Digital twin water conservancy hyper-fusion method for coupling multi-source data and all-in-one machine

Abstract

The invention provides a digital twin water conservancy hyper-fusion method for coupling multi-source data and an all-in-one machine, and relates to the technical field of intelligent water conservancy. By constructing a unified water conservancy spatio-temporal data link protocol, efficient fusion of multi-source heterogeneous data of hydrology, meteorology, engineering operation and the like is realized, and the problem of water conservancy system data islands is solved; a coupling distributed hydrological model and a hydrodynamic model are established, and real-time accurate simulation of the watershed hydrological process and the river channel hydrodynamic process is achieved; a multi-target dynamic weight reinforcement learning algorithm is adopted to optimize a water conservancy scheduling scheme, model parameters are dynamically corrected in real time through a time sequence error regression analysis model, and closed-loop adaptive optimization is achieved; according to the method, closed-loop cooperation of water conservancy data fusion, real-time accurate simulation, intelligent scheduling optimization and dynamic parameter self-correction is realized, and the real-time performance and accuracy of water conservancy intelligent decision making are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of smart water conservancy technology, specifically to a digital twin water conservancy hyper-fusion method and integrated machine that couples multi-source data. Background Technology

[0002] The safe and efficient operation of water conservancy projects is of vital importance for flood control, disaster reduction, and optimal allocation of water resources. Firstly, water conservancy data is diverse, including hydrological, meteorological, and engineering operation data, which are scattered across different departments and systems. Data standards are inconsistent, and there is a lack of effective data fusion methods, making unified management and efficient utilization difficult. This creates a serious "data silo" problem, severely impacting the overall effectiveness of water conservancy decision-making and coordinated scheduling.

[0003] Secondly, traditional hydrological and hydrodynamic simulation models generally rely on a large amount of historical data for parameter calibration, resulting in static model parameters and insufficient real-time performance. Especially in emergency scenarios such as sudden floods, they are unable to quickly and dynamically respond to the real-time needs of water conservancy scheduling decisions and cannot meet the accurate real-time prediction capabilities required by the current complex and ever-changing water conservancy environment.

[0004] Furthermore, the existing hardware resources (such as monitoring sensors and computing servers) and software systems (including data monitoring, simulation, and optimization scheduling) of water conservancy systems are often independent of each other and lack effective coordination mechanisms, making it difficult to form a unified and efficient overall scheduling system, which seriously reduces the response speed and decision-making efficiency of water conservancy systems.

[0005] In summary, how to break the isolation of water conservancy data and build a unified data fusion system; how to improve the real-time performance and prediction accuracy of water conservancy simulation models to meet dynamic response requirements; and how to achieve efficient collaboration between hardware resources and software systems have become the core technical issues that urgently need to be addressed in the field of smart water conservancy. Summary of the Invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art; the present invention proposes a digital twin water conservancy hyperfusion method and integrated machine that couples multi-source data.

[0007] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a digital twin hydraulic hyperfusion method for coupling multi-source data, comprising: A water conservancy spatiotemporal data chain protocol is established, and water conservancy monitoring data is formatted and processed through hierarchical data modeling. Edge computing nodes are determined based on the formatted water conservancy monitoring data and water affairs data. The predicted future load values ​​of the nodes are obtained based on the edge computing nodes, and a cloud-based fusion dataset is generated by selecting an appropriate data transmission format. Based on the cloud-based fusion dataset, a coupled distributed hydrological model and a hydrodynamic model are constructed. An initial scheduling scheme for the forecast period is generated based on rainfall and water information, and a forward simulation is performed to obtain the initial simulation results. Based on the difference data between the initial simulation results and the preset scheduling target, and using the difference data as feedback, a multi-objective dynamic weight reinforcement learning algorithm is used to reverse optimize and adjust the control parameters of the initial scheduling scheme in the foreseeability period to generate an optimized scheduling scheme in the foreseeability period. Based on the optimized scheduling scheme with a forward prediction period, a forward simulation was conducted again to verify the optimized simulation results. Then, based on the real-time deviation between the optimized simulation results and the preset scheduling target, the model parameters of the coupled distributed hydrological model and hydrodynamic model were dynamically corrected using a time-series error regression analysis model to achieve dynamic self-correction of the model parameters.

