A water conservancy digital management method and system based on digital twinning

Abstract

This application discloses a digital management method and system for water conservancy based on digital twins, relating to the field of smart water conservancy technology. In response to early warnings of extreme hydrological events, this application initiates cross-validation of hydrological sensors, identifies and masks abnormal sensor data; when data from the main control monitoring section is interrupted, it infers compensatory state data based on real-time upstream and downstream data and the inherent physical properties of the river channel and injects it into the digital twin; in response to instructions for designated protection targets, it divides the whole-basin flood simulation into high-resolution sub-tasks in the core protection zone and low-resolution sub-tasks in the non-core area, prioritizing the execution of core area simulations, and generating joint flood discharge instructions for the reservoir group based on its output. This application ensures the fidelity and decision support capabilities of the digital twin in emergency situations.

Description

Technical Field

[0001] This application relates to the field of smart water conservancy technology, and in particular to a digital management method and system for water conservancy based on digital twins. Background Technology

[0002] Digital twin technology is widely used in current water conservancy management systems. Generally, digital twin technology is used to build an integrated virtual model of watershed hydrology and engineering. This virtual model can support flood forecasting and scheduling decisions to the greatest extent.

[0003] However, in sudden events such as extreme rainstorms, hydrological sensors deployed in the field are prone to damage or communication disruptions, which may lead to missing or distorted data at key cross-sections. At the same time, flood control command centers need to complete high-precision flood evolution simulations in a very short time. Only when both time and accuracy requirements are met can reservoir discharge and personnel evacuation instructions be issued.

[0004] However, some existing digital twin systems typically rely on complete and continuous sensor data streams to update the state. If the data source is abnormal, it will cause the twin to deviate from the physical world, resulting in a significant decrease in the reliability of subsequent simulation results.

[0005] In addition, unified high-resolution simulation calculations across the entire watershed are time-consuming and cannot meet the timeliness requirements of emergency response. Summary of the Invention

[0006] In view of the aforementioned problems, this application is hereby filed.

[0007] Therefore, this application provides a digital management method and system for water conservancy based on digital twins, which can maintain the accurate mapping of the watershed status of the digital twin when monitoring data is abnormal or interrupted due to extreme hydrological events, and support minute-level emergency dispatch decisions.

[0008] To solve the above-mentioned technical problems, this application provides the following technical solution: In a first aspect, this application provides a digital management method for water conservancy based on digital twins, including: in response to receiving an early warning signal characterizing an extreme hydrological event, activating a digital twin corresponding to a target watershed, and initiating cross-validation of data from all online hydrological sensors within the target watershed; Based on the results of the cross-validation of the data, abnormal sensors whose data deviation exceeds the tolerance are identified, and state correction data excluding the contribution of the abnormal sensor data is generated to update the digital twin. In response to an interruption in the data source of any main control monitoring section in the digital twin, the system calls up the real-time data of the upstream section, the real-time data of the downstream section, and the inherent physical properties of the river channel to deduce and generate compensatory state data of the main control monitoring section, and injects the compensatory state data into the digital twin. In response to receiving an emergency dispatch instruction for a designated protection target, the whole-basin flood evolution simulation task based on the digital twin is divided into a high-resolution simulation subtask for the core protection area and a low-resolution simulation subtask for the non-core area. Prioritize scheduling computing resources to execute high-resolution simulation subtasks in the core protected area, and generate scheduling instructions to regulate the flood discharge of the upstream reservoir group based on the flood risk situation output by the high-resolution simulation subtasks.

[0009] Preferably, the step of initiating cross-validation of data from all online hydrological sensors within the target watershed includes: Define spatially adjacent validation clusters for each type of hydrological sensor; Calculate the average real-time data of similar hydrological sensors within each validation cluster; The real-time data of each hydrological sensor is compared with the average real-time data of its respective validation cluster to determine the deviation. Hydrological sensors whose deviation comparison results exceed the preset tolerance are marked as sensors to be verified.

[0010] Preferably, the step of identifying abnormal sensors whose data deviation exceeds the tolerance and generating state correction data excluding the contribution of the abnormal sensor data includes: If the deviation comparison result of any sensor to be verified continues to exceed a preset time, the sensor to be verified is identified as an abnormal sensor. When updating the hydrological status of the digital twin, real-time data from all confirmed abnormal sensors are masked. The state correction data is generated by fusing real-time data from all non-abnormal sensors.

[0011] Preferably, the deduction to generate compensatory state data for the main control monitoring section includes: Retrieve the stored river cross-section topographic data and riverbed roughness data associated with the main control monitoring section; Based on the real-time flow and water level of the upstream section, combined with the topographic data and riverbed roughness data of the river section, the expected water level of the main control monitoring section is calculated through hydraulic continuity logic. The rationality of the expected water level is verified by reverse analysis based on the real-time water level at the downstream section. The expected water level obtained through reverse verification will be used as the compensating state data.

[0012] Preferably, the method of dividing the whole-basin flood evolution simulation task into a high-resolution simulation sub-task for the core protected area and a low-resolution simulation sub-task for the non-core area includes: Parse the emergency dispatch command and extract the geographical range of the specified protection target; Water conservancy elements within the geographical area are designated as core protection zones, and areas outside the geographical area are designated as non-core zones. A first grid resolution is configured for the core protected area, and a second grid resolution is configured for the non-core area, wherein the first grid resolution is higher than the second grid resolution; Based on the first grid resolution and the second grid resolution, the high-resolution simulation subtask and the low-resolution simulation subtask are constructed respectively.

[0013] Preferably, the prioritized scheduling of computing resources executes the high-resolution simulation subtask of the core protected area and generates scheduling instructions for regulating the flood discharge of the upstream reservoir group, including: Allocate no less than a preset proportion of total computing resources to the high-resolution simulation subtasks of the core protected area; Real-time monitoring of the flood risk situation output by the high-resolution simulation subtask of the core protected area; In response to the flood risk situation reaching the warning level, the current water storage status of the upstream reservoir group is extracted; Based on the flood risk situation and the current water storage status of the upstream reservoir group, a joint flood discharge command for the reservoir group is generated to reduce the flood risk of the core protected area.

[0014] Preferably, the method further includes: In response to the completion of the high-resolution simulation subtask of the core protected area, the low-resolution simulation subtask of the non-core area is initiated. The output results of the low-resolution simulation subtask in the non-core area are coupled with the boundary output results of the high-resolution simulation subtask in the core protection area to generate a comprehensive flood evolution status map of the entire basin.

[0015] Preferably, the method further includes: In response to the end of the extreme hydrological event, cross-validation of data from all online hydrological sensors is stopped. The digital twin is restored to its normal operating mode, and the identity information of all abnormal sensors and the generation log of compensatory status data during this emergency response are recorded.

