth at different times is statistically analyzed, and a dynamic flood risk probability map is generated.

[0034] The expression for the dynamic flood risk probability diagram is:

[0035] in, For a moment The following is a dynamic flood risk probability map. For a moment, For the watershed area, This represents the dynamic probability of flood risk. Spatial location Direction coordinates Spatial location Direction coordinates.

[0036] Furthermore, for each grid cell within the basin at each future time, the number of flood evolution scenarios in which the water depth of that grid cell exceeds the preset disaster-causing water depth for that cell is counted. This number is then divided by the total number of flood evolution scenarios to obtain the dynamic flood risk probability of that grid cell at that time. By traversing all grid cells within the basin and all future times of interest, all dynamic flood risk probability values ​​are calculated and aggregated, thus generating a dynamic flood risk probability map that changes over time.

[0037] S3. Construct a simulation training environment using historical flood data and a high-precision digital twin of the watershed, and conduct offline training of the scheduling strategy network in the simulation training environment based on reinforcement learning algorithm. Output the trained scheduling strategy network according to the real-time watershed status.

[0038] S3.1. Using historical flood data to drive a calibrated high-precision watershed digital twin, a massive flood scenario covering different rainfall patterns and initial conditions is generated to construct a simulation training environment.

[0039] Furthermore, typical rainstorm processes and extreme rainfall events recorded in historical flood data are used as input to drive a high-precision watershed digital twin that has undergone rigorous parameter calibration and verification to perform repeated simulations. In each simulation, different historical rainfall sequences are randomly combined as inputs, and different initial soil moisture and initial river water levels are set as initial conditions to simulate and generate a massive number of synthetic flood scenarios covering various possibilities. This series of massive flood scenarios and their corresponding watershed evolution processes together constitute a simulation training environment for training the algorithm.

[0040] S3.2 The real-time watershed status includes network-wide water level data, areal rainfall data, and soil moisture data.

[0041] Furthermore, the real-time watershed status is defined as a state vector containing specific data dimensions. Specifically, the state vector includes real-time water level data of all key nodes in the watershed obtained from the monitoring network, i.e., the whole network water level data, areal rainfall data representing the average rainfall of the watershed surface, and soil moisture data reflecting the dryness and wetness of the underlying surface of the watershed.

[0042] S3.3 Based on reinforcement learning algorithm, the scheduling strategy network is trained offline in the constructed simulation training environment. The training objective is to maximize a long-term reward function that comprehensively considers flood control safety, water resource utilization and scheduling stability. After offline training is completed, the training completed scheduling strategy network is generated by instantaneously outputting scheduling instructions based on the real-time watershed status.

[0043] Furthermore, a neural network is initialized as the scheduling strategy network, taking real-time watershed status as input and water conservancy project scheduling instructions as output. A near-end policy optimization reinforcement learning algorithm is employed, with the scheduling strategy network as the agent and a simulation training environment as the interactive environment for training. During training, the agent outputs scheduling instructions based on the current real-time watershed status and applies them to the simulation training environment. The environment then calculates to the next state and feeds back a reward value, which is calculated using a long-term reward function. The specific calculation of the long-term reward function comprehensively considers the flood control safety represented by the degree of downstream control point water level exceeding limits during the simulation period, the water resource utilization represented by the reservoir storage at the end of the scheduling period, and the scheduling stability represented by the change in discharge volume. The training process continues until the performance of the scheduling strategy network converges, ultimately obtaining a trained scheduling strategy network capable of instantly outputting corresponding scheduling instructions based on the real-time watershed status.

[0044] S4. Input the current real-time watershed status into the trained scheduling policy network to obtain a preliminary scheduling scheme, and identify the sensitivity factors based on the preliminary scheduling scheme through sensitivity analysis.

[0045] S4.1 Input the current real-time watershed status of the entire network water level data, areal rainfall data and soil moisture data into the trained scheduling strategy network to obtain the preliminary scheduling scheme output by the trained scheduling strategy network.

[0046] Furthermore, the latest real-time watershed status, including water level data, areal rainfall data, and soil moisture data of the entire network, is obtained from the data acquisition and fusion unit. This real-time watershed status data is combined into a vector that meets the input format requirements and directly input into the training-completed scheduling strategy network that has completed offline training. After internal forward propagation calculation, the training-completed scheduling strategy network outputs a set of suggested scheduling instruction sequences corresponding to each water conservancy project in the future period. This set of instruction sequences constitutes the preliminary scheduling scheme.

[0047] S4.2 Based on the preliminary scheduling scheme, sensitivity analysis is conducted by changing the input parameters to identify the sensitive factors affecting the final flood control effect of the preliminary scheduling scheme.