[0008] Secondly, the present invention provides a digital twin hydraulic hyper-converged integrated machine that couples multi-source data. The integrated machine is equipped with multiple modules for implementing the aforementioned digital twin hydraulic hyper-converged method that couples multi-source data. The modules include: Data fusion module: used to establish a water conservancy spatiotemporal data chain protocol, realize the formatted processing of water conservancy monitoring data through hierarchical data modeling; determine edge computing nodes based on formatted water conservancy monitoring data and water affairs data, obtain the predicted future load value of the nodes based on the edge computing nodes, and select an appropriate data transmission format to generate a cloud-based fused dataset; Coupled hydrological simulation module: used to construct a coupled distributed hydrological model and a hydrodynamic model based on the cloud-based fusion dataset, generate an initial scheduling scheme for the forecast period based on rainfall and water information, and perform forward simulation to obtain initial simulation results; The scheduling optimization module is used to optimize and adjust the control parameters of the initial scheduling scheme in the foresight period based on the difference data between the initial simulation results and the preset scheduling target, using the difference data as feedback, and employing a multi-objective dynamic weight reinforcement learning algorithm to generate an optimized scheduling scheme in the foresight period. The parameter self-calibration module is used to perform forward simulation pre-run verification based on the forecast period optimized scheduling scheme, obtain optimized simulation results, and dynamically correct the model parameters of the coupled distributed hydrological model and hydrodynamic model using a time series error regression analysis model according to the real-time deviation between the optimized simulation results and the preset scheduling target, so as to realize the dynamic self-calibration of model parameters.

[0009] Compared with the prior art, the beneficial effects of the present invention are: This invention establishes a unified spatiotemporal data chain protocol for water conservancy, enabling efficient fusion and unified processing of multi-source heterogeneous water conservancy monitoring data. This breaks the "data silo" problem in traditional water conservancy systems and effectively improves the integration and collaborative utilization efficiency of water conservancy monitoring data.

[0010] This invention significantly improves the real-time performance and dynamic response capability of the simulation model by constructing an edge computing-driven coupled distributed hydrological and hydrodynamic simulation model and combining it with a multi-objective dynamic weight reinforcement learning algorithm, thus ensuring accurate prediction and rapid decision-making capabilities under extreme hydrological events such as sudden floods.

[0011] This invention introduces a time-series error regression analysis model to dynamically and in real-time self-correct model parameters, achieving efficient closed-loop collaboration between hardware resources and software systems. This ensures the continuous high precision and dynamic optimization capability of the water conservancy system during long-term operation, effectively improving the overall intelligent management level of water conservancy projects.

[0012] This invention achieves closed-loop collaboration of water conservancy data fusion, real-time accurate simulation, intelligent scheduling optimization and dynamic parameter self-correction, which significantly improves the real-time performance and accuracy of intelligent water conservancy decision-making. Attached Figure Description

[0013] Figure 1 This is a flowchart of the digital twin water conservancy hyperfusion method for coupling multi-source data in Example 1.

[0014] Figure 2 This is an architecture diagram of the digital twin water conservancy hyper-converged integrated machine that couples multi-source data in Example 2. Detailed Implementation

[0015] The present invention will now be further described with reference to the accompanying drawings and embodiments.

[0016] Example 1 Please see Figure 1 This invention provides a digital twin water conservancy hyperfusion method that couples multi-source data, including: A water conservancy spatiotemporal data chain protocol is established, and water conservancy monitoring data is formatted and processed through hierarchical data modeling. Edge computing nodes are determined based on the formatted water conservancy monitoring data and water affairs data. The predicted future load values ​​of the nodes are obtained based on the edge computing nodes, and a cloud-based fusion dataset is generated by selecting an appropriate data transmission format. Based on the cloud-based fusion dataset, a coupled distributed hydrological model and a hydrodynamic model are constructed. An initial scheduling scheme for the forecast period is generated based on rainfall and water information, and a forward simulation is performed to obtain the initial simulation results. Based on the difference data between the initial simulation results and the preset scheduling target, and using the difference data as feedback, a multi-objective dynamic weight reinforcement learning algorithm is used to reverse optimize and adjust the control parameters of the initial scheduling scheme in the foreseeability period to generate an optimized scheduling scheme in the foreseeability period. Based on the optimized scheduling scheme with a forward prediction period, a forward simulation was conducted again to verify the optimized simulation results. Then, based on the real-time deviation between the optimized simulation results and the preset scheduling target, the model parameters of the coupled distributed hydrological model and hydrodynamic model were dynamically corrected using a time-series error regression analysis model to achieve dynamic self-correction of the model parameters.