[0016] Preferably, the inherent physical properties of the river channel include at least the river channel cross-sectional topographic data and the riverbed roughness data.

[0017] Secondly, this application also provides a digital management system for water conservancy based on digital twins, including: a twin activation module, configured to activate a digital twin corresponding to a target watershed in response to receiving an early warning signal characterizing an extreme hydrological event, and to initiate cross-validation of data from all online hydrological sensors within the target watershed; The state correction module is configured to identify abnormal sensors whose data deviation exceeds the tolerance based on the results of the data cross-validation, and generate state correction data that excludes the contribution of the abnormal sensor data to update the digital twin. The state compensation module is configured to respond to an interruption in the data source of any master monitoring section in the digital twin, call the real-time data of the upstream section, the real-time data of the downstream section, and the inherent physical properties of the river channel of the master monitoring section, deduce and generate compensatory state data of the master monitoring section, and inject the compensatory state data into the digital twin. The simulation partitioning module is configured to, in response to receiving an emergency dispatch command for a specified protection target, divide the whole-basin flood evolution simulation task based on the digital twin into a high-resolution simulation subtask for the core protection area and a low-resolution simulation subtask for the non-core area. The instruction generation module is configured to prioritize scheduling computing resources to execute high-resolution simulation subtasks of the core protected area, and generate scheduling instructions to regulate the flood discharge of the upstream reservoir group based on the flood risk situation output by the high-resolution simulation subtasks.

[0018] Implementing this application has the following beneficial effects: This application provides a digital management method and system for water conservancy based on digital twins. In response to early warnings of extreme hydrological events, it initiates cross-validation of hydrological sensors, identifies and masks abnormal sensor data; when data from key sections is interrupted, it infers compensatory state data based on real-time upstream and downstream data and the inherent physical properties of the river channel and injects it into the digital twin; in response to instructions for designated protection targets, it divides the whole-basin flood simulation into high-resolution sub-tasks in the core protection zone and low-resolution sub-tasks in the non-core area, prioritizing the execution of core area simulations and generating joint flood discharge instructions for the reservoir group based on its output. This application ensures the fidelity and decision support capabilities of the digital twin in emergency situations. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1This is an overall flowchart of a digital twin-based water conservancy digital management method involved in this application; Figure 2 This is a timeline flowchart of a digital management method for water conservancy based on digital twins, which is the subject of this application. Figure 3 This is a computer equipment diagram of a digital management method for water conservancy based on digital twins, which is the subject of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0022] In one exemplary embodiment, Figure 1 A flowchart of a digital management method for water conservancy based on digital twins is shown, and Figure 2 A timeline flowchart of a digital management method for water conservancy based on digital twins is shown, which specifically includes: S1, in response to receiving an early warning signal characterizing an extreme hydrological event, activates the digital twin corresponding to the target watershed and initiates cross-validation of data from all online hydrological sensors within the target watershed.

[0023] Currently, the standard practice in watershed management is to rely on complete sensor data streams for state synchronization under normal operating conditions in water conservancy digital twin systems. These solutions typically rely on static data access pipelines or configure fixed data source lists to meet the twin's update needs, but they generally fail to effectively achieve the ability to automatically identify and isolate abnormal data under extreme events.

[0024] While some technologies have proposed multi-source data fusion solutions, their core mechanisms still rely on post-event manual verification or offline data cleaning for anomaly correction. Therefore, such multi-source fusion solutions not only experience significant latency during operation but may also suffer from twin state drift due to the continuous injection of abnormal data. This state drift operation may also produce erroneous flood evolution projection results.

[0025] This application activates the digital twin of the target watershed in response to the early warning signal characterizing extreme hydrological events. Furthermore, it initiates a data cross-validation mechanism for all online hydrological sensors while activating the twin. Then, it forms a dynamic validation cluster of similar sensors that are spatially adjacent to each other to support real-time reliability assessment, ultimately achieving a closed-loop control for data fidelity from passive to active verification.

[0026] It should be noted that early warning signals that characterize extreme hydrological events may include red rainstorm warnings issued by meteorological departments, water level exceeding warning levels reported by hydrological stations, or emergency alarms from reservoir dam safety monitoring systems.

[0027] It should be emphasized that the aforementioned digital twin can be understood as a virtualized model entity used in this application to map the target watershed's river channels, reservoirs, dikes, and hydrological elements, and its state is updated driven by real-time sensor data.

[0028] In this application, the initiation of data cross-validation can prevent abnormal data from contaminating the twin in the early stages of extreme events, providing a reliable input basis for subsequent high-confidence simulations.

[0029] In some embodiments, cross-validation of data from all online hydrological sensors within the target watershed is initiated, including: S11 defines spatially adjacent validation clusters for each type of hydrological sensor.

[0030] It should be understood that the spatially adjacent verification cluster mentioned in S11 can be a circular area centered on a certain reference sensor with a radius not exceeding the typical hydrological response scale of the watershed, or it can be a linear area composed of adjacent upstream and downstream sections defined along the river channel, or it can be a set of multiple points within the same catchment sub-unit divided based on the digital elevation model. This application does not limit this.

[0031] It should also be understood that each type of hydrological sensor mentioned in S11 includes rain gauges, water level gauges, and flow meters. Specifically, they can be classified according to the sensor type first, and then based on the spatial topology relationship in the geographic information system, several non-overlapping or partially overlapping verification clusters can be constructed for each type of sensor.

[0032] It should be noted that if a certain type of hydrological sensor is a water level gauge, then the multiple validation clusters of this water level gauge are a continuous cross-sectional group distributed along the main river channel. For example, in a certain section of the middle reaches of the Yangtze River, three water level stations that are no more than ten kilometers apart can be classified as a validation cluster to reflect the consistency of water level changes in a local section of the river.

[0033] It should also be noted that if a certain type of hydrological sensor is a rain gauge, then the multiple validation clusters of this rain gauge are representative rainfall areas divided based on Thiessen polygons. For example, rain gauges located within the same micro-topographic unit, such as valleys or plateaus, are grouped into a validation cluster to capture the characteristics of local heavy rainfall.

[0034] In this application, defining validation clusters ensures that cross-validation is conducted within spatial units with hydrological similarities, thus avoiding the misjudgment of reasonable data as abnormal due to differences in terrain or distance.

[0035] S12, calculate the average real-time data of the same type of hydrological sensor within each validation cluster.

[0036] It should be noted that in actual operation, even if different hydrological sensors are of the same type, they may experience slight data fluctuations due to differences in the microenvironment of their installation locations, and the magnitude of these fluctuations is closely related to the consistency of data within the validation cluster.