[0048] Furthermore, a local sensitivity analysis method is employed to fine-tune specific parameters in the current real-time watershed status. For example, a small perturbation is added to the areal rainfall data of a certain sub-region upstream while keeping all other parameters unchanged. The perturbed real-time watershed status is then input into the watershed digital twin for rapid simulation, and the simulated peak water level change at key downstream control points is calculated. The ratio of this peak water level change to the parameter perturbation is used as the local sensitivity index of the parameter. All possible uncertain input parameters and sensitivity indices are iterated, and the parameter or several parameters with the largest absolute value of the sensitivity index are identified as the most sensitive factors that have the greatest impact on the final flood control effect of the preliminary scheduling plan.

[0049] S5. Based on the scheduling monitoring resources, the initial scheduling scheme is modified by targeted monitoring of the sensitivity factors to obtain an optimized scheduling scheme.

[0050] S5.1 Use or move monitoring stations to monitor resources, enhance monitoring of the spatial locations corresponding to the identified sensitive factors, and obtain measured data of the sensitive factors.

[0051] Furthermore, based on the identified sensitive factor types and their corresponding geospatial locations, instructions are issued to mobile monitoring resources such as drones or mobile monitoring stations. The instructions require high-frequency and high-precision enhanced monitoring of the spatial locations corresponding to the sensitive factors. Drones equipped with radar rainfall measurement equipment fly to the airspace of designated sub-basins to conduct vertical detection to obtain measured values ​​of areal rainfall data for the area, or mobile monitoring stations are deployed to key river sections to conduct continuous water level and flow monitoring. Through enhanced monitoring, measured data of the sensitive factors at the current moment are obtained.

[0052] S5.2 Using the measured data of the sensitivity factor, correct the corresponding data items in the current real-time watershed state to form the corrected real-time watershed state.

[0053] Furthermore, a data assimilation algorithm is used to optimally fuse the measured data with the original monitoring data, thereby updating the value of the sensitivity factor in the current real-time watershed status and forming a corrected real-time watershed status that is closer to the actual watershed conditions.

[0054] S5.3. The corrected real-time watershed state is then input back into the trained scheduling policy network to obtain the optimized scheduling scheme output by the trained scheduling policy network.

[0055] Furthermore, the corrected real-time watershed state is used as a new input and fed back into the trained scheduling strategy network. Based on this updated, less uncertain corrected real-time watershed state, the trained scheduling strategy network performs a new round of forward propagation calculations and outputs a new set of water conservancy project scheduling instructions. This new set of scheduling instructions is the optimized scheduling scheme obtained after fully considering key uncertainty information.

[0056] S6. Execute the optimized scheduling scheme, and correct the state of the watershed digital twin based on real-time monitoring data, and perform rolling optimization.

[0057] S6.1. The optimized scheduling plan is issued to the reservoir and sluice gate water conservancy project control units within the basin for execution.

[0058] Furthermore, the specific control instructions included in the optimized scheduling scheme, such as the reservoir discharge setpoint or the gate opening height, are transmitted to the corresponding reservoir control unit and gate control unit within the basin via the communication network. After receiving the instructions, the reservoir control unit and gate control unit automatically execute them, thereby changing the actual water flow state.

[0059] S6.2 Collect real-time monitoring data of the watershed after the water conservancy project control unit executes the optimized scheduling scheme, and use the real-time monitoring data to correct the internal state variables of the watershed digital twin through a data assimilation algorithm.

[0060] Furthermore, after the optimized scheduling scheme is implemented, real-time rainfall data, real-time water level data, and real-time flow data of each station in the basin are continuously collected through an integrated air-space-ground monitoring network. Using these newly collected real-time monitoring data, the real-time monitoring data is fused with the simulated forecast values ​​of the basin's digital twin through a Kalman filter data assimilation algorithm. The internal state variables of the basin's digital twin, the flow values ​​of each river channel, and the water content values ​​of each soil zone are dynamically adjusted and corrected, so that the state of the basin's digital twin is closer to the real state of the physical basin.

[0061] S6.3. Based on the corrected digital twin of the watershed, after waiting for a fixed time interval, restart the process from generating multiple flood evolution scenarios in the future period to form a closed-loop rolling optimization.

[0062] Furthermore, after completing the state correction of the watershed digital twin, a fixed time interval, such as one hour, is waited for. When the next decision cycle arrives, the process restarts, using the latest corrected watershed digital twin and the latest numerical weather prediction products to regenerate various flood evolution scenarios for future periods, and then executing all subsequent steps in sequence. This periodic repetitive execution mechanism constitutes a closed-loop rolling optimization process, ensuring that the scheduling strategy can continuously adapt to the dynamic changes in the watershed state.