[0017] Considering the diverse data sources in this embodiment, and the significant differences in data structures generated by different types of sensors due to variations in monitoring principles and functional positioning, the water conservancy monitoring data formatting process described in this embodiment is as follows: A hierarchical modeling of water conservancy monitoring data yields a tuple data structure, which is represented as follows: ,in, Indicates the monitoring subject A unique identifier used to globally and uniquely identify water conservancy monitoring equipment or nodes; specifically, it can be the sensor number, device address, or network node number.

[0018] As the main monitoring subject The monitoring data values ​​may vary depending on the specific monitoring object, including water level, flow rate, water quality parameters, and meteorological parameters. If the monitoring object is a water level sensor, the corresponding monitoring data value is the water level. If the monitoring object is a flow rate sensor, the corresponding monitoring data values ​​are flow velocity, cross-sectional area, flow rate calculation method, etc. If the monitoring object is a water quality sensor, the corresponding monitoring data value is the water level. pH Mapping relationships between multiple water quality parameters and units, such as dissolved oxygen concentration and conductivity.

[0019] As the main monitoring subject Geographical location information, such as national elevation datum, latitude and longitude coordinates, or grid coordinates.

[0020] To collect timestamps, Coordinated Universal Time (UTC) is used. UTC This indicates that it is used to unify the time benchmark of data from different monitoring entities, so as to achieve accurate fusion and correlation of multi-source data.

[0021] In a further embodiment, the tuple data structure further includes: a monitoring subject. Monitoring principle This yields the five-element data structure, represented as: .

[0022] To determine the specific location of data processing, this embodiment uses the following method to identify edge computing nodes: The monitoring area is divided into several sub-areas based on water data; the water data mentioned in this embodiment may include the geographical environment, water system structure, node spatial layout and network communication characteristics of the monitoring area, etc.

[0023] By utilizing the BeiDou satellite positioning system to acquire the spatial locations of sensor nodes in each sub-region in real time, a topology scoring formula is established to calculate the topology centrality score of each sensor node. Topological centrality scoring The spatial priority order of each sensor node is determined. The topology scoring formula is expressed as follows: , in, This represents the total number of sensor nodes within the sub-region. Represents sensor nodes With sensor nodes The number of grids traversed during data transmission between them; Represents sensor nodes With sensor nodes Spatial distance between them and These represent the maximum number of grid cells and the maximum spatial distance between any two sensor nodes, respectively. and Let represent the weight coefficients of the network hop count factor and the spatial distance factor, respectively, and satisfy . ; Within each sub-region, score based on topological centrality. A primary edge computing node and several backup nodes are determined; and in this embodiment, the primary edge computing node and several backup nodes satisfy the following dynamic settings: A health scoring mechanism is established to monitor the health score of the primary edge computing node in real time. When the health score of the primary node is lower than a preset health threshold, the node with the highest health score is selected from the backup nodes as the new primary edge computing node. The health scoring mechanism described in this implementation can be based on... CPU The operating parameters such as occupancy rate, memory usage rate, and network communication latency are set.

[0024] The health score is between 0 and 1, where 1 indicates that the node is in the best condition and 0 indicates that the node is completely ineffective.

[0025] For example, within a certain sub-region, the initial health score of the current primary edge computing node is 0.95. As the operating load increases or network latency rises, the primary node's health score gradually decreases. When the primary node's health score drops below a preset health threshold (e.g., 0.6) (let's say 0.55), the system triggers a standby node switchover mechanism; at this time, a preset standby node (e.g., node...) is selected from the sub-region's backup node...A The rating is 0.75, node B The rating is 0.82, node C From the nodes with a rating of 0.68, the node with the highest rating is automatically selected. B (Score: 0.82) As a new primary edge computing node, it enables adaptive dynamic switching of nodes, ensuring that data processing and communication efficiency remains stable and reliable throughout the monitoring area.

[0026] Based on this, the steps to obtain the predicted future load value of a node are as follows: Obtain the historical runtime load sequence of the main edge computing node and historical bandwidth change sequence The historical runtime load sequence and historical bandwidth change sequence The features are concatenated according to the timestamp to obtain the fused feature sequence.

[0027] In the specific implementation process, the historical operating load sequence is formed by the main edge computing node within a continuous time window in the past. CPU The historical sequence of monitoring computing resources such as occupancy rate and memory usage rate can be represented as: , In the formula, This represents the historical load data sequence of the main edge computing node. Indicates the first q The measured operating load of the main edge computing node at any given time. t Indicates the current moment. T Indicates the length of the historical observation window. Indicates the start time of the historical observation window The operating load measurement value, and so on. Current moment The measured value of the operating load.

