Ecological water replenishing regulation and control method and system for Yellow River basin

Abstract

The invention relates to the technical field of water resource scheduling, in particular to a Yellow River basin ecological water supplementing regulation and control method and system. The method comprises the following steps: deploying sensor units at key ecological nodes of the Yellow River basin, and collecting Yellow River ecological data; performing ecological water demand fingerprint extraction on the Yellow River ecological data to obtain ecological water demand fingerprint data; obtaining Yellow River hydrological data, and constructing a Yellow River digital twin river channel model according to the Yellow River hydrological data to obtain the Yellow River digital twin river channel model; constructing a water replenishing unit according to the ecological water demand fingerprint data and the Yellow River digital twin river channel model to obtain water replenishing unit data; and performing particle routing simulation according to the water replenishing unit data to obtain water replenishing simulation scheduling data. By constructing a coupling mechanism of the ecological water demand fingerprint and the digital twin river channel model, accurate identification and dynamic simulation of the ecological water use state of the Yellow River basin are realized, and the ecological response suitability of water replenishing path planning and the intelligence of regulation and control decision are improved.

Description

Technical Field

[0001] This invention relates to the field of water resource allocation technology, and in particular to a method and system for ecological water replenishment regulation in the Yellow River Basin. Background Technology

[0002] Under the current strategy of ecological protection and high-quality development in the Yellow River Basin, ecological water replenishment has become one of the important means to ensure the healthy operation of the river. Traditional water replenishment regulation methods mostly rely on static hydrological parameters and human experience-based decision-making, lacking dynamic perception and precise adaptation to local ecological response mechanisms, and are unable to cope with the challenges of seasonal changes, ecological factors, and contradictions in water resource allocation. At the same time, existing ecological scheduling methods generally lack the ability to model the coupling relationship between hydrological behavior and ecological factors, resulting in phenomena such as "ineffective replenishment" or "over-replenishment" in the water replenishment process, which neither fully guarantees the basic ecological water demand nor avoids wasting limited water resources. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes a method and system for ecological water replenishment regulation in the Yellow River Basin, thereby resolving at least one of the aforementioned technical issues.

[0004] This application provides a method for ecological water replenishment and regulation in the Yellow River Basin, including the following steps:

[0005] Step S1: Deploy sensor units at key ecological nodes in the Yellow River Basin to collect Yellow River ecological data; extract ecological water demand fingerprints from the Yellow River ecological data to obtain ecological water demand fingerprint data;

[0006] Step S2: Obtain Yellow River hydrological data and construct a digital twin river channel model of the Yellow River based on the Yellow River hydrological data to obtain the Yellow River digital twin river channel model;

[0007] Step S3: Construct water replenishment units based on ecological water demand fingerprint data and the Yellow River digital twin channel model to obtain water replenishment unit data;

[0008] Step S4: Perform particle routing simulation based on the water replenishment unit data to obtain water replenishment simulation scheduling data.

[0009] This invention deploys multiple types of sensors at key ecological nodes in the Yellow River Basin to achieve dynamic perception of ecological factors, hydrological characteristics, and environmental conditions. Based on the collected data, it extracts ecological water demand fingerprints to accurately characterize the time-varying response needs of different ecological units to water resources. By constructing a digital twin river model that integrates topography, hydrodynamics, and ecological response mechanisms, it realistically recreates the water conveyance path and ecological coupling structure within the basin, improving the spatial accuracy of the scheduling scheme. By particleizing ecological demand, decoupling path simulation from the scheduling process, it forms a water replenishment unit that can be optimized in parallel, achieving efficient allocation of flow resources and path simulation under ecological drive. This generates scheduling data with ecological response targeting and execution feasibility, effectively improving the intelligence level and ecological restoration efficiency of the water transfer system.

[0010] Preferably, the Yellow River ecological data collection specifically includes:

[0011] Hydrological sensors were deployed at key ecological nodes in the Yellow River Basin to acquire hydrological data;

[0012] Soil moisture sensors were deployed at key ecological nodes in the Yellow River Basin to acquire soil moisture data;

[0013] Deploy environmental sensors at key ecological nodes in the Yellow River Basin to acquire Yellow River environmental data;

[0014] By integrating hydrological data, soil hydrological data, and Yellow River environmental data, we obtain Yellow River ecological data.

[0015] This invention achieves precise monitoring of multiple ecological factors within the Yellow River basin by deploying hydrological sensors, soil moisture sensors, and environmental sensors at key ecological nodes. It enables the simultaneous collection of raw data reflecting hydrodynamic changes, soil moisture conditions, and climate fluctuations at different time scales. Through spatiotemporal collaborative data integration and processing, the completeness and responsiveness of ecological data are effectively improved, providing high-resolution and real-time support for accurate modeling of ecological water demand.

[0016] Preferably, the ecological water demand fingerprint extraction specifically involves:

[0017] By performing river segment geographic mapping on the Yellow River ecological data, river segment mapping data is obtained;

[0018] Lightweight behavior-triggered identification is performed based on river section mapping data to obtain hydrological behavior data;

[0019] Water fingerprint data is obtained by extracting water fingerprint data based on hydrological behavior data.

[0020] Ecological water demand fingerprint data is obtained by performing ecological water demand fingerprint clustering and coding based on water fingerprint data.

[0021] This invention employs segment-level geographic mapping of Yellow River ecological data, giving ecological observation data a clear spatial attribution and enhancing the correspondence between data and actual river channel structure. Combining the segment mapping results with a behavior-triggered identification mechanism, key hydrological behavior segments can be automatically extracted from continuous monitoring data, reducing manual intervention and improving the real-time performance and accuracy of event response. Furthermore, water fingerprinting technology is used to extract high-dimensional vector representations reflecting hydrological behavior characteristics, achieving a patterned representation of ecological water demand processes. By generating structured ecological water demand fingerprint data through clustering coding, not only is the comparability and classification of water demand characteristics across different ecological units enhanced, but a quantifiable and traceable ecological expression basis is also provided for water replenishment scheduling strategies, significantly improving the scientific rigor and ecological adaptability of water transfer schemes.

[0022] Preferably, water is used for fingerprint extraction, specifically as follows:

[0023] Hydrological tension modulus decomposition is performed based on hydrological behavior data to obtain tension regulation characteristic data;

[0024] A variational modal adversarial network model is constructed based on tension regulation characteristic data;

[0025] Modal residual learning is performed based on the variational modal adversarial network model to obtain natural impulse data and artificial impulse data;

[0026] Causal source modeling was performed based on natural pulse data and artificial pulse data to obtain hydrological behavior causal diagram data;

[0027] Modal feature compression was performed on the hydrological behavior causal graph data to obtain water fingerprint data.

[0028] This invention employs hydrological tension mode decomposition to extract nonlinear coupling and tension response characteristics among elements such as flow velocity, flow rate, and water level, effectively characterizing the dynamic evolution trend of hydrological behavior. The variational modal adversarial network model constructed in this invention possesses the ability to identify differences between naturally driven and artificially regulated behaviors, achieving modal separation while maintaining data generation consistency. By combining the natural and artificial impulse data separated through modal residual learning with causal attribution modeling, the influence paths and transmission mechanisms behind water diversion operations and rainfall events can be revealed. Through modal feature compression to form water fingerprint data, a structured expression and compressed encoding of hydrological driving patterns are achieved. This not only enhances the granularity of behavior recognition in scheduling modeling but also provides more interpretable and decision-supporting foundational data for the intelligent matching of ecological water replenishment schemes.

[0029] Preferably, step S2 specifically includes:

[0030] Obtain hydrological data of the Yellow River;

[0031] Based on the Yellow River hydrological data, the geometric structure of the river channel was reconstructed to obtain a spatial model of the river channel.

[0032] By performing attribute mapping on the river channel spatial model, a river channel mapping model is obtained;

[0033] Ecological factors were injected into the river channel mapping model to obtain a digital twin river channel model of the Yellow River.

[0034] This invention utilizes multi-source hydrological data from the Yellow River Basin to establish a spatiotemporal modeling foundation for key parameters such as flow rate, water level, and velocity. By reconstructing the river channel geometry using hydrological data, the invention accurately recreates geomorphic units such as the main channel, tributaries, floodplains, and cross-sections, forming a river spatial model with topological connectivity and spatial scale constraints. Attribute mapping is then implemented on this spatial model, injecting static and dynamic attributes such as hydraulic boundary conditions, basin classification, and water conveyance capacity into the structural units, thus constructing a semantically complete river channel mapping model. Ecological factors such as vegetation type, aquatic organism distribution, and ecological red line zones are superimposed to form a digital twin river channel model of the Yellow River with ecological response capabilities. This model not only supports simulation of water replenishment paths under real hydrological conditions but also dynamically responds to changes in ecological water demand, improving the spatial accuracy and target targeting of ecological scheduling.

[0035] Preferably, step S3 specifically includes:

[0036] Ecological response point data is obtained by matching ecological water demand fingerprint data with ecological response points in the digital twin river channel model of the Yellow River.

[0037] Water replenishment coupling points are identified based on ecological response point data to obtain water replenishment coupling point data;

[0038] Based on the water replenishment coupling point data, water replenishment particle units are constructed to obtain water replenishment particle candidate data;

[0039] The candidate data for water replenishment particles are encapsulated with attributes to obtain the water replenishment particle data;

[0040] Based on the water replenishment particle data, the potential field is used to find the water replenishment path data.

[0041] Water element dissipation cloud map calculation is performed on the water replenishment path data to obtain water replenishment unit data.