[0037] For example, if the readings of three water level gauges in a certain verification cluster differ by 0.1 meters due to local scouring of the riverbed, it is still within the normal range and will not cause misjudgment.

[0038] If the data within the verification cluster is highly consistent, it indicates that the current hydrological state is stable. If only a single sensor reading is used as the basis for twin updates, the consensus information of the group will be ignored, which will lead to a lack of robustness to isolated outliers.

[0039] Therefore, the average real-time data of similar hydrological sensors within each validation cluster can be calculated to accurately depict the true hydrological state of the local area.

[0040] Optionally, calculating the average real-time data of similar hydrological sensors within each validation cluster may include the following operations performed in sequence: Obtain real-time data from all online hydrological sensors at the current moment from the data acquisition gateway; Furthermore, the data is grouped according to sensor type and the validation cluster to which it belongs; Furthermore, obviously invalid values, such as negative water levels or excessive rainfall, are removed from the real-time data within each group. Furthermore, the arithmetic mean of the remaining valid data is calculated and used as the real-time data mean of the validation cluster.

[0041] It should be noted that the data acquisition gateway mentioned above includes, but is not limited to, RTU remote terminal units or edge computing nodes. The rules for removing invalid values ​​can be set according to the actual needs of relevant technical personnel, and this application does not impose any restrictions.

[0042] It should also be noted that the aforementioned real-time data average can be understood as the core reference value used in this application to characterize the consensus level of hydrological status within the verification cluster.

[0043] In this application, the calculation of the real-time data mean can suppress individual sensor noise through group averaging, thereby improving the stability of state estimation.

[0044] S13, compare the real-time data of each hydrological sensor with the average real-time data of its respective validation cluster to determine the deviation.

[0045] It is worth noting that different hydrological sensors have varying reliability in sensing the watershed state in practice. Furthermore, reliability is closely related to the degree to which their real-time data deviates from the population mean. A larger deviation indicates that the sensor is more likely to be affected by local interference or equipment failure, leading to unreliable data and a weakened contribution to the twin's state update. Therefore, it is necessary to classify reliability according to the degree of deviation.

[0046] Optionally, for each hydrological sensor, the absolute or relative deviation from the average real-time data of its respective validation cluster can be calculated. That is, the absolute value of the difference between the two is taken as the absolute deviation, or the ratio of the absolute deviation to the average is taken as the relative deviation, and this is used to determine whether the tolerance is exceeded. For example, in a certain scenario, the average water level of the validation cluster is 25.0 meters, and a water level gauge reading is 25.8 meters, with an absolute deviation of 0.8 meters. If the preset tolerance is 0.5 meters, it is determined to be an over-limit.

[0047] If a relative deviation is used, it can adapt to the fluctuation characteristics of different water level stages. For example, when the average water level is 5 meters during the dry season, a deviation of 0.3 meters is considered abnormal; while when the average water level is 30 meters during the flood season, a deviation of 0.8 meters may still be within a reasonable range.

[0048] After completing the above deviation comparison operation, you can decide whether to mark the sensor as an object to be verified based on the comparison results.

[0049] The aforementioned deviation comparison can be understood as a key judgment operation used in this application to quantify the difference between single sensor data and group consensus.

[0050] In this application, deviation comparison can achieve quantitative identification of abnormal data, providing a basis for decision-making for subsequent shielding or compensation.

[0051] S14, mark hydrological sensors whose deviation comparison results exceed the preset tolerance as sensors to be verified.

[0052] Understandably, if the sensor data is permanently removed from the twin update when the deviation comparison result exceeds the preset tolerance, it will result in the loss of a valid data source after the sensor is briefly interfered with, such as when leaves block the rain gauge.

[0053] However, if the deviation results are completely ignored, severely faulty sensors, such as floating level gauges, may continue to inject erroneous data, resulting in falsely high twin water levels.

[0054] Therefore, after detecting that the deviation exceeds the limit, this application initiates a transient marking mechanism to mark the hydrological sensor as a sensor to be verified, rather than immediately removing it.

[0055] In this embodiment of the application, the specific operation of marking a hydrological sensor whose deviation comparison result exceeds a preset tolerance as a sensor to be verified may include the following steps: First, if the deviation comparison result exceeds the preset tolerance, a sensor verification identifier is generated; Furthermore, the identifier to be verified is bound to the sensor identity information and written to the abnormal sensor cache queue; Furthermore, continuous monitoring of the abnormal sensors is initiated to record whether subsequent data recovers to within the tolerance range.

[0056] The aforementioned sensor to be verified can be understood as a data source in a state of undetermined credibility in this application. Its data will not participate in the twin state update for the time being, but the recovery channel is reserved.

[0057] In this application, the sensor marked as to be verified can retain the ability to recover from temporary anomalies while ensuring the current accuracy of the twin, thus avoiding data sparsity caused by excessive culling.

[0058] S2, based on the results of data cross-validation, identifies abnormal sensors whose data deviation exceeds the tolerance, and generates state correction data that excludes the contribution of abnormal sensor data to update the digital twin.

[0059] It should be pointed out that the problem of digital twins failing to automatically correct their state when sensor data is abnormal, as present in existing technologies, can be solved by dynamically identifying the source of the anomaly and reconstructing reliable state data based on cross-validation results. Only after completing the identification of abnormal sensors and the generation of state correction data can the digital twin maintain a true mapping of the physical watershed in extreme events, thereby supporting high-confidence emergency dispatch decisions and avoiding misjudgments of flood discharge commands due to state distortion.

[0060] In some implementations, the final confirmation and state correction of anomalous sensors can be achieved using a dual verification method based on spatiotemporal consistency and persistence. For example, a two-stage mechanism of marking followed by confirmation can be adopted, combined with a data fusion strategy from non-anomalous sensors, to reconstruct the hydrological state of the target watershed. This dual verification method ensures that the anomaly detection is neither too sensitive nor too sluggish, facilitating the accurate execution of subsequent state corrections.

[0061] In practice, all sensors to be validated and their deviation sequences can be extracted from the cross-validation results. For example, in a rainstorm event, the system identifies five water level gauges in a state of needing validation and records their deviation values ​​for 10 consecutive minutes.

[0062] Furthermore, for the aforementioned sensors to be verified, it is determined whether the deviation exceeding the limit is persistent. For example, if a water level gauge has a deviation exceeding 0.6 meters for five consecutive sampling cycles, it is identified as an abnormal sensor. Upon confirmation, an abnormality flag is generated for the abnormal sensor, and it is removed from the list of valid data sources.

[0063] Furthermore, based on the real-time data from the remaining non-abnormal sensors, weighted averaging or interpolation methods are used to generate state correction data covering the entire watershed. For example, linear interpolation of upstream and downstream water levels is used for missing sections, and inverse distance weighted interpolation is used for the rainfall field.