[0063] This embodiment also provides a watershed intelligent flood control scheduling system based on digital twins, including: a data fusion module, which collects static and dynamic data of the watershed, and constructs a hydrological and hydrodynamic coupling model based on the static and dynamic data to form a watershed digital twin; The simulation engine module inputs the received numerical weather forecasts into the watershed digital twin for simulation, generates multiple flood evolution scenarios for future periods, and calculates dynamic flood risk probability maps. The decision-making module uses historical flood data and a high-precision digital twin of the watershed to construct a simulation training environment, and performs offline training on the scheduling strategy network in the simulation training environment based on reinforcement learning algorithms. Based on the real-time watershed status, it outputs the training completed scheduling strategy network with scheduling instructions. The monitoring and optimization module inputs the current real-time watershed status into the trained scheduling strategy network to obtain a preliminary scheduling scheme, and identifies sensitive factors based on the preliminary scheduling scheme through sensitivity analysis. The execution control module performs targeted monitoring and correction of the preliminary scheduling scheme based on the scheduling monitoring resources to obtain an optimized scheduling scheme; The optimization control module executes the optimized scheduling scheme, corrects the state of the watershed digital twin based on real-time monitoring data, and performs rolling optimization.

[0064] This embodiment also provides a computer device applicable to the case of a watershed intelligent flood control scheduling method based on digital twins, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the watershed intelligent flood control scheduling method based on digital twins as proposed in the above embodiment.

[0065] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0066] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the intelligent flood control scheduling method for watersheds based on digital twins as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0067] In summary, this invention constructs a high-fidelity digital twin of a watershed by building an integrated air-space-ground monitoring network and a hydrological and hydrodynamic coupling model, and generates multi-scenario flood evolution risk probability maps using numerical weather prediction. It innovatively adopts a reinforcement learning offline training scheduling strategy network to achieve second-level decision response, and combines sensitivity analysis to dynamically identify sensitive factors for targeted monitoring and optimization, forming a closed-loop intelligent scheduling system of monitoring-simulation-decision-verification-rolling optimization. This effectively solves the core problems of insufficient decision-making timeliness and weak adaptability under uncertain environments in traditional methods.

[0068] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A digital-twin-based intelligent flood control scheduling method for a river basin, characterized in that: This includes collecting static and dynamic data of the watershed, and constructing a hydrological and hydrodynamic coupling model based on the static and dynamic data to form a digital twin of the watershed; The received numerical weather forecasts are input into the watershed digital twin for simulation, generating multiple flood evolution scenarios for future periods and calculating dynamic flood risk probability maps. A simulation training environment is constructed using historical flood data and a high-precision digital twin of the watershed. The scheduling strategy network is then trained offline in the simulation training environment based on a reinforcement learning algorithm. The trained scheduling strategy network outputs scheduling instructions based on the real-time watershed status. The current real-time watershed status is input into the trained scheduling policy network to obtain a preliminary scheduling scheme, and sensitivity factors are identified based on the preliminary scheduling scheme through sensitivity analysis. Based on the scheduling monitoring resources, the initial scheduling scheme is modified by targeted monitoring of the sensitivity factors to obtain an optimized scheduling scheme. The optimized scheduling scheme is executed, and the state of the watershed digital twin is corrected based on real-time monitoring data, and a rolling optimization is performed.

2. The digital-twin-based basin intelligent flood control scheduling method of claim 1, wherein: Collecting static and dynamic data of the watershed, and constructing a hydrological-hydrodynamic coupling model based on the static and dynamic data to form a digital twin of the watershed, includes the following steps: An integrated air-space-ground monitoring network, consisting of sensors deployed on satellite platforms, airborne platforms, and ground platforms, collects topographic elevation data, river cross-section data, land use type data, geometric model data of water conservancy projects, as well as real-time rainfall data, real-time water level data, and real-time flow data for the watershed. Using static and dynamic data, the parameters of the hydrological model describing the runoff generation and confluence process of the watershed and the hydrodynamic model describing the flood evolution process are calibrated and coupled together, and a hydro-hydrodynamic coupling model with a data exchange interface between the hydrological model and the hydrodynamic model is constructed. The hydrological and hydrodynamic coupling model, after parameter calibration and embedding a real-time driving interface, is defined as a watershed digital twin capable of mapping the physical watershed state.

3. The digital-twin-based basin intelligent flood control scheduling method of claim 2, wherein: The received numerical weather forecasts are input into the watershed digital twin for simulation to generate various flood evolution scenarios for future periods and calculate a dynamic flood risk probability map, including the following steps: The system receives numerical weather prediction products from multiple forecast members, and sequentially inputs the rainfall forecast data of each forecast member into the watershed digital twin for parallel simulation, generating multiple flood evolution scenarios corresponding to the number of forecast members. Based on various flood evolution scenarios, the probability that the water depth at each location within the basin exceeds the preset disaster-causing water depth at different times is statistically analyzed to generate a dynamic flood risk probability map.