[0042] This invention uses spatial semantic matching between ecological water demand fingerprint data and a digital twin river model to accurately locate ecological response points, achieving precise mapping of ecological needs within the river system. Subsequently, it identifies water replenishment coupling points that are hydraulically connected and dispatchably accessible to the response points, constructing an ecologically driven water resource intervention interface. Water replenishment particle units constructed based on these coupling points granularize and structure water replenishment behavior, facilitating independent dispatching and path optimization in complex watershed structures. By encapsulating the attributes of these particle units and introducing multi-dimensional information such as ecological benefit weights, water transfer costs, and response time windows, it forms adjustable, comparable, and evaluable water replenishment task entities. Through potential field pathfinding, driven by ecological response intensity, hydraulic accessibility, and dispatching priority, it seeks the optimal ecological water replenishment path, effectively avoiding path congestion and water waste. Water element dissipation cloud map calculations spatially quantify evaporation, leakage, and time delay losses along the path, evaluating the actual ecological water replenishment efficiency and selecting efficient path combinations to ensure the optimal spatial distribution, energy utilization, and ecological response of the water replenishment process.

[0043] Preferably, the pathfinding in the potential field is specifically as follows:

[0044] Multi-level energy potential processing is performed on the water replenishment particle data to obtain multi-level energy potential data, which includes path dissipation energy potential processing and water source scheduling accessibility processing.

[0045] Potential energy field grid data is obtained by mapping the potential energy field grid based on the multi-level potential energy data.

[0046] Energy potential-driven path search is performed on the potential energy field grid data to obtain water replenishment path data.

[0047] This invention performs multi-level energy potential processing on water replenishment particle data, quantifying the path dissipation characteristics and water source scheduling accessibility during ecological water replenishment into energy potential indicators, thereby reflecting the differences in energy utilization efficiency and scheduling feasibility of different paths. By constructing multi-level energy potential data, the physical losses, topographic resistance, and management boundary conditions of each water replenishment path can be integrated into a unified potential energy expression system. This energy potential information is then spatially gridded to form potential energy field grid data, giving the water replenishment path selection process clear spatial reference and gradient-driven characteristics. Through energy potential-driven path search, water particles are prioritized to advance along low-dissipation, high-accessibility paths, dynamically avoiding unfavorable terrain and resource conflict zones. This invention not only improves the ecological response adaptability and resource utilization efficiency of path planning but also significantly enhances the intelligent adaptation and global optimization capabilities of the scheduling system under the watershed structure.

[0048] Preferably, the calculation of the water element dissipation cloud map is as follows:

[0049] Based on the water replenishment path data, the geographical path is segmented to obtain path segment data;

[0050] The path segment data is divided into node energy units to obtain node unit data;

[0051] Dissipation calculations are performed on the node cell data to obtain node dissipation data;

[0052] Ecological response coverage maps are constructed based on node dissipation data to obtain water replenishment unit data.

[0053] This invention segments water replenishment path data geographically, dividing continuous water replenishment paths into path segments with independent physical meaning based on factors such as topographic features, hydraulic nodes, and ecological distribution. This helps improve the spatial resolution accuracy of water replenishment paths. Dividing path segments into refined node energy units allows the water element volume carried by each node to be individually modeled and tracked during transport. Dissipation calculations based on node unit data quantify energy attenuation processes such as evaporation loss, seepage dissipation, and hydraulic delay along the path, thus accurately describing the physical attenuation trend of water resources during transmission. Based on this, an ecological response coverage map is constructed, spatially mapping the remaining energy of water elements with downstream ecological response areas to generate water replenishment unit data reflecting the effectiveness of water resource transport and ecological replenishment capacity. This invention not only achieves the linked evaluation of water replenishment paths in physical and ecological spaces but also provides data support and response matching basis for water replenishment path optimization and ecological resource allocation, significantly improving the ecological adaptability and resource allocation efficiency of scheduling schemes.

[0054] Preferably, step S4 specifically includes:

[0055] Based on the water replenishment unit data, a path candidate map is constructed to obtain water replenishment path candidate map data;

[0056] Cost weighting is applied to the candidate water replenishment path map data to obtain the weighted water replenishment path map data.

[0057] Based on the weighted data of the water replenishment path, a routing strategy enhancement simulation is performed to obtain water replenishment simulation scheduling data.

[0058] This invention constructs a candidate graph of water replenishment paths using data from water replenishment units, modeling each path with ecological value and resource accessibility into a graph structure, thus providing a structural foundation for the entire water replenishment scheduling problem through graph computation. A cost weighting mechanism is introduced into this graph structure, assigning each edge a cost weight based on a combination of factors such as energy dissipation, ecological response time, and water transport distance, ensuring that scheduling optimization can accurately measure the systemic cost of path selection. By dynamically training the path selection strategy using reinforcement learning, water replenishment particles can autonomously learn the optimal scheduling path in multi-source, multi-objective scenarios, continuously optimizing ecological benefits and resource utilization efficiency in uncertain environments. The overall process not only achieves global optimization and dynamic adaptation of ecological water replenishment routes but also copes with disturbances caused by sudden events and hydrological changes, improving the intelligence, robustness, and ecological coordination capabilities of the scheduling system.

[0059] Preferably, this application also provides an ecological water replenishment and regulation system for the Yellow River Basin, used to execute the ecological water replenishment and regulation method for the Yellow River Basin as described above. The ecological water replenishment and regulation system for the Yellow River Basin includes:

[0060] The ecological sensing and fingerprint extraction module is used to deploy sensor units at key ecological nodes in the Yellow River Basin to collect Yellow River ecological data; and to extract ecological water demand fingerprints from the Yellow River ecological data to obtain ecological water demand fingerprint data.

[0061] The digital twin river channel modeling module is used to acquire Yellow River hydrological data and construct a digital twin river channel model of the Yellow River based on the Yellow River hydrological data, thus obtaining the Yellow River digital twin river channel model;

[0062] The water replenishment unit generation module is used to construct water replenishment units based on ecological water demand fingerprint data and the Yellow River digital twin channel model, and obtain water replenishment unit data.

[0063] The particle scheduling and path simulation module is used to perform particle routing simulation based on the water replenishment unit data to obtain water replenishment simulation scheduling data.

[0064] The beneficial effects of this invention are as follows: Multiple types of sensors are deployed at key ecological nodes along the Yellow River to achieve real-time acquisition of multi-source data, including hydrological, soil, and environmental data. An ecological water demand fingerprint extraction method is used to extract water demand characteristics of different river sections during specific phenological or ecologically sensitive periods. A digital twin river model constructed using hydrological data achieves high-precision replication of the river's geometric structure, hydraulic properties, and ecological factors, providing a dynamic and computable spatial carrier for water replenishment paths. By identifying ecological response points and hydraulic coupling points, water replenishment particle units with attribute encapsulation are constructed. Energy potential pathfinding and dissipation assessment techniques are employed to construct water element transport paths and efficiency assessment maps. Based on these water replenishment units, a path candidate map is constructed and cost weighting is integrated. An enhanced simulation mechanism is introduced to form an adaptive particle routing scheduling strategy, enabling intelligent matching and dynamic scheduling of multi-source water resources to multi-objective ecological needs. Attached Figure Description

[0065] Other features, objects, and advantages of this application will become more apparent from the following detailed description of the non-limiting embodiments, taken with reference to the accompanying drawings:

[0066] Figure 1 A flowchart illustrating the steps of an embodiment of an ecological water replenishment and regulation method for the Yellow River Basin is shown.

[0067] Figure 2 A flowchart illustrating the steps of a Yellow River ecological data acquisition method according to an embodiment is shown.

[0068] Figure 3 A flowchart illustrating the steps of a digital twin river modeling method according to one embodiment is shown.

[0069] Figure 4 A flowchart illustrating the steps of a water replenishment unit generation method according to an embodiment is shown;

[0070] Figure 5 A flowchart illustrating the steps of a particle scheduling and path simulation method according to one embodiment is shown. Detailed Implementation

[0071] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0072] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. Functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0073] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0074] Please see Figures 1 to 5 This application provides a method for ecological water replenishment and regulation in the Yellow River Basin, comprising the following steps:

[0075] Step S1: Deploy sensor units at key ecological nodes in the Yellow River Basin to collect Yellow River ecological data; extract ecological water demand fingerprints from the Yellow River ecological data to obtain ecological water demand fingerprint data;

[0076] Specifically, for river sections in the Yellow River Basin with significant ecological characteristics and key ecological functions, the following types of sensors are selected and deployed in designated areas: hydrological sensors, such as ultrasonic water level gauges and flow velocity radars, are preferably deployed at typical hydrological cross-sections to collect basic hydrological elements such as river level and flow velocity in real time; soil moisture sensors are prioritized for deployment in riparian wetland areas or areas with typical vegetation root distribution; environmental sensors, including light intensity, wind speed, and air humidity sensing units, are deployed in relatively open, unobstructed bare land to monitor regional microclimate changes; and special monitoring devices, such as underwater acoustic sensors and infrared thermal imaging equipment, are deployed in key biological activity areas (e.g., spawning beaches) to dynamically monitor aquatic biological activity and body temperature changes. All sensor nodes are uniformly configured with a data acquisition cycle, for example, synchronous sampling every 15 minutes, and are connected to edge computing units for local caching and preliminary processing to ensure data stability and transmission efficiency. The collected multi-source ecological data will be spatially located and aggregated according to a pre-defined river section geographic information model to form a river section mapping dataset. The system identifies behavioral triggers and marks abrupt behavioral events (such as sudden rises in water level, rapid drying, and sudden changes in wind speed) in real time to extract key hydrological behavioral data. The system employs a hydrological tension mode decomposition method to deconstruct the behavioral data into three-dimensional features of frequency, amplitude, and gradient, identifying hydrological response patterns at different scales. By constructing a variational modal learning model, it extracts "natural impulses" dominated by natural processes (such as precipitation and runoff) and "artificial impulses" dominated by human-controlled behaviors (such as reservoir scheduling and water diversion operations). The system uses a density-based spatial clustering algorithm (such as DBSCAN) to cluster the impulse response features, extracting typical water demand behavior templates, and encoding them as structured vector fingerprints, outputting ecological water demand fingerprint data.