[0064] Furthermore, after generating the state correction data, it is injected into the digital twin, triggering a state reinitialization of its internal hydrodynamic model. The resulting state correction data constitutes the reliable hydrological state set after eliminating interference from abnormal sensors.

[0065] It should be noted that the aforementioned state correction data can be understood as a key intermediate product used in this application to replace the original sensing input and ensure the fidelity of the twin.

[0066] In this application, the generated state correction data can still maintain the accurate representation of the overall state of the watershed by the twin even when the data source partially fails.

[0067] In some embodiments, aberrant sensors whose data deviation exceeds the tolerance are identified, and state correction data excluding contributions from aberrant sensor data is generated, including: S21, in response to the deviation comparison result of any sensor to be verified continuously exceeding a preset time, the sensor to be verified is identified as an abnormal sensor.

[0068] In actual operation, hydrological sensors may generate momentary abnormal data due to brief interference, such as obstruction by floating objects or communication jitter. Such momentary abnormalities are difficult to distinguish from permanent equipment failures in the early stages.

[0069] If a sensor to be verified only exceeds the deviation limit at a single moment, it is likely that it is only subject to temporary interference. If it is immediately identified as an abnormal sensor, it will lead to the incorrect removal of valid data sources, and may even result in no data available for critical sections.

[0070] If the deviation exceeds the limit and persists, it indicates that the equipment is likely to have a structural failure, such as a stuck float in the water level gauge or a clogged funnel in the rain gauge. If only a single deviation is used as the basis for confirmation, it will mask the essential difference between interference and failure, leading to false negatives or false negatives.

[0071] Therefore, if the deviation comparison result of any sensor to be verified continues to exceed a preset time, the sensor to be verified is identified as an abnormal sensor in order to distinguish between transient interference and real faults.

[0072] In this application, judging by duration can improve the accuracy of anomaly confirmation and avoid data source loss due to transient noise.

[0073] S22, when updating the hydrological status of the digital twin, shields the real-time data of all confirmed abnormal sensors.

[0074] In some embodiments, the identifiers of confirmed anomalous sensors are written into a data filtering blacklist, which is loaded in real time by the twin state update module. Furthermore, at the beginning of each state update cycle, the system iterates through the identifiers of all online sensors; if any identifier matches one of the blacklist identifiers, the data reading and processing flow for that sensor is skipped. The anomalous sensors mentioned here include water level gauges, rain gauges, or flow meters that have been confirmed to be faulty due to hardware damage, communication interruption, or data drift.

[0075] Furthermore, each status update cycle is treated as an execution unit, and the blacklist is dynamically refreshed and applied within each cycle.

[0076] In addition, to ensure the timeliness of the blocking operation, the blacklist update delay will not exceed one sampling interval.

[0077] It should be noted that relevant technical personnel can flexibly adjust the refresh frequency and matching strategy of the blacklist according to the specific watershed size and sensor density, and this application does not impose any restrictions.

[0078] It should be noted that although the above description provides a specific shielding mechanism, in practical applications, the specific implementation of the shielding method can be adjusted according to the site conditions and system architecture, such as software filtering or hardware disconnection.

[0079] For example, in some edge computing deployment scenarios, abnormal sensor data can be discarded directly at the data acquisition end; while in a centralized architecture, filtering is performed at the server end. Similarly, the stringency of the shielding strategy should be appropriately configured according to the actual situation for flood protection levels with different reliability requirements.

[0080] In this application, shielding abnormal sensor data can prevent erroneous data from contaminating the twin's state and ensure the purity of the simulation input.

[0081] S23, based on real-time data from all non-abnormal sensors, fuses and generates state correction data.

[0082] It is worth noting that in practice, different non-anomaly sensors provide varying degrees of coverage of the watershed state. Furthermore, the quality of state correction data is closely related to the spatial distribution density of non-anomaly sensors. If non-anomaly sensors are sparsely distributed, it means that some areas lack direct observation, leading to blind spots in state correction and creating uncertainty in the boundary condition settings for flood evolution simulation. Therefore, data fusion based on spatial representativeness is necessary.

[0083] Optionally, state-corrected data can be generated by fusing real-time data from all non-abnormal sensors using a combination of spatial interpolation and physical constraints. That is, original values ​​are retained for sections with direct observations, while calculated values ​​based on river topography and hydraulic continuity are used for sections without observations, thereby constructing a consistent state field across the entire basin. For example, in a scenario where upstream station A and downstream station C have normal data, but intermediate station B is abnormal, the water level at station B can be inferred using the Manning formula combined with the water levels at stations A and C and the river roughness.

[0084] If non-abnormal sensors are completely missing in a certain sub-region, it is necessary to fill in the pattern by referring to historical hydrological patterns of the same period or similar events in neighboring watersheds. For example, when all sensors fail due to a landslide in a small mountain watershed, historical rainstorm response curves from watersheds of the same climate zone and area can be used as a temporary state reference.

[0085] After completing the above fusion operations, the digital twin can be driven to complete state synchronization based on the generated state correction data.

[0086] The aforementioned state correction data refers to the reconstructed dataset used in this application to replace the original sensor input and cover the hydrological elements of the entire watershed. Its generation process integrates measured data and physical model deduction results.

[0087] In this application, the fused generated state correction data can still generate a physically consistent and spatially complete watershed state field even when some data is missing, supporting subsequent simulations.

[0088] S3, in response to the interruption of the data source of any main control monitoring section in the digital twin, calls the real-time data of the upstream section, the real-time data of the downstream section, and the inherent physical properties of the river channel of the main control monitoring section, deduces and generates compensatory state data of the main control monitoring section, and injects the compensatory state data into the digital twin.

[0089] It should be noted that if only historical averages or linear interpolation are used to fill in the interrupted data, there may be a risk of distortion of the hydrological process. That is, the actual water level at the interrupted section may deviate significantly from the simple interpolation result due to local confluence or channel contraction.

[0090] If the interpolated data is mechanically injected into the twin at this time, it may lead to errors in the calculation of flood wave propagation speed, which in turn may cause misjudgment of the timing of reservoir flood discharge.

[0091] For example, if a section is located downstream of the confluence of a tributary, and the water level of the main river is stable but a flash flood occurs in the tributary, the water level of this section will rise sharply, and linear interpolation cannot capture this sudden change.

[0092] The water level in a certain canyon section is rising due to the narrowing of the cross-section. If the influence of the river topography is ignored, the compensation water level will be seriously low.

[0093] In such cases, if the system continues to use a compensation strategy without physical constraints, it will interfere with the accuracy of flood evolution simulation and cause significant deviations in scheduling decisions.