[0028] The historical bandwidth change rate data is a sequence composed of the bandwidth change rate of the main edge computing node at each moment within the same continuous time window relative to the previous moment, which can be represented as: , In the formula, This represents a sequence of historical bandwidth change rate data for the main edge computing nodes. Indicates the node at the 1st i The relative rate of change of bandwidth at each moment. Indicates the start time of the historical observation window The relative rate of change of bandwidth, and so on. For the current moment The relative rate of change of bandwidth.

[0029] A load prediction data encoding model is deployed on a defined edge computing node. This model embeds a multi-head attention mechanism, which includes... Attention units set up in parallel.

[0030] The load prediction data encoding model described in this embodiment adopts a gated cyclic unit ( GRU The network serves as the basic structure to capture the temporal dependencies in historical sequence data, thereby accurately predicting future node operating load trends.

[0031] The multi-head attention mechanism outputs a feature vector which is then processed by a fully connected layer and a linear activation function to obtain the predicted future load value of the node.

[0032] exist GRU Based on the network, this embodiment further embeds a multi-head attention mechanism, which includes several attention units set up in parallel, each attention unit focusing on the... GRU The historical data feature vectors output by the network are independently weighted to enhance the model's differentiated expression of the importance of node load features at different times, thereby improving the accuracy and reliability of node load trend prediction. Specifically, the comprehensive feature vector of the multi-head attention mechanism is obtained by concatenating the outputs of each attention unit and then performing a linear mapping. The specific calculation expression is as follows: , in, This represents the combined feature vector output by the multi-head attention mechanism. The first attention unit outputs a sub-output feature vector. The second attention unit outputs a sub-output feature vector. Indicates the first The output feature vector of each attention unit The fully connected mapping matrix obtained through training. This indicates a splicing operation.

[0033] The load prediction data encoding model described in this embodiment is trained and optimized using a joint loss function. Furthermore, the joint loss function, composed of node running load prediction error and bandwidth prediction error, is used as the optimization objective. The specific joint loss function can be expressed as follows: , In the formula, This represents the value of the joint loss function. This represents the total number of sample pairs in the training dataset. This indicates the load forecasting data encoding model for the first... The predicted value of the operating load of each sample node. Indicates the first The actual measured values ​​of the load running on each sample node. The model represents the first The predicted value of the bandwidth of each sample node. Indicates the first The actual measured value of the bandwidth of each sample node; and These represent the relative weighting coefficients of load prediction error and bandwidth prediction error in the joint loss function, respectively.

[0034] It should be noted that the weighting coefficients are set by technical personnel based on the actual scenario and the relative importance of load prediction error and bandwidth prediction error, and satisfy the following conditions: .

[0035] Using the above technical solution to obtain the predicted future load value of the node, the data transmission format is further determined. The specific process is as follows: The future node load value output by the node load prediction model is compared with a preset load threshold. If the predicted future node load is greater than or equal to the preset load threshold, it indicates that the node is about to enter a high-load state, and the node's computing resources are insufficient to support additional data compression operations. In this case, uncompressed data is used directly. JSON Data is transmitted in a specific format to reduce computational load and ensure stable node operation; If the predicted future load of the node is less than the preset load threshold, the data encoding and transmission format is further determined based on the relationship between the model's predicted future bandwidth and the preset bandwidth threshold. If the model predicts that the future bandwidth of the node is sufficient and greater than or equal to the preset bandwidth threshold, then uncompressed data is selected. JSON Data is transmitted in a specific format to avoid unnecessary compression computation overhead. If the model predicts that the future bandwidth of a node is less than a preset bandwidth threshold, a format with a high compression ratio and relatively low computational overhead is selected. CBOR Data is encoded and transmitted using compressed formats to reduce the data transmission burden and improve bandwidth utilization.

[0036] After the above load prediction and data encoding format selection process, the nodes encode the monitoring data in the determined data encoding format to form a fused dataset, and then upload the data to the cloud.

[0037] In one embodiment, the coupled distributed hydrological model and hydrodynamic model are multi-scale coupled models embedded with three lines of defense, and their construction process includes: A distributed hydrological model is constructed based on a cloud-based fusion dataset. A spatial weight matrix is ​​established based on the topographic relief and spatial heterogeneity of rainfall in sub-basins, and the model structure of the distributed hydrological model is constructed using the spatial weight matrix. By comprehensively considering the topographic and rainfall distribution characteristics of sub-basins, the spatial weight matrix makes differentiated adjustments to the model parameters and weights at different spatial locations, enabling the model to more accurately reflect the spatial differences in hydrological response in the actual watershed, thereby effectively improving the simulation accuracy of the model.