[0077] Step S2: Obtain Yellow River hydrological data and construct a digital twin river channel model of the Yellow River based on the Yellow River hydrological data to obtain the Yellow River digital twin river channel model;

[0078] Specifically, the system first acquires multi-source hydrological observation data within the Yellow River basin. This data includes, but is not limited to, historical and real-time cross-sectional water level data, flow rate data, and sediment concentration data collected by hydrological monitoring stations along the river. Simultaneously, remote sensing data and UAV aerial images are incorporated to extract and vectorize the river's planar morphology, shoreline changes, and water area boundaries. Based on this data, the system constructs a river channel geometric model: using a digital elevation model (DEM) generated from remote sensing satellite and aerial mapping data, the riverbed geomorphology is reconstructed in three dimensions, forming a detailed geometric description including longitudinal profiles, cross profiles, and local geomorphic abrupt change zones. Furthermore, based on remote sensing time-series data of water surface changes, the system reconstructs the water area boundary changes during flood and dry seasons, capturing the dynamic evolution characteristics of the river channel. Based on the completed geometric reconstruction of the river channel, the aforementioned hydrological time-series data, hydrogeological parameters (such as soil permeability and groundwater depth), and typical aquatic ecological classification information (such as aquatic vegetation types and fish breeding area distribution) are mapped and projected onto the three-dimensional river channel structure model. The system performs regular grid division (such as based on triangular irregular grids or honeycomb hexagonal grids) to divide the river channel space into several ecological functional units. The system injects multiple ecological factors into the river channel mapping model to form ecological coupling characteristics. These ecological factors include key habitat nodes (such as migratory bird stopover zones and dense aquatic species areas), vegetation cover time-series evolution indicators (reflecting the ecological restoration status), and wetland response lag factors (such as soil water retention capacity and water replenishment response delay coefficient). A structured digital twin river channel model data structure is constructed, comprising four modules: geometric morphology, flow field information, ecological factors, and historical process traces.

[0079] Step S3: Construct water replenishment units based on ecological water demand fingerprint data and the Yellow River digital twin channel model to obtain water replenishment unit data;

[0080] Specifically, in the Yellow River digital twin river channel model, response point matching is performed based on the extracted ecological water demand fingerprint data, which includes ecological response time windows and target hydrological parameters (such as flow velocity thresholds and humidity ranges). Through geospatial indexing and temporal matching mechanisms, river segment nodes matching the fingerprint features are located in the model; these are the ecological response points. Each response point represents a specific spatial location and time window with ecological water replenishment needs. To achieve coordinated water source regulation, the system identifies water replenishment coupling points for each response point. Using spatial topology retrieval methods (such as KD-tree indexing or multidimensional nearest neighbor search algorithms), controllable water source inlet nodes connected to the hydraulic paths of the response points are identified in the digital twin river channel structure. These nodes may include reservoir inlets, artificial gate control nodes, or pump station intakes. The aforementioned identification process considers factors such as hydrological path connectivity, node control capability, and historical response records. The system determines path connectivity through terrain modeling and flow network analysis, calculates control capability scores by combining parameters such as the water diversion capacity and response time of the inlet node, and selects the optimal water replenishment coupling point by integrating three indicators based on the success rate and ecological response effect of historical water diversion records. The coupling relationship between each pair of response points and water replenishment points is encapsulated in the form of particle units, constructing water replenishment particle units. Each water replenishment particle unit contains key fields such as: unit water replenishment volume, geographic identifier of the target response node, action time window, ecological water replenishment value score, and path flow loss coefficient. After the particle unit construction is completed, the water replenishment path and energy potential field construction operation is performed. First, based on the hydraulic connectivity and terrain structure between the particle origin and the target response point, combined with factors such as flow slope, vegetation resistance, and hydraulic structure, a path dissipation factor matrix is ​​constructed; then, based on this matrix, energy potential field pathfinding calculation is performed to extract the set of water replenishment paths with the lowest energy consumption or optimal ecological benefits. Based on the existing path set, a water element dissipation cloud map assessment operation is performed on the path nodes. That is, the attenuation degree of unit flow and the ecological response intensity during the transmission process are calculated sequentially at each key node of the path to obtain complete water replenishment unit data.

[0081] Step S4: Perform particle routing simulation based on the water replenishment unit data to obtain water replenishment simulation scheduling data.

[0082] Specifically, based on the constructed water replenishment unit data, a candidate graph structure for particle transport paths is generated. This candidate graph is represented as a directed graph, where each node represents a key control point or ecological response point in the ecological regulation system, such as a gate node, pumping station node, or wetland inlet. Each directed edge represents a feasible flow path for particles along the control path, and the edge weight characterizes the comprehensive energy consumption and scheduling cost of that path. Based on this path graph structure, the system introduces a cost weighting mechanism to assign dynamic weights to each edge. Specifically, the weight of each edge can be calculated using the following weighting formula: W i→j =a·Cdissipation +β·T delay +γ·G eco-gap +δ·C operation C dissipation T represents the energy loss cost of a particle along this path. delay Indicates the expected delay time, G eco-gap C represents the difference between the target ecological response and the current water volume. operation The system represents the technical and human costs required for operation implementation. α, β, γ, and δ are adjustable weighting coefficients used to dynamically adjust the influence weights of each factor to adapt to different scheduling strategies. A particle routing simulation mechanism based on reinforcement learning is adopted to dynamically optimize particle transmission paths. The system can use reinforcement learning algorithms such as deep Q-networks or near-end policy optimization to construct a particle router model. This model takes the current particle's state information and its adjacency matrix in the candidate graph as input and outputs the optimal path selection strategy for the particle's next hop. Through continuous training and policy iteration, the model can learn the optimal particle scheduling path under multi-objective conditions, thereby achieving a synergistic improvement in maximizing ecological response benefits and resource scheduling efficiency. The system generates complete water replenishment simulation scheduling data based on the scheduling paths of all particles. This scheduling data includes the opening and closing timing plans of each key control point, the expected flow control curve, and the sequence of operation execution nodes. It supports outputting in a structured format and connecting to the scheduling execution system or ecological governance government platform to achieve automated ecological water replenishment control and intelligent linkage management.

[0083] Preferably, the Yellow River ecological data collection specifically includes:

[0084] Hydrological sensors were deployed at key ecological nodes in the Yellow River Basin to acquire hydrological data;

[0085] Specifically, the selection principle for the ecological nodes is based on the analysis results of a Geographic Information System (GIS), prioritizing representative typical cross-sections, including but not limited to the following three types of areas: upstream wetland areas, facilitating monitoring of the initial water replenishment effect; confluence areas where tributaries flow into the main river channel, facilitating the capture of changes in tributary water volume and quality; and areas with severe riverbed desertification, facilitating the tracking of dynamic changes in water-sediment relationships. The following types of hydrological sensors are deployed for the aforementioned ecological nodes: water level monitoring equipment, using radar-type water level gauges with a measurement accuracy of ±5 mm and IP68 protection capability, suitable for long-term underwater or humid environments; flow velocity monitoring equipment, equipped with Doppler-based flow meters, capable of monitoring from 0 to 5 meters per second, supporting automatic temperature compensation to adapt to measurement accuracy requirements under different seasons and flow velocity conditions; and water quality parameter sensors, including pH, conductivity, and dissolved oxygen concentration, all using electrode-type online monitoring sensors with plug-and-play and continuous stable monitoring capabilities. The sensor installation method is flexibly selected based on the site water depth and terrain conditions, with two deployment methods: a floating installation method for areas with large water depth fluctuations, and a fixed pile foundation installation method for areas with stable water flow and good foundation conditions. Each sensor is connected to its corresponding local data acquisition unit and performs data acquisition operations at a uniform sampling period. The sampling period is 5 minutes to collect complete hydrological data. The acquired data is transmitted to the edge gateway node deployed in the area via a low-power wide-area network (LoRa) or a 4G communication module. The edge gateway node is responsible for data aggregation and preliminary preprocessing, and assigns a unified data structure encoding format to the uploaded data.

[0086] Soil moisture sensors were deployed at key ecological nodes in the Yellow River Basin to acquire soil moisture data;

[0087] Specifically, the soil moisture sensors are deployed in a layered burial method, with three depth levels: 0–20 cm, 20–40 cm, and 40–60 cm, corresponding to the shallow, middle, and deep layers of typical vegetation root distribution, respectively, comprehensively reflecting changes in soil moisture content in the root zone. The soil moisture sensors used are TDR (Time Domain Reflectometry) sensors, with key performance parameters including a measurement error of less than 2%, a response time of less than 1 second, and equipped with a corrosion-resistant housing and automatic temperature compensation. Within each ecological monitoring node, the deployment density of soil moisture sensors is set based on regional spatial heterogeneity and ecological monitoring needs, selecting 3 to 5 sampling points. The deployment method is a grid-based arrangement with a spacing of 20 meters × 20 meters. For special terrain areas, the grid size is adjusted appropriately to balance representativeness and deployment feasibility. Sensor communication uses the RS-485 bus communication protocol, connecting data from each level of soil moisture sensors to the local data acquisition terminal. The DTU aggregates data from multiple sensors and uploads it to the central management system via wired or wireless means, enabling centralized management and analysis of multi-node, multi-level soil moisture information. Soil moisture data acquired from each sampling is encoded and transmitted according to a unified data structure. The encoded content includes the following information fields: node number, used to identify the specific ecological monitoring location; acquisition time, recording the specific time point of data acquisition; depth level, indicating the vertical depth of the sensor burial (unit: cm); soil moisture content, the volumetric moisture content of the soil layer at the current depth (unit: %); and soil temperature, the temperature value of the soil layer at the current depth (unit: °C). In each data structure, the information of each depth level is arranged sequentially, ensuring that the system can correctly reconstruct the sensor source and measurement parameters based on the field order. For example, if a sampling contains information from three depth levels, the data structure should sequentially present the three depths and their corresponding moisture content and temperature values.