[0094] Therefore, after detecting the interruption of the data source of the main control monitoring section, this application further introduces real-time hydrological data from upstream and downstream and the inherent physical properties of the river channel, generates compensatory state data through hydraulic model deduction, and verifies it based on downstream feedback, ultimately forming a physically consistent compensation result.

[0095] The aforementioned compensatory state data can be understood as reliable hydrological state values ​​generated after hydraulic constraints and boundary verification, used to replace missing observations.

[0096] In this application, even under data interruption conditions, step S3 can still generate compensation data that conforms to hydrodynamic laws, thus maintaining the physical consistency of the twins.

[0097] In some embodiments, the compensatory state data of the main control monitoring section is simulated, including: S31, retrieve the stored river cross-section topographic data and riverbed roughness data associated with the main control monitoring section.

[0098] It is easy to understand that users can obtain geometric features such as river width, water depth, and slope at the main monitoring section by retrieving the river cross-section topographic data associated with the main monitoring section. These geometric features determine the ability of water to flow through the main monitoring section. If these features are ignored, the water level cannot be accurately predicted.

[0099] It should also be understood that retrieving riverbed roughness data essentially involves obtaining parameters regarding the impact of riverbed roughness on flow resistance. Riverbed roughness data is usually stored in the form of Manning coefficients; the larger the value, the rougher the riverbed, and the higher the water level at the same flow rate.

[0100] In this application, the river cross-sectional topographic data is a cross-sectional elevation sequence obtained by lidar or sonar mapping, and the riverbed roughness data is a Manning coefficient table determined based on historical hydrological inversion or on-site investigation.

[0101] In this application, the river channel properties can be retrieved to provide the necessary physical boundary conditions for hydraulic simulation.

[0102] S32, based on the real-time flow and water level of the upstream section, combined with the river section topography data and riverbed roughness data, calculates the expected water level of the main control monitoring section through hydraulic continuity logic.

[0103] It should be noted that, in order to effectively address the potential problem of missing data at key cross-sections, the key lies in clarifying how to reconstruct the internal state using limited boundary information. This process should begin with establishing a physically interpretable inference model. Once the inference model is accurately constructed, specific water level estimates for the missing cross-sections can be carried out based on it. This will not only help to reasonably restore the twin state but also reduce inefficient and physically unfounded blind interpolation.

[0104] In some embodiments, water level estimation can be achieved using a simplified form of the one-dimensional unsteady flow Saint-Venant equations. For example, the Muskingum-Cunge method under the dynamic wave approximation can be used to explicitly calculate the propagation process through the flow-water-roughness-slope relationship. This approach is suitable for handling river sections with clearly defined upstream and downstream boundaries, balancing computational efficiency and physical fidelity.

[0105] In practice, the real-time flow rate at the upstream section can be used as the input boundary condition. For example, the measured flow rate at upstream station A can be used. As the model entry point.

[0106] Further, a discretized calculation unit is constructed for this river section. For example, continuing the aforementioned scenario, when processing the river section from station A to the target section B, the storage weight coefficient K and the tracer coefficient X of the calculation unit are based on the river length L, average slope S0, cross-sectional area A, water surface width B, and Manning coefficient n.

[0107] Furthermore, after initializing the model parameters, based on Muskingen's formula... Among them, the coefficient Co, , The flow rate at target section B is iteratively calculated based on K, X, and the time step Δt. This section utilizes existing technologies, and further details will not be elaborated upon.

[0108] Furthermore, using the river cross-sectional topographic data of section B, a table of water level-area-discharge relationship is established. This relationship can be obtained through table lookup or interpolation. Corresponding expected water level .

[0109] Furthermore, the expected water level of the main control monitoring section can be determined through the above-mentioned hydraulic deduction chain.

[0110] It should be noted that the above-mentioned expected water level can be understood as a preliminary compensation value derived in this application based on the physical model and upstream boundary conditions, without considering downstream feedback.

[0111] In this application, the compensation process is based on hydraulic principles, thus avoiding the physical distortion of purely statistical methods.

[0112] S33, based on the real-time water level of the downstream section, reversely verify the rationality of the expected water level; In practice, the real-time water level at the downstream section can be used as a reverse boundary constraint. For example, the measured water level Hc at downstream station C can be used as a reference for the upper limit of reasonableness.

[0113] Furthermore, determining whether the expected water level satisfies the non-negative hydraulic gradient condition requires ensuring that the water level decreases monotonically (or remains relatively flat) from upstream point A to target point B and then downstream point C. The constructed judgment rule can quickly filter out obviously unreasonable projection results. For example, when dealing with plain river networks, if the projected water level at point B is higher than that at upstream point A, it is judged as an anomaly.

[0114] Furthermore, after completing the initial screening for water level monotonicity, a hydraulic gradient tolerance check is performed on these expected water levels that pass the initial screening. For each expected water level that passes the initial screening, the average hydraulic gradient from A to B and from B to C is calculated. For example, in a gently sloping river channel, if the gradient from A to B is 0.0002 and the gradient from B to C is 0.0015, then the estimated value at point B may be too high. By comparing whether the two gradients are within a preset ratio range (e.g., not exceeding 3 times), if the ratio of the two gradients exceeds the tolerance, it can be determined that this expected water level is unreasonable.

[0115] Furthermore, after dual verification of monotonicity and slope tolerance, the expected water level that is ultimately retained while satisfying the hydraulic continuity conditions of upstream and downstream constitutes a reasonable compensation candidate that passes the reverse verification.

[0116] The aforementioned reverse verification can be understood as a key mechanism used in this application to verify the physical consistency of upstream inference results using downstream observations.

[0117] In this application, performing reverse verification can prevent inference drift caused by upstream data errors or model simplification, thereby improving the reliability of compensation data.

[0118] S34 uses the expected water level obtained through reverse verification as compensating state data.

[0119] Understandably, in order to facilitate the identification and loading of compensation data by the subsequent digital twin state update module, it is necessary to attach a compensation identifier to the expected water level that has passed the reverse verification. Then, the expected water level with the compensation identifier can be written into the twin state cache area, so as to facilitate the subsequent triggering of the state synchronization of the hydrodynamic model based on the compensation identifier.

[0120] The compensation identifier mentioned in S34 is a metadata tag used to distinguish between the original observation data and the inferred compensation data. That is, the "data source = compensation" field is appended to the attribute of the water level value. The expected water level in S34 is the water level value calculated by S32 and verified by S33.