[0038] A hydrodynamic model is established, and a local water level gradient correction term is introduced into the numerical discretization scheme to correct the hydrodynamic model. It should be noted that the hydrodynamic model used in this embodiment is based on... Saint - Venant Unsteady flow equations are used to simulate water flow in a river channel; to effectively reduce numerical solution errors, this embodiment uses traditional... Saint - Venant Based on the numerical discrete solution scheme of the equation, an additional local water level gradient correction term is introduced for error compensation and correction.

[0039] The flow data output by the distributed hydrological model is linked with the boundary conditions of the hydrodynamic model through asynchronous data interaction to ensure the real-time and coordinated data transmission between the two models, ultimately forming a multi-scale coupled model with three lines of defense.

[0040] It should be noted that the data interaction between the coupled distributed hydrological model and the hydrodynamic model is asynchronous, meaning that the two models run independently at different time scales and exchange the data required for model calculation through a data interface; specifically, the watershed outlet discharge data calculated by the distributed hydrological model is used as the boundary condition input of the hydrodynamic model, and the time steps between the two are inconsistent. It should be understood that distributed hydrological models generally use a larger time step for calculation, such as a model time step of 1 hour, in order to ensure the efficiency of hydrological process calculation; while hydrodynamic models need to use a smaller time step, such as 10 minutes or less, to simulate the dynamic changes of water flow in the river channel more precisely. Therefore, there is a significant difference in the data time scale between the two models, and the flow data cannot be directly matched one-to-one. Asynchronous data exchange must be used for processing.

[0041] To facilitate the implementation of the above technical solution, the method for determining the spatial weight matrix in this embodiment includes: The watershed was divided into multiple sub-units, and the topographic relief and spatial variation coefficient of rainfall were obtained for each sub-unit.

[0042] The geomorphic weight of each sub-unit is determined based on the topographic relief and the spatial variability coefficient of rainfall. With rainfall weight The specific calculation formula is as follows: , , In the formula, and These represent the maximum and minimum values ​​of the terrain relief for all sub-units, respectively. Indicates rainfall weight, This represents the spatial variability coefficient of rainfall in the sub-unit. and This represents the minimum and maximum values ​​of the spatial variation coefficient of rainfall across all sub-units.

[0043] The spatial weight of each sub-unit is obtained by combining the landform weight and the rainfall weight, and the spatial weight matrix is ​​constructed using the spatial weights of each sub-unit.

[0044] It should be noted that the spatial weight matrix is ​​composed of spatial weight values ​​determined by combining the topographic weight and rainfall weight of the sub-units. The specific calculation method is as follows: , In the formula, This represents the spatial weight value of the sub-unit, and the weight coefficient. and Let these represent the relative importance of geomorphic features and rainfall features in the spatial weights, respectively, and satisfy the following conditions: The weighting coefficients are obtained based on field experience and long-term data analysis.

[0045] The construction of the spatial weight matrix can be used to optimize the parameter configuration of the distributed hydrological model, so as to achieve accurate optimization of the model structure at the spatial scale.

[0046] In practical implementation, the local water level gradient correction term is used to correct the local water level gradient in the Saint-Venant equation calculation, and the calculation method is as follows: , In the formula, This represents the corrected local water level gradient. This indicates that the model initially calculates the local water level gradient. This represents the local water level gradient correction coefficient to be self-corrected; the correction coefficient is dynamically adjusted. To effectively reduce the error between model-simulated water levels and actual observations; In another embodiment, the preset scheduling objectives include, but are not limited to, various quantitative indicators such as: the in-channel flow control objective, the in-channel water level exceeding the warning time control objective, the inundation range control objective outside the channel, and the inundation duration control objective.

[0047] The difference data is determined based on the difference between the initial simulation results and the preset scheduling target. The calculation formula for the difference data includes: The calculation is based on the initial simulation results and the preset scheduling target, and the formula is as follows: , In the formula, Indicates difference data, Indicates the initial simulation result index. Indicates the preset scheduling target; Judge the difference data, if A value greater than zero indicates that the simulation performance exceeds the preset target and needs to be optimized and reduced; if A value less than zero indicates that the simulation performance has not met the preset target and needs to be optimized and improved; if... A value of zero indicates that the simulation performance has met the preset target and no optimization is needed.