[0088] Deploy environmental sensors at key ecological nodes in the Yellow River Basin to acquire Yellow River environmental data;

[0089] Specifically, the key environmental factors collected by the environmental sensor system include air temperature and humidity, measured by combining humidity-sensitive capacitors and NTC (negative temperature coefficient) thermistors, which features fast response, high accuracy, and suitability for long-term field monitoring; light intensity, using a silicon photovoltaic array light sensor for linear response detection of solar radiation intensity, adapting to all-weather light condition changes; wind speed and direction, using ultrasonic anemometers to measure wind speed and direction on both axes using the time difference principle, which is maintenance-free, has no rotating parts, and is suitable for stable operation in the field; auxiliary monitoring factors, depending on ecological monitoring needs, can be expanded to include a CO2 concentration monitoring module (based on infrared absorption principle) and a dust collection device (gravity sedimentation method or filter membrane collection) for environmental quality trend tracking and pollutant monitoring. All environmental sensor equipment is uniformly installed on poles approximately 1.5 to 2 meters high to ensure that the sensor sampling space is not affected by near-ground disturbances (such as surface radiation reflection, water accumulation disturbance, etc.). The pole system is equipped with lightning protection grounding devices to effectively avoid the risk of lightning strikes. Power is supplied by a combination of solar panels and battery banks, ensuring continuous system operation even in low-light conditions or prolonged rainy weather. Data acquisition is set to once every 10 minutes. Data transmission utilizes the MQTT protocol for low-bandwidth, high-efficiency data transmission, facilitating lightweight communication between edge nodes and the central server, suitable for large-scale ecological monitoring network environments.

[0090] By integrating hydrological data, soil hydrological data, and Yellow River environmental data, we obtain Yellow River ecological data.

[0091] Specifically, all raw monitoring data are aligned and stored using a unified Coordinated Universal Time (UTC) format timestamp and a unique ecological node ID as the primary key. If there are differences in time intervals between data from different sources—for example, hydrological data is sampled every 5 minutes while environmental data is sampled every 10 minutes—linear interpolation is applied to numerical continuous variables (such as water level and soil moisture) to unify them to a 5-minute time resolution. For discrete data that is not suitable for interpolation or changes slowly (such as wind direction and topographic labels), the nearest-value filler method is used. All integrated data is encapsulated in structured data blocks.

[0092] Preferably, the ecological water demand fingerprint extraction specifically involves:

[0093] By performing river segment geographic mapping on the Yellow River ecological data, river segment mapping data is obtained;

[0094] Specifically, the river segment division is based on either a fixed threshold or the assumption that any changes in the river channel angle within a given range constitute the same river channel. Spatial analysis is then conducted using high-resolution remote sensing imagery, digital elevation model (DEM) data, and typical hydrological cross-sectional information from the Yellow River's main stream and tributaries over the years. The entire Yellow River basin is divided into three levels of geographical units—upper, middle, and lower reaches—based on geomorphological structure and hydrological behavior characteristics, serving as the foundation for the first-level river segment division. Within each first-level segment, further refinement is achieved based on key ecological spaces such as tributary confluence areas, typical wetland distribution areas, and ecological protection red line coverage areas, forming second-level river segment units to accurately map ecological functional areas. After completing the river segment division, geographic information system tools (such as ArcGIS or QGIS) are used to spatially register the deployed ecological monitoring nodes (including hydrological, soil, and environmental sensors). A buffer circle with a radius of no more than 300 meters (r≤300m) is constructed centered on each monitoring node. Within this circle, the centerline of the nearest river segment is searched, and the monitoring node is associated with the matched river segment number, thus completing the spatial correspondence between ecological data and river segment structure. The output river segment mapping data is recorded in a structured format. Each record includes a monitoring node identifier, the river segment number, the river segment name, spatial coordinate information (latitude and longitude or projected coordinates), and the river segment hydrological classification code.

[0095] Lightweight behavior-triggered identification is performed based on river section mapping data to obtain hydrological behavior data;

[0096] Specifically, a lightweight behavior recognition rule engine is established in the system to dynamically determine events related to ecological monitoring parameters (such as water level, flow velocity, and sediment concentration) contained in the river section mapping data. This rule engine constructs recognition templates based on preset typical hydrological behavior patterns. For example, if the water level at a monitoring point rises by more than 0.2 meters within three consecutive hours and the flow velocity increases significantly, the system marks this phenomenon as a sudden water replenishment event. Similarly, if the water level in a river section is below a set ecological threshold for five consecutive days, it is identified as an ecological water demand disturbance event. This rule engine employs finite state machine modeling (i.e., assigning corresponding states to each type of hydrological behavior, such as "initial state," "monitoring," "behavior triggered," and "behavior ended." Specific conditions are set for state transitions, such as "water level rising more than 0.2 meters within 3 hours" triggering the "behavior triggered" state. A state transition matrix or table is constructed to clearly define the path from one state to another under different conditions. Hydrological data is progressively processed within a fixed time window, with real-time checks to determine if a transition condition is met. Once the state machine enters the "behavior triggered" state and meets the termination condition, the corresponding behavior label and related river segment and time information are output). State transition encoding is applied to various event templates to ensure clear trigger conditions and state transition paths between different types of behaviors, enabling rapid judgment with low computational overhead. In actual data processing, the system uses a sliding window mechanism to scan and identify ecological monitoring data. A fixed time window (e.g., 24 hours) and time step (e.g., 1 hour) are set. Within each sliding window, the target monitoring variable sequence is extracted and compared segment by segment according to the rule engine, outputting behavior label information. The tag information includes the time of the event, the corresponding river section number, and the type of behavioral event, including "pulse_injection", "ecological_drought" or "normal".

[0097] Water fingerprint data is obtained by extracting water fingerprint data based on hydrological behavior data.

[0098] Specifically, data augmentation preprocessing is performed. For each identified hydrological behavior event, the system constructs a corresponding event cycle window to capture the dynamic evolution characteristics before and after the event. The event cycle window is defined as the previous day as the pre-event window and the following three consecutive days as the post-event window, covering the preparation state before the event triggers, the main reaction period during the event, and the recovery transition period after the event. Next, water fingerprint extraction is performed. The system employs a hydrological tension mode decomposition algorithm (specifically, it extracts an analysis window centered on the target event from the original hydrological time series (e.g., 1 day before the event and 3 days after). Variational mode decomposition (VMD) or empirical mode decomposition (EMD) methods are used to decompose the signal into several intrinsic mode functions (IMFs), each representing a change pattern at different frequency scales. Low-frequency IMFs are synthesized into trend modes, reflecting slow, systematic hydrological changes; high-frequency IMFs are synthesized into disturbance modes, corresponding to short-duration, strong-response hydrological pulses. The energy proportion and center frequency of each mode are calculated to identify the influence intensity of significant modes in the event. The peak amplitude, duration, and spectral distribution of the disturbance modes are analyzed to construct a tension feature vector). Frequency domain analysis is then performed on the core variables (such as water level change curves, flow velocity change curves, and soil moisture content change curves) within each event window. By constructing a decoupled framework for trend and disturbance modes, hydrological signals are decomposed into trend terms (such as continuous slow changes) and high-frequency fluctuation terms (such as short-term pulse disturbances), thereby more accurately locating areas of hydrological abrupt changes. The system uses a variational mode adversarial network to perform intervention causal analysis on hydrological response patterns. This model takes precipitation data, water level curves, and reservoir scheduling data during the event as input. With the help of the mode adversarial training mechanism, it decouples changes driven by natural factors (such as flooding caused by rainfall) from changes caused by artificial regulation (such as reservoir discharge), and constructs fingerprint templates for "natural pulse patterns" and "artificial intervention patterns" respectively. In the feature modeling process, the system extracts the following key index parameters as the core dimensions of water fingerprint, including: (1) pulse intensity, which refers to the maximum change in the value during the event; (2) pulse duration, which is the time span from the appearance of the pulse behavior to its disappearance; (3) recovery time, which is the time required for the system to recover from the peak state to the baseline state; (4) waveform offset angle, which is the phase difference between the rising period of water level or flow velocity and the rainfall period; (5) response lag time, which is the time delay between the external factor trigger and the internal response; (6) dissipation slope, which is the average rate of change of the descending segment after the pulse peak, used to measure the dissipation capacity of the hydrological system.

[0099] Ecological water demand fingerprint data is obtained by performing ecological water demand fingerprint clustering and coding based on water fingerprint data.

[0100] Specifically, the system collects key hydrological information of the Yellow River Basin through multi-source fusion. It utilizes the open and shared hydrological monitoring platform interface of the Yellow River Conservancy Commission to obtain real-time and historical hydrological data for major control sections, including but not limited to indicators such as water level, flow velocity, and sediment concentration. The system uses high-resolution remote sensing data, such as MODIS and Sentinel-2 multispectral imagery, for water body identification, water surface boundary change detection, and remote sensing verification of hydrological anomalies. The system synchronously receives data from hydrological sensor nodes deployed locally in key river sections, including radar current meters, ultrasonic water level gauges, and laser particle sediment concentration sensors. The sensor network employs an edge acquisition and synchronous upload mechanism to ensure data timeliness and continuity. Regarding data acquisition strategies, all data sources are uniformly set to sample at fixed time intervals, with a preferred time resolution of once every 15 minutes. The system performs layered anomaly screening. Specifically, the first layer uses the IQR (interquartile range) method to remove extreme outliers, and the second layer uses the three-standard-deviation (3σ) rule for secondary screening of local fluctuations, thereby improving data quality and stability. During the data structure standardization process, the system uniformly transforms raw data from different sources into a structured data table format. Fields include collection time, collection location number, water level, flow velocity, and sediment concentration. All fields use a unified unit system (e.g., m, m / s, kg / m³). 3 (and include a timestamp and data validity marker).