[0121] It is easy to understand that through the above implementation method, the identity information of the compensation data can be clearly defined, enabling the internal modules of the twin to distinguish between the original data and the compensation data. Furthermore, by associating with the verification results, a traceable chain of the credibility of the compensation data is established, providing a basis for adjusting the simulation weights based on the data source in subsequent steps. This helps to more accurately integrate the compensation data into the overall state field and ensures sufficient traceability.

[0122] The aforementioned compensation identifier can be understood as key metadata used in this application to support data lineage tracing.

[0123] In this application, using the verified water level as compensation data ensures that the data injected into the twin is both physically reasonable and of clear origin.

[0124] S35, the inherent physical properties of the river channel include at least the topographic data of the river channel cross section and the roughness data of the riverbed.

[0125] Among them, the inherent physical properties of the river channel are a set of static parameters that describe the geometric and resistance characteristics of the river channel itself.

[0126] It should be noted that the inherent physical properties of a river channel have a decisive impact on the water level-discharge relationship. However, due to significant differences in cross-sectional shape and bed material across different river sections, users cannot accurately construct hydraulic simulation models without clearly understanding the specific composition of the river channel's inherent physical properties. The river channel data used in the actual implementation of this application includes cross-sectional elevation point clouds obtained through UAV aerial surveys and roughness distribution data retrieved through ADCP mobile surveys.

[0127] Therefore, before extrapolating the compensation data, it is necessary to clearly define the inherent physical properties of the river channel, including at least the river channel cross-sectional topographic data and the riverbed roughness data. Among them, the river channel cross-sectional topographic data is an elevation-distance sequence that characterizes the geometry of the river channel cross-section. The river channel cross-sectional topographic data set in this application is to facilitate the subsequent calculation of hydraulic radius and flow area.

[0128] In this application, river channel attributes are explicitly defined as providing a standardized physical input basis for state compensation.

[0129] S4, in response to receiving an emergency dispatch instruction for a designated protection target, divides the whole-basin flood evolution simulation task based on digital twins into a high-resolution simulation sub-task for the core protection area and a low-resolution simulation sub-task for the non-core area.

[0130] It is understandable that the whole-basin flood evolution simulation task mentioned in S4 consists of multiple computational grids. Therefore, the spatial coverage of the core protection zone can be determined by a set of grids based on geographical scope. In addition, the computational granularity of the simulation can be precisely controlled by the grid resolution level. The aforementioned spatial coverage and computational granularity can be used to balance simulation accuracy and computational timeliness.

[0131] In S4 above, the grid set is the coordinate set of all computational units after the entire watershed is discretized. This grid set is used to reflect the spatial extent of the simulation domain, while the grid resolution level is the side length of each computational unit in the horizontal direction, which represents the level of detail of the simulation.

[0132] Through the above implementation methods, this application can perform regionalized hierarchical processing of the whole watershed simulation in order to prioritize the simulation accuracy of key areas, and also facilitate the reduction of simulation time in core areas under limited computing power, ultimately avoiding decision delays caused by global high resolution.

[0133] In some embodiments, the whole-basin flood evolution simulation task is divided into a high-resolution simulation sub-task for the core protected area and a low-resolution simulation sub-task for the non-core area, including: S41, parse the emergency dispatch command and extract the geographical range of the specified protection target.

[0134] In some embodiments, the protection targets in emergency dispatch instructions are encoded as standard geographic identifiers, including administrative division codes, reservoir names, or dike chainage intervals. Furthermore, a mapping table between geographic identifiers and latitude / longitude boundaries is pre-configured in the instruction parsing module. The designated protection targets mentioned here cover urban built-up areas, critical infrastructure (such as substations and railway bridges), and densely populated villages and towns.

[0135] Furthermore, each instruction reception event is treated as a parsing cycle, and semantic parsing of the instruction text or structured message is completed within each cycle.

[0136] In addition, to ensure the accuracy of geographic range extraction, multiple input formats are supported, including GeoJSON, WKT, or natural language descriptions.

[0137] It should be noted that relevant technical personnel can flexibly adjust the command parsing rules and the expression of geographical scope according to the specific command system interface standard, and this application does not impose any restrictions.

[0138] It should be noted that although the above description provides a specific analysis scheme, in practical applications, the specific way of expressing geographical scope, such as points, lines, and areas, can be adjusted according to the type of protection target.

[0139] For example, in some situations, it may be necessary to protect a section of riverside road, in which case the geographic extent would be a linear buffer zone; while when protecting an urban area, it would be a polygonal area. Similarly, the precision requirements for the geographic extent should be appropriately configured according to the actual situation for different levels of emergency response.

[0140] The geographical range extracted in this application can provide a clear spatial basis for subsequent simulation partitioning.

[0141] S42 designates water conservancy elements within the geographical area as core protection zones, and areas outside the geographical area as non-core zones.

[0142] It is easy to understand that users can identify reservoirs, rivers, dikes and settlements within a specified protection target by overlaying the geographical range of the protection target onto the water feature layer of the digital twin. At this time, these features constitute the physical entities of the core protection area. If the feature attribution judgment is ignored, the high-resolution area boundary cannot be accurately defined.

[0143] It should also be understood that designating non-core areas essentially means automatically classifying the remaining area of ​​the entire watershed after subtracting the core protection zone as a low-priority simulation area. The designation of non-core areas ensures that computing resources are not excessively consumed by non-critical regions.

[0144] In this application, the water conservancy elements are vector primitives representing water engineering and socio-economic objects in a digital twin, and the core protection zone is a set of connected regions containing at least one key protected object.

[0145] In this application, the delineation of core / non-core areas can achieve spatial alignment of simulation resources and protection priorities.

[0146] S43 sets a first grid resolution for the core protected area and a second grid resolution for the non-core area, with the first grid resolution being higher than the second grid resolution.

[0147] It is worth noting that different simulation areas have different requirements for result accuracy in actual operation. Since the accuracy requirement is directly related to the grid resolution, if a coarse grid is used in the core protection area, it means that the edge of the flood inundation range will be blurred, which will lead to misjudgment of the personnel evacuation range and negatively affect the effectiveness of emergency response. Therefore, it is necessary to configure differentiated resolutions according to the importance of the area.

[0148] Optionally, a first grid resolution with a side length not exceeding 50 meters can be configured for the core protected area, and a second grid resolution with a side length not less than 200 meters can be configured for the non-core area. That is, 1 / 5 of the scale of the smallest critical facility in the core area is taken as the upper limit of the first grid resolution, and 1 / 2 of the scale of the largest runoff unit in the non-core area is taken as the lower limit of the second grid resolution, and two levels of grids are generated accordingly. For example, in a scenario where a bridge spanning a river is being protected with a pier spacing of 30 meters, the grid in the core area is set to 20 meters; the grid in the non-core area, which is farmland with a large runoff area, is set to 300 meters.