[0048] The difference data is determined based on the difference between the initial simulation results and the preset scheduling target. This difference data is then used as the feedback input of the multi-objective dynamic weight reinforcement learning algorithm to guide the reverse optimization adjustment of the control parameters, thereby generating an optimized predictive scheduling scheme.

[0049] Furthermore, the control parameters are optimized based on the feedback differential data through a reinforcement learning policy network. In this embodiment, the reinforcement learning policy network is trained using a deep deterministic policy gradient algorithm to achieve efficient prediction of control parameter optimization.

[0050] In another implementation, based on the specific control parameter settings in the optimized scheduling scheme, the distributed hydrological model and the hydrodynamic model are re-invoked for joint forward simulation to obtain predicted values ​​of various hydrological indicators reflecting the actual effect of the optimized scheduling scheme, thus forming optimized simulation results. However, the simulation model itself has certain parameter errors and simulation errors, which may still result in a certain real-time deviation between the optimized simulation results and the actual preset scheduling target. In this embodiment, a time-series error regression analysis model is used to dynamically correct the model parameters, realizing dynamic self-correction of the parameters of the distributed hydrological model and the hydrodynamic model, thereby improving the simulation accuracy of the model and the accuracy of future predictions.

[0051] For example, by dynamically adjusting the coefficients in the correction of the local water level gradient To effectively reduce the error between model-simulated water levels and actual observations; by dynamically adjusting the weighting coefficients. and weighting coefficients This ensures that the spatial weight matrix reflects changes in current watershed rainfall and geomorphological features in real time, thereby improving the spatial prediction accuracy of the hydrological model. The parameter self-calibration process described in this embodiment is achieved through a time-series error regression analysis model.

[0052] Specifically, the time-series error regression analysis model is a long short-term memory network that integrates the error accumulation trend. It uses the time-series error data between the simulation results of the optimized scheduling scheme and the actual scheduling target as input, and the model parameter correction amount as output to construct a training set. It is trained using the root mean square error between the error prediction value and the actual error as the loss function, and outputs a regression model for correcting the parameters of the distributed hydrological model and the hydrodynamic model.

[0053] In summary, to address the issues of data interaction and computational accuracy between hydrological and hydrodynamic models, a multi-scale coupled model with three embedded defense lines is established, comprising multiple computational stages. These three defense lines, as described above, are: The first line of defense is to use a spatial weight matrix structure based on the geomorphic relief and spatial heterogeneity of rainfall within the distributed hydrological model, so as to accurately capture the spatial heterogeneity characteristics of the watershed hydrological response and prevent systematic errors caused by insufficient description of spatial heterogeneity from the model structure.

[0054] The second line of defense involves introducing a local water level gradient correction term during the numerical solution process within the hydrodynamic model. This term is dynamically corrected by real-time monitoring data feedback to prevent the propagation and accumulation of local errors in the model from the computational stage.

[0055] The third line of defense is to use an asynchronous interaction combined with interpolation processing data fusion mechanism at the data interaction interface between the distributed hydrological model and the hydrodynamic model. This prevents the amplification of data errors caused by time scale mismatch from the data transmission stage, and ensures the real-time and coordinated transmission of data between the two models.

[0056] Example 2 like Figure 2 As shown, this embodiment discloses a digital twin water conservancy hyper-converged integrated machine that couples multi-source data, used to implement the digital twin water conservancy hyper-converged method described in Embodiment 1, including: Data fusion module: used to establish a water conservancy spatiotemporal data chain protocol, realize the formatted processing of water conservancy monitoring data through hierarchical data modeling; determine edge computing nodes based on formatted water conservancy monitoring data and water affairs data, obtain the predicted future load value of the nodes after edge computing processing based on the edge computing nodes, and generate a cloud fusion dataset by selecting an appropriate data transmission format. Coupled hydrological simulation module: used to construct a coupled distributed hydrological model and a hydrodynamic model based on the cloud-based fusion dataset, generate an initial scheduling scheme for the forecast period based on rainfall and water information, and perform forward simulation to obtain initial simulation results; The scheduling optimization module is used to optimize and adjust the control parameters of the initial scheduling scheme in the foresight period based on the difference data between the initial simulation results and the preset scheduling target, using the difference data as feedback, and employing a multi-objective dynamic weight reinforcement learning algorithm to generate an optimized scheduling scheme in the foresight period. The parameter self-calibration module is used to perform forward simulation pre-run verification based on the forecast period optimized scheduling scheme, obtain optimized simulation results, and dynamically correct the model parameters of the coupled distributed hydrological model and hydrodynamic model using a time series error regression analysis model according to the real-time deviation between the optimized simulation results and the preset scheduling target, so as to realize the dynamic self-calibration of model parameters.