[0101] Preferably, water is used for fingerprint extraction, specifically as follows:

[0102] Hydrological tension modulus decomposition is performed based on hydrological behavior data to obtain tension regulation characteristic data;

[0103] Specifically, hydrological behavior data acquired in time series form is denoted as H(t). This data typically includes continuous measurements of water level, flow velocity, and water volume as they change over time. The system performs analytical calculations using the hydrological tension operator, denoted as... This operator is used to capture the changing trends of hydrological behavior data across multiple derivative dimensions and amplify or suppress them by combining the watershed's response characteristics to disturbance changes. Its operational logic comprises the following two core components, denoted as... This indicates that the i-th order difference processing is performed on the hydrological behavior data H(t) to extract the dynamic change trend at different orders. For example, This represents the first-order rate of change (i.e., the approximate derivative). Represents changes in acceleration, etc.; the watershed disturbance response function, denoted as R. iLet (t) represent the coupled response factor of the watershed system to external disturbances (such as rainfall, upstream scheduling, soil erosion, etc.) at the i-th time scale. This function can be modeled by fitting the correlation between historical disturbances and the current hydrological state, and has time-varying characteristics. Combining the above two parts, the hydrological tension regulation characteristics at each order are defined as the tension characteristics F at the i-th time scale. i (t), obtained by applying the hydrological tension operator to the original hydrological behavior data, is calculated by multiplying the i-th order difference result with the corresponding disturbance response function point by point. That is, the i-th order tension characteristic is the coupling result between the change trend of this order and the watershed disturbance response. The system will continuously output a set of tension regulation characteristic sequences, denoted as F1(t), F2(t), ..., F n (t), where n is the maximum order required for the analysis, which is set empirically or automatically selected through signal sensitivity assessment.

[0104] A variational modal adversarial network model is constructed based on tension regulation characteristic data;

[0105] Specifically, the constructed network model consists of two main modules: a variational encoder, which performs modal compression and feature extraction on the input tension modulation feature sequence; and a variational inference, which maps the original high-dimensional tension feature sequence into low-dimensional latent variables in the latent variable space. Formally, each set of input features F is encoded as a latent variable z conforming to a certain conditional probability distribution, denoted as q. φ (z|F), where φ is the set of parameters for the encoder. The modal discriminator is responsible for determining the modal type corresponding to the latent variable z, distinguishing whether it originates from natural hydrological regulation (such as seasonal fluctuations, groundwater recharge) or human-induced regulation (such as upstream water release, sluice gate operation, etc.). The discriminator outputs a modal label to guide the distribution learning of the latent variable space. This model uses an adversarial learning mechanism for joint training. Its training objective function consists of two parts: a variational regularization objective (KL divergence), used to constrain the latent variable distribution to approximate the prior distribution (usually a standard normal distribution), achieving regularization and stability enhancement of the latent space structure by minimizing the Kullback-Leibler divergence, thus avoiding overfitting; and a modal adversarial objective (GAN loss), where the encoder and discriminator form an adversarial structure using generative adversarial networks. The encoder attempts to generate "fuzzy modal" latent variables to deceive the discriminator, while the discriminator continuously enhances its ability to discriminate the source of the latent variable modality. Through this adversarial process, the model gradually learns to distinguish between natural and artificial modes in the latent space, thereby enhancing the modal separability of tension features.

[0106] Modal residual learning is performed based on the variational modal adversarial network model to obtain natural impulse data and artificial impulse data;

[0107] Specifically, the system performs dual-channel residual decomposition to extract natural and artificial impulse mode signals under latent spatial conditions. Natural impulse modes primarily reflect pulsating changes driven by natural processes, such as rainfall-induced runoff and seasonal hydrological fluctuations; artificial impulse modes primarily reflect abrupt flow changes caused by human interventions such as reservoir scheduling and dam operations. The module structure includes two independent decoder branches, used to reconstruct the estimated signals of natural and artificial modes from latent variables, including the natural mode decoder branch (denoted as...). ): The input is the latent variable z obtained from the VAE encoder, and the output is the system's estimate of the natural modal components; the artificial modal decoder branch (denoted as...) Similarly, using z as input, the output is the system's estimate of the artificial modal components. The system calculates the residuals between the reconstructed outputs of each branch and the original hydrological behavior data H(t) to infer the modal independence components. The residual calculation method involves removing the artificial modal estimation results from the original data to obtain the natural impulse residual data, i.e., natural impulse data = original hydrological behavior data minus the output of the artificial modal decoder branch; similarly, removing the natural modal estimation results from the original data yields the artificial impulse residual data, i.e., artificial impulse data = original hydrological behavior data minus the output of the natural modal decoder branch. Through this bidirectional residual learning mechanism, the system can automatically separate and interpret hydrological impulses at the modal level without explicit labeling, allowing the hydrological response characteristics driven by nature and those influenced by human intervention to be independently modeled and tracked.

[0108] Causal source modeling was performed based on natural pulse data and artificial pulse data to obtain hydrological behavior causal diagram data;

[0109] Specifically, each hydrological event is abstracted as a node in the graph, denoted as e. i Nodes can correspond to the following types: natural impulse events (such as a sudden rise in water level caused by a rainfall event); artificial impulse events (such as a change in flow velocity caused by the opening of a reservoir gate); and compound events (such as a response to a water level exceeding the warning level caused by the combined effects of previous accumulated rainfall and water level regulation). In a causal graph, edges represent causal relationships between hydrological events and are directed edges (i.e., e...). i →e j Indicates event e i It is event e j (Potential causal antecedents). The weights of the edges represent the causal strength and are defined as the output values ​​of the modality fusion scoring function. This causal scoring function contains two items, including the Granger causal score, which measures the statistical significance of the preceding event e. i For subsequent events e j Does it possess significant predictive power, i.e., does it work without considering e? i time ej The fitting effect decreased significantly; the temporal consistency score, used to measure the synchronicity or logical consistency between two events in their temporal evolution trend, can be obtained by dynamic time warping (DTW) or self-attention weight learning. These two indicators are weighted and fused to form a causal strength score with edge weights. in For event e i to e j The causal strength score, where exp is the natural exponential function, γ1 is the Granger causal score decay coefficient, and G ij For event pairs (e i ,e j The Granger causality score, γ2 is the time consistency score decay coefficient, and D ij For event pairs (e i ,e j The time consistency score is calculated based on the above modeling logic. A directed weighted causal graph is constructed, denoted as G = (V, E, W), where V represents the set of event nodes; E represents the set of causal edges; and W is the causal strength weight matrix corresponding to all edges.

[0110] Modal feature compression was performed on the hydrological behavior causal graph data to obtain water fingerprint data.

[0111] Specifically, the hydrological behavior causal graph is input into a graph neural network (GNN) framework to extract node-level modality embedding features. The selected graph neural network can be a graph convolutional network (GCN) or a sampling graph representation method (such as GraphSAGE), and outputs a modality embedding feature matrix X. * Modal embedding feature matrix X * Each row in the embedding matrix represents the feature representation of an event node in the latent space, including its causal context information with other events. The K principal modal dimensions with the highest information content are selected from the embedding matrix. Information entropy calculation can be performed based on the distribution of each feature dimension across all nodes, selecting the top K features with the highest information entropy values. The selected K principal modal features are integrated to form a fingerprint vector with fixed dimensions.

[0112] Preferably, step S2 specifically includes:

[0113] Obtain hydrological data of the Yellow River;

[0114] Specifically, the system collects key hydrological information of the Yellow River Basin through multi-source fusion. It utilizes the open and shared hydrological monitoring platform interface of the Yellow River Conservancy Commission to obtain real-time and historical hydrological data for major control sections, including but not limited to indicators such as water level, flow velocity, and sediment concentration. The system acquires high-resolution remote sensing data, such as MODIS and Sentinel-2 multispectral imagery, for water body identification, water surface boundary change detection, and remote sensing verification of hydrological anomalies. The system synchronously receives data from hydrological sensor nodes deployed locally in key river sections, including radar current meters, ultrasonic water level gauges, and laser particle sediment concentration sensors. The sensor network employs an edge acquisition and synchronous upload mechanism.

[0115] Based on the Yellow River hydrological data, the geometric structure of the river channel was reconstructed to obtain a spatial model of the river channel.

[0116] Specifically, the system uses multi-source spatial mapping data as input to reconstruct the geometric structure of the Yellow River's main channel and its key tributaries. The input data includes water depth distribution data obtained from cross-sectional measurements, point cloud coordinates from shoreline boundary sampling, a digital elevation model (DEM) of the basin topography, and velocity distribution maps collected at specific cross-sections or river sections. To obtain detailed structures of underwater geomorphological changes, the system integrates cross-sectional deformation data provided by underwater laser scanning (such as green lidar) and ADCP (Acoustic Doppler Current Profiler) measurements. Regarding the geometric reconstruction method, the system preferentially uses the Delaunay triangulation algorithm to construct a three-dimensional mesh structure. The initial mesh is harmonicized using a Laplace smoothing algorithm to suppress anomalous gradients or sharp boundaries. Considering the significant hydrological variations in the Yellow River during the rainy and dry seasons, the system models the river morphology for different time periods separately, constructing seasonal river geometric models to characterize the dynamic range of changes and morphological drift trends. The output is a four-dimensional river spatial mesh model.

[0117] By performing attribute mapping on the river channel spatial model, a river channel mapping model is obtained;

[0118] Specifically, based on the constructed three-dimensional spatial mesh model of the river channel, the system further performs attribute injection and mapping operations to form a river channel mapping model with hydraulic, geological, and ecological responsiveness. The mapped attributes include, but are not limited to, the following categories: hydrodynamic attributes, such as the Manning roughness coefficient (n), which characterizes the velocity and water level changes under the influence of cross-sectional roughness; geological and ecological permeability attributes, including geological permeability (characterizing the ability of water to infiltrate in different substrates) and the slope of the wetting curve (describing the sensitivity of soil moisture content changes to permeability), used for eco-hydrological simulation; hydraulic geometric features, such as the cross-sectional velocity distribution function, used to simulate non-uniform flow structures; and boundary control conditions, including the geographical location, structural state, and control logic parameters of control nodes (such as reservoir dam sites, intakes, and spillway gates). In terms of specific implementation, the system constructs an attribute node table (denoted as A). node This is used to record the aforementioned attribute values ​​and is bound to each unit node of the river channel spatial grid model through a unique node identifier. For regions with missing attribute values, the system uses spatial interpolation algorithms such as Inverse Distance Weighted (IDW) or Kriging to interpolate and complete the values, making the attributes spatially continuous and smooth.