[0149] If computing resources are extremely limited, the resolution of non-core areas can be further compressed, but the hydrological response time error must be ensured to be within a preset threshold. For example, when the total number of computing nodes is less than 10, the grid in the non-core area can be widened to 500 meters, but it must be verified that its impact on the boundary conditions of the downstream core area is within 5%.

[0150] After completing the resolution configuration steps described above, subsequent simulation tasks can be built based on the generated two-level mesh.

[0151] The first grid resolution mentioned above refers to the size of the computing unit used for high-precision flood calculation in the core protected area in this application, and the second grid resolution refers to the size of the computing unit used for generalized simulation in the non-core area.

[0152] In this application, configuring differentiated grid resolution can maximize the simulation accuracy of key areas while keeping the total computational load under control.

[0153] S44, based on the first grid resolution and the second grid resolution, respectively constructs high-resolution simulation subtasks and low-resolution simulation subtasks.

[0154] It should be pointed out that the problem of low computational efficiency caused by a unified grid across the entire watershed in existing technologies can be solved by constructing multi-resolution simulation tasks based on regional importance. Only after completing the two-level grid division can refined hydrodynamic calculations be initiated for the core protected area, thereby enabling the output of a high-confidence risk situation within 10 minutes and avoiding missing the scheduling window due to the excessive time required for global high-resolution simulation.

[0155] In some implementations, multi-resolution task construction can be achieved using nested meshes or unstructured mesh stitching. For example, local mesh refinement techniques supported by finite volume solvers can be employed to refine the mesh in the core region while maintaining a coarse mesh in the non-core region, and data coupling can be achieved at the boundaries through flux conservation interpolation. This nested mesh method ensures hydrological continuity between different resolution regions, facilitating subsequent joint simulation execution.

[0156] In practice, the entire basin's basic grid (corresponding to the second grid resolution) can be initialized first. For example, in a tributary of the Yangtze River, a 200-meter basic grid can be constructed to cover the entire basin.

[0157] Furthermore, for the core protected area, a 50-meter-high density grid is overlaid on the basic grid, and a mapping table between coarse and fine grids is established. For example, in the office building area, each coarse grid is subdivided into 16 fine grids. Simultaneously, a task dependency graph is generated, indicating that tasks in the fine grids must wait for the boundary conditions of the coarse grids to be ready.

[0158] Furthermore, allocate independent computing thread groups to high-resolution subtasks and allocate the remaining resources to low-resolution subtasks; for example, allocate 80% of the GPU memory to the core area for simulation.

[0159] Furthermore, after all subtasks are constructed, the high / low resolution simulation subtasks are registered to the task scheduling queue. The simulation subtasks thus determined are independent computing units with differentiated accuracy and resource binding.

[0160] It should be noted that the aforementioned high-resolution simulation subtask can be understood as a key computational module used in this application to quickly generate flood risk maps around the core protection targets.

[0161] In this application, the construction of multi-resolution subtasks can enable computing resources to be tilted towards high-value areas, supporting minute-level emergency decision-making.

[0162] S5 prioritizes scheduling computing resources to execute high-resolution simulation sub-tasks in the core protected area, and generates scheduling instructions to regulate the flood discharge of the upstream reservoir group based on the flood risk situation output by the high-resolution simulation sub-tasks.

[0163] It should be noted that if flood discharge orders are formulated solely based on low-precision simulation results of the entire basin, there may be risks of delayed scheduling or excessive flood discharge. That is, the actual peak flood level in the core area may be underestimated, leading to untimely evacuation of personnel.

[0164] If the conventional scheduling rules are still followed mechanically at this time, it may lead to further flooding downstream or waste of water released from the reservoir.

[0165] For example, the flood control standard of a city's levee is designed for a 50-year flood event, but due to localized heavy rainfall, the actual flood is close to a 100-year flood event. Low-resolution models cannot capture this super-standard process.

[0166] The independent operation of multiple upstream reservoirs, without considering their joint storage capacity, led to the superposition of downstream flood peaks.

[0167] In such situations, if the system continues to use a simulation and scheduling strategy that does not differentiate between priorities, it will interfere with the effectiveness of emergency response and cause significant public safety risks.

[0168] Therefore, after completing the high-resolution simulation of the core area, this application further introduces the coupling analysis of flood risk situation and reservoir group status to generate coordinated flood discharge instructions, realizing a closed loop from accurate forecasting to fine scheduling.

[0169] The aforementioned scheduling instructions can be understood as structured control commands used in this application to coordinate the discharge flow and timing of multiple reservoirs.

[0170] In this application, executing S5 can directly transform high-precision simulation results into executable joint scheduling actions, thus shortening the decision-making process.

[0171] In some embodiments, computing resources are prioritized to execute high-resolution simulation subtasks within the core protected area and to generate scheduling instructions for regulating the flood discharge of the upstream reservoir group, including: S51, allocate no less than a preset proportion of total computing resources to the high-resolution simulation sub-tasks in the core protected area.

[0172] In some embodiments, total computing resources are quantified as the number of CPU cores, GPU memory capacity, or the total number of parallel task slots. Furthermore, the resource scheduler pre-sets a core area resource guarantee ratio of no less than 70%. The total computing resources mentioned here include all available computing units deployed on cloud platforms or edge servers.

[0173] Furthermore, each simulation task start event is treated as a resource allocation cycle, and the computing resources are dynamically allocated within each cycle.

[0174] In addition, to ensure the real-time performance of the core area simulation, resource allocation operations must be completed before task queue scheduling.

[0175] It should be noted that relevant technical personnel can flexibly adjust the preset ratio value according to the specific hardware configuration and emergency level, and this application does not impose any restrictions.

[0176] It should be noted that although the above description provides a specific resource guarantee mechanism, in practical applications, the specific resource allocation strategy can be adjusted according to the system load, such as preemption or reservation.

[0177] For example, in some lightly loaded scenarios, non-core tasks can be completely paused to free up resources; while in heavily loaded scenarios, time-slice rotation is used to ensure minimum computing power in the core area. Similarly, for watersheds of different sizes, the resource ratio should be appropriately configured according to the actual situation.

[0178] In this application, prioritizing the allocation of computing resources ensures that the core area simulation can be completed within the time window.

[0179] S52 is a high-resolution simulation subtask that monitors the flood risk situation in the core protected area in real time.

[0180] It is easy to understand that users can assess the flood risk level of key facilities by reading the water depth, flow velocity, and inundation duration output by the core area simulation in real time. At this time, these indicators constitute a quantitative representation of the flood risk situation. If only the final result snapshot is relied upon, the dynamic evolution process of the risk cannot be captured.