[0057] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0058] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A digital twin-based hyperfusion method for hydraulic engineering that couples multi-source data, characterized in that, include: Establish a water conservancy spatiotemporal data chain protocol and realize the formatted processing of water conservancy monitoring data through hierarchical data modeling; Edge computing nodes are determined based on formatted water conservancy monitoring data and water affairs data. The predicted future load values ​​of the nodes are obtained based on the edge computing nodes, and a cloud-based fusion dataset is generated by selecting an appropriate data transmission format. Based on the cloud-based fusion dataset, a coupled distributed hydrological model and a hydrodynamic model are constructed. An initial scheduling scheme for the forecast period is generated based on rainfall and water information, and a forward simulation is performed to obtain the initial simulation results. Based on the difference data between the initial simulation results and the preset scheduling target, and using the difference data as feedback, a multi-objective dynamic weight reinforcement learning algorithm is used to reverse optimize and adjust the control parameters of the initial scheduling scheme in the foreseeability period to generate an optimized scheduling scheme in the foreseeability period. Based on the optimized scheduling scheme with a forward prediction period, a forward simulation was conducted again to verify the optimized simulation results. Then, based on the real-time deviation between the optimized simulation results and the preset scheduling target, the model parameters of the coupled distributed hydrological model and hydrodynamic model were dynamically corrected using a time-series error regression analysis model to achieve dynamic self-correction of the model parameters.

2. The digital twin hydraulic super-fusion method for coupling multi-source data according to claim 1, characterized in that, The process of formatting and processing the water conservancy monitoring data is as follows: A hierarchical modeling of water conservancy monitoring data yields a tuple data structure, which is represented as follows: ,in, Indicates the monitoring subject Unique identifier, As the main monitoring subject The monitoring data values, As the main monitoring subject Location information For collection timestamps.

3. The digital twin hydraulic hyperfusion method for coupling multi-source data according to claim 1, characterized in that, The process of determining the edge computing node includes: Based on water resources data, the monitoring area is divided into several sub-regions; the spatial location information of sensor nodes within each sub-region is obtained, a topology scoring formula is established, and the topology centrality score of each sensor node is calculated. , Let be the number of the sensor node; wherein, the topology scoring formula is expressed in the following form: ， in, This represents the total number of sensor nodes within the sub-region. Represents sensor nodes With sensor nodes The number of grids traversed during data transmission; Represents sensor nodes With sensor nodes Spatial distance between them and These represent the maximum number of grid cells and the maximum spatial distance between any two sensor nodes, respectively. and These represent the weighting coefficients for the network hop count factor and the spatial distance factor, respectively. Within each sub-region, score based on topological centrality. A primary edge computing node and several backup nodes are determined; the primary edge computing node and the several backup nodes satisfy the following dynamic settings: Establish a health scoring mechanism to monitor the health score of the main edge computing node in real time. When the health score of the main node is less than the preset health threshold, select the node with the highest health score from the backup nodes as the new main edge computing node.

4. The digital twin hydraulic super-fusion method for coupling multi-source data according to claim 1, characterized in that, The steps for obtaining the predicted future load value of the node are as follows: Obtain the historical runtime load sequence of the main edge computing node and historical bandwidth change sequence The features are then concatenated according to the timestamps to obtain a fused feature sequence. A load prediction data encoding model is deployed on a defined edge computing node. This model embeds a multi-head attention mechanism and a joint loss function. The multi-head attention mechanism includes... A parallel set of attention units; The fused feature sequence is taken as input and processed by an attention unit. Output sub-output feature vector And obtain the comprehensive feature vector by using the mapping relationship. The multi-head attention mechanism outputs a comprehensive feature vector. After processing by a fully connected layer and a linear activation function, the predicted future load value of the node is obtained; The mapping relationship is expressed in the following form: In the formula, This indicates a splicing operation. The fully connected mapping matrix obtained through training. The first attention unit outputs a sub-output feature vector. The second attention unit outputs a sub-output feature vector.