[0119] Ecological factors were injected into the river channel mapping model to obtain a digital twin river channel model of the Yellow River.

[0120] Specifically, the system incorporates key ecological factors into the river channel mapping model. By constructing a multi-dimensional ecological response layer, it achieves a structured expression of the river channel's ecological state, forming a digital twin river channel model. The injected ecological factors mainly include: fish habitat density, expressed as individuals per cubic meter, representing the distribution concentration of representative fish populations in different river sections, used to determine whether the habitat has biological carrying capacity; root water content and distribution characteristics of riparian vegetation, used to reflect the ecosystem's drought resistance and recovery potential; root water content is obtained through soil moisture sensors and remote sensing inversion models; and the bank slope ecological stability coefficient, assessed by combining vegetation cover indicators (such as NDVI value) and slope information. For example, when vegetation is sparse (NDVI < 0.3) and the slope is steep (greater than 35 degrees), it is marked as an ecologically fragile area; when the soil type is silty loam... Areas with an NDVI value below 0.4 are marked as ecologically vulnerable; areas with a historical shoreline retreat rate > 2 meters / year and no man-made slope protection structure are marked as ecologically vulnerable; areas located in flood-prone zones with a slope greater than 30 degrees are marked as ecologically vulnerable; areas with a slope between 20° and 35° but a continuously decreasing NDVI value for three periods are marked as potentially ecologically vulnerable. The wetland cover ratio, representing the proportion of wetland areas in the total river section area, reflects the scale of the ecological buffer zone. Human intrusion impact factors, such as the frequency of gate opening and flow velocity fluctuations, are used to measure the degree of impact of human intervention on ecological stability. Each of the above factors is constructed as an independent ecological factor mapping layer, denoted as... Using three-dimensional spatial coordinates as a reference, the system models the spatial distribution of ecological impact parameters. Through a rule-based conditional injection mechanism, ecological factors are incorporated into the river channel node structure. For example, in spatial areas satisfying "NDVI value less than 0.3 and slope greater than 35 degrees," the system sets the ecological stability factor of that node to a "low" level. After completing the injection of all ecological factors, the system constructs a multi-layered ecological mapping system and maps it to the river channel spatial geometric model (M). geo ), Attribute Node Table (A) node Together they form a digital twin river structure.

[0121] Preferably, step S3 specifically includes:

[0122] Ecological response point data is obtained by matching ecological water demand fingerprint data with ecological response points in the digital twin river channel model of the Yellow River.

[0123] Specifically, a spatial search algorithm (such as a KD-tree-based data structure) is used to quickly index the river channel units in the digital twin model. Taking the water demand area types and their spatial distribution ranges described in the ecological fingerprint as a reference, a set of location candidates with potential response capabilities in the model is located. Alternatively, ecological semantic constraint matching is further performed on the spatial candidate areas. Taking the "wetland type" water demand as an example, the system will preferentially select river section units that meet the following conditions as matching targets. For example, the Normalized Difference Vegetation Index (NDVI) value is higher than 0.6, indicating that the area has good green coverage and a basis for hygrophytic vegetation; the slope is lower than 10 degrees, indicating that the area has the topographic conditions to form or maintain wetland water bodies; if 0.3 < NDVI < 0.6 and the slope ∈ [15°, 35°] and the shoreline recession rate > 1.5 m / year, it is determined as a priority shore zone restoration point; if the river section is the main river and the flow velocity ∈ [0.3, 1.0] m / s and there are no hydraulic obstacles, it is determined as a suitable section for the water demand of the migration channel; if NDVI < 0.3 and the wind erosion sensitivity level is high and the average annual evaporation is > rainfall, the area is marked as a priority sand pressing unit for water demand; if the area is located in the core area of a natural wetland and the standard deviation of historical water level fluctuations < threshold (such as 0.3 m) and there is little human intervention around, it is a priority waterbird water replenishment point.

[0124] Identify the water replenishment coupling points based on the ecological response point data to obtain the water replenishment coupling point data;

[0125] Specifically, for each ecological response point P resp (j), in the geospatial space, with P resp (j) as the center, search for the set of river channel nodes within a range of 2 to 5 kilometers upstream to form a preliminary candidate set. During screening, it is necessary to ensure that the candidate points are in the upstream position in the hydrodynamic simulation model and the flow direction between the candidate points and the response points is consistent. Use a graph embedding learning algorithm (such as Node2Vec) to embed the candidate nodes into the latent space and construct a vector expression model of the river channel structure diagram. For each response point, calculate the coupling affinity score between it and the candidate nodes, which is obtained through vector similarity calculation. The higher the affinity, the stronger the path accessibility and the better the structural continuity. Select the node with the highest coupling affinity score from the candidate nodes as the best water replenishment coupling point for the current response point, denoted as P couple (k). If the scores of multiple candidate points are similar, factors such as the cost of water transfer, the length of the water conveyance path, or the intervention priority can be combined for sorting or marking redundant alternative points.

[0126] Construct water replenishment particle units based on the water replenishment coupling point data to obtain candidate data for water replenishment particles;

[0127] Specifically, a water replenishment particle unit is an abstract structure representing the smallest functional unit of a water diversion path, embodying the process of water being transported from a coupling point to a designated ecological target area. Each particle unit carries necessary hydrophysical constraints and ecological regulation attributes to characterize its water replenishment feasibility and ecological priority. Example units include {starting point coordinates, ecological target, minimum water demand, maximum flow time, ecological level}. Starting from the coupling point, and combining topography, hydrodynamic network, and river accessibility maps, connected ecological target points are identified, and the physical accessibility of the path is evaluated. Based on the satisfactory physical accessibility assessment, candidate particle units are constructed. Each particle simulates the amount of ecological water it can transport and the target area it can reach within a fixed water replenishment duration (e.g., 1 hour), generating a preliminary candidate set, denoted as the water replenishment particle candidate set; each particle unit in the set satisfies basic constraints (e.g., maximum flow time, minimum water demand), and its structural attribute fields are recorded.

[0128] The candidate data for water replenishment particles are encapsulated with attributes to obtain the water replenishment particle data;

[0129] Specifically, the system introduces the following dynamic attribute fields for candidate particle units to express the controllability, behavior type, resource consumption, and environmental constraints of the path; Adjustability indicates whether the particle path is controlled by regulatory facilities such as reservoirs and sluices; if the path origin or path contains artificial regulation nodes, it is marked as "Yes"; otherwise, it is marked as "No," indicating that the path is a natural terrain water diversion path and does not have external controllability. Behavior type refers to the main functional behavior of the particle, which can be divided into the following categories, such as "water transfer": indicating cross-regional water allocation; "water injection": indicating the injection of ecological water replenishment into a specific area; "water diversion": indicating the extraction of water from natural water bodies for ecological regulation. Energy consumption budget is estimated based on the hydraulic friction and head loss of the particle path, reflecting the unit energy consumption or head resource cost required for water replenishment operations; it is expressed as the energy required per unit volume of water (e.g., kilowatt-hours / cubic meter) or head loss (e.g., meters). The environmental risk level assesses the potential disturbance risk to the ecosystem along the particle path; it is obtained by mapping from a pre-defined parameter library, referencing anthropogenic activity density layers (such as industrial and mining distribution, farmland development, and urban construction); and is divided into "Low," "Medium," and "High" levels; these are used for ecological water replenishment priority ranking and path elimination. Each encapsulated water replenishment particle unit is recorded in a structured form, forming a standardized data structure.

[0130] Based on the water replenishment particle data, the potential field is used to find the water replenishment path data.

[0131] Specifically, based on the transport requirements of water replenishment particles in the river channel, the system constructs an energy potential field model and performs path optimization accordingly to obtain the optimal flow path for ecological water replenishment. The system defines two types of basic energy potential indices, such as path dissipation potential φ. diss (x,y): Used to measure the energy loss of water flow under the influence of factors such as topographic slope and riverbed resistance; achievable energy potential φ for scheduling. ctrl (x,y): The dispatchability of water flow is assessed based on the allocation capacity of the water source control unit, path accessibility, and historical reachability probability. The above two types of potential values ​​are weighted and combined to form the potential function: Φ total (x,y)=α· diss (x,y)+β·φ ctrl (x,y), where Φ total (x,y) represents the total potential value, α is the path dissipation potential energy weighting coefficient, and φ diss (x,y) represents the path dissipation potential, β represents the scheduling control potential weight coefficient, and φ ctrl (x, y) represents the achievable energy potential for scheduling, and parameters α and β are the strategy setting weights, which can be adjusted according to the real-time target of ecological water replenishment. The system will then allocate the energy potential Φ... total (x,y) is mapped onto a two-dimensional grid structure in the river region to construct a potential energy field diagram.

[0132] Water element dissipation cloud map calculation is performed on the water replenishment path data to obtain water replenishment unit data.

[0133] Specifically, the path set Path opt (n) Perform equal-length segmentation (e.g., every 500 meters as a segment) to generate a path segment sequence Segment. s (l). Within each path segment, hydraulic parameters are extracted to form node elements, including node spatial coordinates, local velocity values, drag factor R, and unit segment length L. The formula for calculating unit energy consumption is E. l =γ·Q·R·L, where E l Let γ be the dissipation coefficient, Q represent the unit flow rate per particle, R be the channel resistance, and L be the segment length. The unit energy consumption E at each node is... l A hydrological heat dissipation map (CloudMap) is generated by back-mapping along the path to the ecological response area and using interpolation. resp (i) Heatmaps are used to identify areas with significant dissipation gradients and key points for ecological water replenishment. Based on dissipation cloud maps and ecological target response level assessment criteria, the system outputs local water replenishment unit data.