4. The digital-twin-based basin intelligent flood control scheduling method of claim 3, wherein: A simulation training environment is constructed using historical flood data and a high-precision digital twin of the watershed. A scheduling strategy network is then trained offline within this environment using a reinforcement learning algorithm. Based on the real-time watershed status, the trained scheduling strategy network outputs scheduling instructions. The process includes the following steps: A calibrated high-precision basin digital twin is driven by historical flood data to simulate a large number of flood scenarios covering different rainfall patterns and initial conditions, and a simulation training environment is constructed; The real-time basin state includes network water level data, surface rainfall data and soil moisture data; Based on the reinforcement learning algorithm, the dispatching strategy network is trained offline in the constructed simulation training environment, and the training target is to maximize a long-term reward function that comprehensively considers flood control safety, water resource utilization and dispatching stability. After offline training, the trained dispatching strategy network instantaneously outputs dispatching instructions according to the real-time basin state.

5. The digital-twin-based basin intelligent flood control scheduling method of claim 4, wherein: The current real-time basin state is input into the trained dispatching strategy network to obtain a preliminary dispatching scheme, and based on the preliminary dispatching scheme, sensitive factors are identified through sensitivity analysis, including the following steps: The current real-time basin state of the network water level data, surface rainfall data and soil moisture data is input into the trained dispatching strategy network to obtain the preliminary dispatching scheme output by the trained dispatching strategy network; Based on the preliminary dispatching scheme, sensitivity analysis is performed by changing the input parameters to identify sensitive factors that affect the final flood control effect of the preliminary dispatching scheme.

6. The digital-twin-based basin intelligent flood control scheduling method of claim 5, wherein: Based on the dispatching monitoring resources, the sensitive factors are monitored and the preliminary dispatching scheme is corrected to obtain an optimized dispatching scheme, including the following steps: Using or moving monitoring stations, the spatial positions corresponding to the identified sensitive factors are intensively monitored to obtain measured data of the sensitive factors; Using the measured data of the sensitive factors, the corresponding data items in the current real-time basin state are corrected to form a corrected real-time basin state; The corrected real-time basin state is input into the trained dispatching strategy network again to obtain the optimized dispatching scheme output by the trained dispatching strategy network.

7. The digital-twin-based basin intelligent flood control scheduling method of claim 6, wherein: The optimized dispatching scheme is executed, and the state of the basin digital twin is corrected based on real-time monitoring data, and a rolling optimization is performed, including the following steps: The optimized dispatching scheme is sent to the reservoir and gate water conservancy engineering control unit in the basin for execution; Real-time monitoring data of the basin after the water conservancy engineering control unit executes the optimized dispatching scheme is collected, and the internal state variables of the basin digital twin are corrected using the real-time monitoring data through a data assimilation algorithm; Based on the corrected basin digital twin, after a fixed time interval, the process of generating multiple flood evolution scenarios in the future period is restarted, forming a closed-loop rolling optimization.

8. A digital-twin-based intelligent flood control scheduling system for a river basin, based on the digital-twin-based intelligent flood control scheduling method of any one of claims 1-7. The data fusion module collects static and dynamic data of the basin, and constructs a hydrological and hydrodynamic coupled model based on the static and dynamic data to form a basin digital twin; The simulation engine module inputs the received numerical weather prediction into the basin digital twin for simulation to generate multiple flood evolution scenarios in the future period and calculate a dynamic flood risk probability map; The decision module constructs a simulation training environment by using historical flood data and a high-precision digital basin twin, and performs offline training on a scheduling strategy network in the simulation training environment based on a reinforcement learning algorithm, and outputs a trained scheduling strategy network according to a real-time basin state; The monitoring optimization module inputs a current real-time basin state into the trained scheduling strategy network to obtain a preliminary scheduling scheme, and identifies a sensitivity factor through sensitivity analysis based on the preliminary scheduling scheme; The execution control module performs targeted monitoring on the sensitivity factor based on scheduling monitoring resources, corrects the preliminary scheduling scheme, and obtains an optimized scheduling scheme; The optimization control module executes the optimized scheduling scheme, corrects the state of the digital basin twin based on real-time monitoring data, and performs rolling optimization. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to implement the steps of the digital-twin-based basin intelligent flood control scheduling method of any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the digital-twin-based basin intelligent flood control scheduling method of any one of claims 1-7.

Images