5. The digital twin hydraulic super-fusion method for coupling multi-source data according to claim 1, characterized in that, The method for adapting the data transmission format is as follows: The future node load value output by the node load prediction model is compared with the preset load threshold. If the predicted future load on a node is greater than or equal to a preset load threshold, uncompressed load will be used. JSON Data format; If the predicted future load of the node is less than the preset load threshold, the data encoding and transmission format is further determined based on the relationship between the predicted future bandwidth of the node and the preset bandwidth threshold in the load prediction data encoding model. If the predicted future bandwidth of a node is greater than or equal to a preset bandwidth threshold, then uncompressed bandwidth is selected. JSON Data format; If the predicted future bandwidth of a node is less than the preset bandwidth threshold, then select... CBOR Compressed format.

6. The digital twin hydraulic hyperfusion method for coupling multi-source data according to claim 1, characterized in that, The coupled distributed hydrological model and hydrodynamic model are multi-scale coupled models with three lines of defense embedded in them. The three lines of defense include: the first line of defense, the second line of defense, and the third line of defense.

7. The digital twin hydraulic hyperfusion method for coupling multi-source data according to claim 6, characterized in that, The coupled distributed hydrological model and hydrodynamic model include a distributed hydrological model, which has a spatial weight matrix structure based on the geomorphic relief of sub-basins and the spatial heterogeneity of rainfall, forming the first line of defense; wherein, the method for determining the spatial weight matrix includes: The sub-basin was divided into multiple sub-units, and the topographic relief and spatial variability coefficient of rainfall were obtained for each sub-unit. The topographic weight and rainfall weight of each sub-unit are determined based on the topographic relief and the spatial variation coefficient of rainfall. The spatial weight value of each sub-unit is obtained by combining the topographic weight and the rainfall weight, and a spatial weight matrix is ​​constructed based on the spatial weight values ​​of each sub-unit. The formula for calculating the spatial weight value of each sub-unit is as follows: ， In the formula, This represents the spatial weight value of the sub-unit. Indicates the weight of the terrain. Indicates rainfall weight, weighting coefficient and Let these represent the relative importance of geomorphic features and rainfall features in the spatial weights, respectively, and satisfy the following conditions: .

8. The digital twin hydraulic hyperfusion method for coupling multi-source data according to claim 6, characterized in that, The coupled distributed hydrological model and hydrodynamic model include: a hydrodynamic model coupled to the distributed hydrological model, which introduces a local water level gradient correction term to correct the hydrodynamic model and form a second line of defense; The correction of the local water level gradient correction term is achieved through a parameter self-calibration module. The parameter self-calibration module adopts a time-series error regression analysis model and dynamically adjusts the local water level gradient correction coefficient k in the local water level gradient correction term based on the real-time deviation between the optimized simulation results and the preset scheduling target.

9. The digital twin hydraulic super-fusion method for coupling multi-source data according to claim 6, characterized in that, At the data interaction interface between the distributed hydrological model and the hydrodynamic model, an asynchronous interaction combined with interpolation processing data fusion mechanism is adopted to form a third line of defense.

10. A digital twin hydraulic hyper-converged integrated machine coupling multi-source data, characterized in that, The integrated machine is equipped with multiple modules for implementing the digital twin water conservancy hyper-fusion method for coupling multi-source data as described in any one of claims 1-9, wherein the modules include: Data fusion module: used to establish a water conservancy spatiotemporal data chain protocol, realize the formatted processing of water conservancy monitoring data through hierarchical data modeling; determine edge computing nodes based on formatted water conservancy monitoring data and water affairs data, obtain the predicted future load value of the nodes based on the edge computing nodes, and select an appropriate data transmission format to generate a cloud-based fused dataset; Coupled hydrological simulation module: used to construct a coupled distributed hydrological model and a hydrodynamic model based on the cloud-based fusion dataset, generate an initial scheduling scheme for the forecast period based on rainfall and water information, and perform forward simulation to obtain initial simulation results; The scheduling optimization module is used to optimize and adjust the control parameters of the initial scheduling scheme in the foresight period based on the difference data between the initial simulation results and the preset scheduling target, using the difference data as feedback, and employing a multi-objective dynamic weight reinforcement learning algorithm to generate an optimized scheduling scheme in the foresight period. The parameter self-calibration module is used to perform forward simulation pre-run verification based on the forecast period optimized scheduling scheme, obtain optimized simulation results, and dynamically correct the model parameters of the coupled distributed hydrological model and hydrodynamic model using a time series error regression analysis model according to the real-time deviation between the optimized simulation results and the preset scheduling target, so as to realize the dynamic self-calibration of model parameters.

Images