[0134] Preferably, the pathfinding in the potential field is specifically as follows:

[0135] Multi-level energy potential processing is performed on the water replenishment particle data to obtain multi-level energy potential data, which includes path dissipation energy potential processing and water source scheduling accessibility processing.

[0136] Specifically, the system performs dissipative potential energy analysis on the path nodes that each water replenishment particle may traverse. The dissipation factor is defined as the degree of potential energy lost by the water flow due to topographic resistance and riverbed roughness. Its main influencing factors include riverbed slope, representing the degree of water level change per unit distance; friction coefficient, reflecting the degree of kinetic energy loss between the fluid and the riverbed; and riverbed roughness coefficient, representing the resistance to water flow propagation caused by the unevenness of the topographic surface. Based on these parameters, the system calculates the corresponding dissipative potential energy value φ for each spatial node position (x, y) in the path. diss (x,y)=f r ·η·tan(θ), where φ diss (x,y) represents the path dissipation potential energy, f r η is the friction coefficient, η is the riverbed roughness coefficient, tan(θ) is the tangent of the river slope, and the calculation results are uniformly normalized so that the dissipated potential energy values ​​of all nodes are normalized to the [0,1] interval.

[0137] The system constructs a water source reachability graph representing the path relationships between water source control nodes. Nodes in the graph represent various scheduling and control units (such as reservoirs, sluice gates, and water intakes), and edges represent the water flow feasibility of scheduling paths. Where φ ctrl (x,y) represents the scheduling control potential energy value, W control The system assigns weighted scores to control nodes, with the number of blockages representing the number of path blockages. For any target water replenishment area, the system calculates its coverage capacity by different control nodes, forming a control coverage index. Key factors considered include the connectivity status of the path (e.g., whether it is blocked by man-made facilities or natural obstacles); the weighted scores of the control nodes (e.g., water transfer capacity, response time); and empirical data such as historical water transfer success probabilities. Combining this information, the system calculates the achievable energy potential value for each location (x, y) at the spatial grid level, obtaining the corresponding control potential energy index.

[0138] The system weights and superimposes the two types of energy potential indices to generate a total energy potential value. For each grid point, its total energy potential is obtained by a linear weighted combination of path dissipation potential energy and scheduling control potential energy, i.e., Φ. total (x,y)=α·φ diss (x,y)+β·φ ctrl (x,y),Φ total (x,y) represents the total potential value, α is the path dissipation potential energy weighting coefficient, and φ diss (x,y) represents the path dissipation potential energy, β represents the scheduling control potential energy weighting coefficient, and φ ctrl (x,y) represents the scheduling control potential energy value.

[0139] Potential energy field grid data is obtained by mapping the potential energy field grid based on the multi-level potential energy data.

[0140] Specifically, the Yellow River channel and its related water replenishment areas are divided into regular two-dimensional spatial grids, with each grid cell defined as having a unique geographic coordinate index (x, y). Preferably, the grid division precision is set to 50 meters × 50 meters. Each grid cell serves as the smallest unit bearing the potential energy field, storing the total potential value at its corresponding location. To prevent the path planning process from exceeding boundaries or deviating from the ecological water replenishment permit area, the system performs special marking processing on grid cells that do not belong to the effective river channel range. For areas falling outside the river channel boundary or explicitly restricted by topography or ecological regulations, their corresponding potential energy values ​​are set to maximum values, indicating an "impassable" state.

[0141] Energy potential-driven path search is performed on the potential energy field grid data to obtain water replenishment path data.

[0142] Specifically, the potential energy field grid data is constructed into a weighted grid graph G. grid = (N, E), where node N represents each grid cell, and edge set E represents the walkable paths between adjacent grid cells. The weight of each edge is defined as the total potential energy Φ corresponding to the nodes at its two ends. total The average value is used to represent the potential energy consumed to move from one point to another. The system uses an improved A* (A-star) path planning algorithm for path calculation. The heuristic function h(n) is defined as the Euclidean distance between the current node n and the target node; the cumulative path cost function f(n) is defined as f(n) = g(n) + h(n), where g(n) is the total potential energy accumulated from the starting point to the current node n, in the form g(n) = ∑Φ total (x i ,y i The maximum path length is set (e.g., 15 km); the path is prohibited from crossing areas with zero regulation capacity (such as enclosed water bodies or inaccessible cross-sections); for nodes in the path that cross aquatic ecologically sensitive areas (such as fish spawning grounds or wetland core zones), a penalty factor (e.g., ΔΦ = +0.2) is added to their potential energy value to inhibit the path from preferentially passing through these areas. The output water replenishment path is represented in the form of a coordinate sequence, denoted as the path point set.

[0143] Preferably, the calculation of the water element dissipation cloud map is as follows:

[0144] Based on the water replenishment path data, the geographical path is segmented to obtain path segment data;

[0145] Specifically, the overall water replenishment path is broken down into segmented units with geographical consistency and controllable response. Physical topography is used to segment the river based on changes in river topography (such as bends, bifurcations, and abrupt changes in slope). Each segment is defined as a continuous river section with similar hydrodynamic conditions, with a length of 500m-1km. If the turning angle of the path is >45°, it is used as the starting point of a new segment.

[0146] The path segment data is divided into node energy units to obtain node unit data;

[0147] Specifically, the energy propagation structure within each segment is refined, and a high-resolution nodal energy map is established. Each path segment is interpolated at 50m intervals to generate a node set N = {n1, n2, ..., n}. k}, where N is the node set, n1 is the first node, n2 is the second node, and n k For the k-th node; assign an initial hydropower index, such as unit flow velocity v, to each node. i water depth h i Cross-sectional area A i Calculate the kinetic energy density of this node. Where ρ is the density of the water.

[0148] Dissipation calculations are performed on the node cell data to obtain node dissipation data;

[0149] Specifically, the energy attenuation process of the water body at each node is simulated due to factors such as topography and resistance. The node energy consumption function is defined as the energy dissipation process for each adjacent node n. i ,n i+1 Calculate the energy consumption difference ΔE i =E i -E i+1 , where ΔE i E represents the energy difference between nodes. i E represents the energy value of the current node. i+1 Let ΔE be the energy value of the next node. i A value less than 0 indicates a node with external energy supply, and it is recorded as an anti-dissipation point; the cumulative dissipation index is defined as the dissipation degree of the path segment. Where D i Here, j represents the path segment dissipation, j is the node index number, and k is the path segment dissipation. i Let E be the number of nodes within the segment. j The energy consumption difference of the j-th pair of nodes is represented by the sum of the dissipation values ​​of all segments to form a spatial heat dissipation intensity map.

[0150] Ecological response coverage maps are constructed based on node dissipation data to obtain water replenishment unit data.

[0151] Specifically, energy consumption data is spatially matched with ecological water demand points to output interpretable water replenishment units. The system spatially projects all ecological response points onto the existing path segmentation and node grid system to determine whether each ecological point falls within the influence range of a specific river segment, thereby establishing a response area-river segment mapping relationship. Local dissipation values ​​within a path segment or node grid are denoted as D. local The system calculates the average dissipation level of the overall path. And set the dissipation fluctuation threshold δ. If it meets the following conditions... This area is then identified as a high-dissipation area, meaning that the energy consumed for water replenishment is significantly depleted in this section, leading to a decrease in ecological response efficiency or insufficient water transfer. If the ecological response point r i Within the projection range of a node falling on a high-dissipation segment, the system records the response coupling relationship between that point and that segment to assess the energy consumption-effect matching degree of the ecological water transfer target. The above analysis results are used to generate an ecological response coverage map, which graphically displays the energy consumption-response coupling relationship between the water replenishment path segment and the ecological target.

[0152] Preferably, step S4 specifically includes:

[0153] Based on the water replenishment unit data, a path candidate map is constructed to obtain water replenishment path candidate map data;

[0154] Specifically, based on the extracted ecological water replenishment unit data, the system constructs a candidate path map covering multiple water source outlets and multiple target ecological response points. The node set in the graph structure includes three categories: source nodes, corresponding to the water source points set in each water replenishment unit, typically reservoirs, sluice gates, or irrigation canal inlets; sink nodes, corresponding to the target areas of the ecological response region, such as water-demanding river sections or wetland areas; and relay nodes, representing spatial units traversed in the water replenishment path, such as river sections, watersheds, culverts, or geomorphic control points. The division of relay nodes is based on spatial segmentation rules and can be automatically extracted using DEM, river connectivity maps, or geographic cross-sectional structures. The edge set in the graph is used to indicate traversable hydrological paths between two nodes. Each edge contains the following attributes: edge length, representing the spatial distance of the path segment in meters; type, indicating the category of the channel, such as "river," "irrigation canal," or "culvert"; maximum flow rate, representing the maximum water replenishment flow that the path segment can support under current hydraulic conditions, in cubic meters per second; velocity limit, used to control the dynamic movement speed in particle simulation; and scheduling capability level, used to indicate whether manual control capability is available, such as setting valves or gate control mechanisms. The graph structure supports arbitrary combinations of multiple source points and multiple target points, including complex water replenishment path topologies such as tributaries, bypass paths, or redundant loops. The system uses an adjacency list structure to represent the candidate path graph, where each edge records its spatial and hydraulic attributes in a labeled format. The candidate path graph is output as structured graph data.

[0155] Cost weighting is applied to the candidate water replenishment path map data to obtain the weighted water replenishment path map data.

[0156] Specifically, a cost weight for water scheduling is assigned to each edge in the path candidate graph. The cost dimension consists of distance cost (C). d ), path physical length; dissipation cost (C e Energy loss per unit water flow in this section; Environmental impact factor (C) eco ), whether it involves sensitive areas (such as wetlands, protected areas); management risk factors (C r The difficulty and accessibility of human intervention, such as historical water diversion success rates, are considered. The weighted cost model is as follows: Among them W ij for, C is the distance cost weighting coefficient. d The physical length of the path. C is the weighting factor for dissipation cost. e The energy loss cost per unit flow rate can be calculated based on factors such as gradient, roughness coefficient, and friction. C represents the environmental impact weighting coefficient. eco Environmental sensitivity indicators, such as ecological protection level, NDVI vegetation cover index, and wetland overlap ratio, To manage risk weighting coefficients, C r To regulate accessibility indicators. Based on W ij Label each edge; transform the graph structure into a weighted graph.

[0157] Based on the weighted data of the water replenishment path, a routing strategy enhancement simulation is performed to obtain water replenishment simulation scheduling data.

[0158] Specifically, the optimal scheduling path is simulated based on reinforcement learning or heuristic optimization strategies, and an executable scheduling sequence is output. A state vector S is constructed. t The status includes the current node number (location code) v t ; Remaining target distance (e.g., Euclidean distance to the target point) dist(v t ,v target Remaining available water Q rem The weighted scheduling cost W of the current node ij Ecological response intensity estimate (e.g., water demand matching score) R eco Define the action set A for each state. t This corresponds to the selection of reachable adjacent nodes, i.e., starting from the current node v. t The set A of all downstream nodes reachable from the starting point t ={v j ∈adj(v t )∣Wtj <∞∧φ ctrl (v j )>0}, where v j Let adj(v) be the candidate adjacent node number. t W represents the set of adjacent nodes of the current node. tj φ represents the scheduling cost of the current path edge. ctrl (v j ) to regulate the potential energy of the node, each action a t Choose an adjacent node and its scheduling traffic (as the dimension of continuous actions). Model using a dual network structure, including a policy network (Actor), with the current state S as input. t Output action a t That is, the next hop node and its allocated water volume; the value network (Critic), input state-action pairs (S... t ,a t Output the expected cumulative return Q(S) t ,a t The network structure adopts a fully connected layer design with ReLU activation function. The output strategy uses Tanh to limit the flow range, ensuring rational water allocation. An immediate reward function R is set. t : R t For instant reward value, W represents the path cost weighting coefficient. ij The path edge weights (scheduling costs) R represents the weighting coefficient for the ecological response score. eco To score the ecological response, Q is the flow deviation penalty coefficient, and diversity is the deviation function between the current allocated flow and the target water demand. t Q represents the traffic value scheduled by the current node. target The reference flow value is set for the target node. Initialize the policy network and value network parameters; randomly initialize the starting point and target in a weighted graph environment; generate path node sequences and water volume actions according to the policy network; calculate the actual reward and update the Critic network; optimize the Actor network using policy gradient backpropagation; update the target network parameters and perform a soft-update; repeat training until the average cumulative reward converges. Output an optimal ecological scheduling path sequence, Route = v1→v2→v3→…→v target Where Route is the path sequence, v1 is the initial node, v2 is the first intermediate node, v3 is the second intermediate node, and v target For example, node v2 is set to a traffic of 20 (unit: m). 3 / s), with a control time window from 08:00 to 10:00; node v5 is set to a flow rate of 18 (unit: m). 3 / s), with a control time window of 10:00 to 12:00.

[0159] Preferably, this application also provides an ecological water replenishment and regulation system for the Yellow River Basin, used to execute the ecological water replenishment and regulation method for the Yellow River Basin as described above. The ecological water replenishment and regulation system for the Yellow River Basin includes:

[0160] The ecological sensing and fingerprint extraction module is used to deploy sensor units at key ecological nodes in the Yellow River Basin to collect Yellow River ecological data; and to extract ecological water demand fingerprints from the Yellow River ecological data to obtain ecological water demand fingerprint data.

[0161] The digital twin river channel modeling module is used to acquire Yellow River hydrological data and construct a digital twin river channel model of the Yellow River based on the Yellow River hydrological data, thus obtaining the Yellow River digital twin river channel model;

[0162] The water replenishment unit generation module is used to construct water replenishment units based on ecological water demand fingerprint data and the Yellow River digital twin channel model, and obtain water replenishment unit data.

[0163] The particle scheduling and path simulation module is used to perform particle routing simulation based on the water replenishment unit data to obtain water replenishment simulation scheduling data.

[0164] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended application documents rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of the equivalents of the application documents be incorporated into the invention.

[0165] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for ecological water replenishment and regulation in the Yellow River Basin, characterized in that, Includes the following steps: Step S1: Deploy sensor units at key ecological nodes in the Yellow River Basin to collect Yellow River ecological data; extract ecological water demand fingerprints from the Yellow River ecological data to obtain ecological water demand fingerprint data; Step S2: Obtain Yellow River hydrological data and construct a digital twin river channel model of the Yellow River based on the Yellow River hydrological data to obtain the Yellow River digital twin river channel model; Step S3: Construct water replenishment units based on ecological water demand fingerprint data and the Yellow River digital twin channel model to obtain water replenishment unit data; Step S4: Perform particle routing simulation based on the water replenishment unit data to obtain water replenishment simulation scheduling data.

2. The method according to claim 1, characterized in that, The specific data collection for the Yellow River's ecology includes: Hydrological sensors were deployed at key ecological nodes in the Yellow River Basin to acquire hydrological data; Soil moisture sensors were deployed at key ecological nodes in the Yellow River Basin to acquire soil moisture data; Deploy environmental sensors at key ecological nodes in the Yellow River Basin to acquire Yellow River environmental data; By integrating hydrological data, soil hydrological data, and Yellow River environmental data, we obtain Yellow River ecological data.

3. The method according to claim 1, characterized in that, Ecological water demand fingerprint extraction specifically involves: By performing river segment geographic mapping on the Yellow River ecological data, river segment mapping data is obtained; Lightweight behavior-triggered identification is performed based on river section mapping data to obtain hydrological behavior data; Water fingerprint data is obtained by extracting water fingerprint data based on hydrological behavior data. Ecological water demand fingerprint data is obtained by performing ecological water demand fingerprint clustering and coding based on water fingerprint data.

4. The method according to claim 3, characterized in that, Water fingerprint extraction specifically involves: Hydrological tension modulus decomposition is performed based on hydrological behavior data to obtain tension regulation characteristic data; A variational modal adversarial network model is constructed based on tension regulation characteristic data; Modal residual learning is performed based on the variational modal adversarial network model to obtain natural impulse data and artificial impulse data; Causal source modeling was performed based on natural pulse data and artificial pulse data to obtain hydrological behavior causal diagram data; Modal feature compression is performed on the hydrological behavior causal graph data to obtain water fingerprint data.

5. The method according to claim 1, characterized in that, Step S2 is as follows: Obtain hydrological data of the Yellow River; Based on the Yellow River hydrological data, the geometric structure of the river channel was reconstructed to obtain a spatial model of the river channel. By performing attribute mapping on the river channel spatial model, a river channel mapping model is obtained; Ecological factors were injected into the river channel mapping model to obtain a digital twin river channel model of the Yellow River.

6. The method according to claim 1, characterized in that, Step S3 is as follows: Ecological response point data is obtained by matching ecological water demand fingerprint data with ecological response points in the digital twin river channel model of the Yellow River. Water replenishment coupling points are identified based on ecological response point data to obtain water replenishment coupling point data; Based on the water replenishment coupling point data, water replenishment particle units are constructed to obtain water replenishment particle candidate data; The candidate data for water replenishment particles are encapsulated with attributes to obtain the water replenishment particle data; Based on the water replenishment particle data, the potential field is used to find the water replenishment path data. Water element dissipation cloud map calculation is performed on the water replenishment path data to obtain water replenishment unit data.

7. The method according to claim 6, characterized in that, The specific process of finding a path in the potential field is as follows: Multi-level energy potential processing is performed on the water replenishment particle data to obtain multi-level energy potential data, which includes path dissipation energy potential processing and water source scheduling accessibility processing. Potential energy field grid data is obtained by mapping the potential energy field grid based on the multi-level potential energy data. Energy potential-driven path search is performed on the potential energy field grid data to obtain water replenishment path data.

8. The method according to claim 6, characterized in that, The specific calculation of the water element dissipation cloud map is as follows: Based on the water replenishment path data, the geographical path is segmented to obtain path segment data; The path segment data is divided into node energy units to obtain node unit data; Dissipation calculations are performed on the node cell data to obtain node dissipation data; Ecological response coverage maps are constructed based on node dissipation data to obtain water replenishment unit data.

9. The method according to claim 1, characterized in that, Step S4 is as follows: Based on the water replenishment unit data, a path candidate map is constructed to obtain water replenishment path candidate map data; Cost weighting is applied to the candidate water replenishment path map data to obtain the weighted water replenishment path map data. Based on the weighted data of the water replenishment path, a routing strategy enhancement simulation is performed to obtain water replenishment simulation scheduling data.

10. An ecological water replenishment and regulation system for the Yellow River Basin, characterized in that, For implementing the Yellow River Basin ecological water replenishment and regulation method as described in claim 1, the Yellow River Basin ecological water replenishment and regulation system comprises: The ecological sensing and fingerprint extraction module is used to deploy sensor units at key ecological nodes in the Yellow River Basin to collect Yellow River ecological data; and to extract ecological water demand fingerprints from the Yellow River ecological data to obtain ecological water demand fingerprint data. The digital twin river channel modeling module is used to acquire Yellow River hydrological data and construct a digital twin river channel model of the Yellow River based on the Yellow River hydrological data, thus obtaining the Yellow River digital twin river channel model; The water replenishment unit generation module is used to construct water replenishment units based on ecological water demand fingerprint data and the Yellow River digital twin channel model, and obtain water replenishment unit data. The particle scheduling and path simulation module is used to perform particle routing simulation based on the water replenishment unit data to obtain water replenishment simulation scheduling data.

Images