[0181] It should also be understood that monitoring flood risk status essentially involves identifying points of sudden risk abrupt change through time series analysis. The flood risk status is the direct input driving subsequent flood discharge decisions; the higher the value, the more urgent the threat to the protected targets.

[0182] In this application, the flood risk situation is a risk index based on a weighted composite of multiple indicators, with the weights pre-configured according to the type of protected object.

[0183] In this application, real-time monitoring of risk status can support dynamic triggering of scheduling actions, thus avoiding response delays.

[0184] S53, in response to the flood risk situation reaching the warning level, extracts the current water storage status of the upstream reservoir group.

[0185] In practice, the water storage status of a reservoir group during the flood season directly affects its flood control capacity, which is closely related to the current water level, reservoir capacity, and availability of discharge facilities.

[0186] When the water level of a reservoir approaches the flood control limit, its potential to participate in joint flood regulation is relatively small, and it may even need to release water in advance.

[0187] If a reservoir is at a low water level, it has significant potential for peak shaving. However, if only a fixed flood discharge rule is adopted without considering the current water storage status, the individual differences among reservoirs will be ignored, leading to an infeasible scheduling plan.

[0188] Therefore, in response to the flood risk situation reaching the warning level, the current water storage status of the upstream reservoir group is extracted to support the generation of personalized flood discharge strategies.

[0189] The extraction of water storage status in this application can ensure that the dispatching instructions match the actual capacity of the reservoir, thereby improving the feasibility of the plan.

[0190] S54 generates a joint flood discharge command for the reservoir group based on the flood risk situation and the current water storage status of the upstream reservoir group to reduce the flood risk in the core protected area.

[0191] It should be noted that, to effectively address the potential challenges of coordinated operation of multiple reservoirs, the key lies in clarifying how to meet downstream flood control objectives while also considering the safety constraints of each reservoir. This process should begin with the construction of a multi-objective optimization model. Once the optimization model is accurately established, specific instructions for the current risk scenario can be generated accordingly. This not only helps to rationally allocate the flood discharge burden, but also reduces inefficient and uncoordinated independent scheduling operations.

[0192] In some embodiments, linear programming or model predictive control (MPC) algorithms can be used to generate joint flood discharge commands. For example, an optimization model can be used with the objective function of minimizing the peak flood flow in the core area, and with constraints such as the upper and lower limits of reservoir water levels, the capacity of flood discharge facilities, and the safe flow of downstream rivers. This approach is suitable for handling the coordinated problems of multi-input, multi-output reservoir groups and can systematically balance flood control and water storage objectives.

[0193] In practice, the flood risk situation can be transformed into a target flood discharge process line at the downstream control section first. For example, the maximum allowable flow rate at the core area inlet section can be set as Qmax(t).

[0194] Further, we construct a set of scheduling variables for the upstream reservoir group. For example, continuing the aforementioned scenario, we define the future T-hour discharge sequence of N reservoirs as follows: , , ..., , as an optimization decision variable.

[0195] Furthermore, after defining the variables, based on the water balance equation... ,in For interval inflow, For changes in reservoir storage, For the error term, a mapping relationship is established between the outflow from the reservoir group and the downstream flow. This error term can be designed according to actual needs and can be a linear function of time.

[0196] Furthermore, the mapping relationship between the outflow of the above-mentioned reservoir group and the downstream flow is embedded into the optimization model to solve for the combination of discharge sequences that satisfy all constraints and minimize the flood peak in the core area.

[0197] Furthermore, the optimization results are encapsulated into structured scheduling instructions, including the start time of flood discharge, target flow, duration, and gate opening recommendations for each reservoir, which are then sent to the control centers of each reservoir through the scheduling interface.

[0198] It should be noted that the aforementioned joint flood discharge command of the reservoir group can be understood as an automated decision-making process used in this application to coordinate the disaster reduction efforts of multiple projects.

[0199] In this application, the generation of joint flood discharge instructions can realize a closed loop from risk perception to multi-project coordinated response, maximizing the overall flood control benefits.

[0200] In some embodiments, the method further includes: S61, in response to the completion of the high-resolution simulation subtask in the core protected area, initiates the low-resolution simulation subtask in the non-core area.

[0201] In practice, once the high-resolution simulation of the core protected area is completed, it means that the risk assessment of the key area has been preliminarily determined. At this time, starting the low-resolution simulation of the non-core area can effectively utilize computing resources while ensuring the overall risk assessment of the entire watershed.

[0202] In practice, the scheduling system can automatically detect the simulation status of the core protection zone and trigger the start command for the simulation of non-core areas. This step aims to ensure that even with limited resources, accurate flood risk forecasting within the core protection zone can be prioritized.

[0203] For example, in a flood disaster early warning system for a large river basin, an event listener is set up to monitor the simulation progress of the core protected area. Once completed, the remaining computing resources are immediately allocated to the simulation work of the non-core area.

[0204] In this application, the simulation of non-core areas can achieve full coverage of the entire watershed without sacrificing the accuracy of core protected areas.

[0205] S62 couples the output of the low-resolution simulation subtask in the non-core area with the boundary output of the high-resolution simulation subtask in the core protection area to generate a comprehensive flood evolution map of the entire basin.

[0206] To achieve this goal, the first step is to process the low-resolution simulation results of the non-core area and extract its boundary condition information. Then, a specialized data fusion algorithm is used to match and integrate this information with the boundary data from the high-resolution simulation of the core area.

[0207] In some embodiments, Geographic Information System (GIS) technology is used as a tool to facilitate seamless integration between models of different resolutions. GIS not only provides powerful spatial analysis capabilities but also supports the visualization of multi-source data, enabling decision-makers to intuitively understand the flood evolution trends across the entire basin.

[0208] Furthermore, by constructing a unified spatial reference framework, all simulation results can be accurately superimposed in the same coordinate system, thereby generating a comprehensive flood evolution map of the entire basin. This not only helps improve prediction accuracy but also provides a scientific basis for formulating effective flood control and disaster reduction strategies.

[0209] In this application, coupling simulation results from two different resolutions can generate a more complete and accurate flood evolution prediction, supporting more scientific flood control decisions.

[0210] In some embodiments, the method further includes: S71, in response to the end of extreme hydrological events, stops cross-validation of data from all online hydrological sensors.

[0211] After an extreme hydrological event has ended, continuing high-intensity data cross-validation is both wasteful of resources and unnecessary. Therefore, it is crucial to design a mechanism to automatically identify the event termination point and adjust the system's operating mode accordingly.

[0212] Specific implementation methods may include determining whether key indicators such as water level and flow rate have returned to normal levels based on preset thresholds or machine learning models. Once the event is confirmed to be over, the system should immediately issue an instruction to stop unnecessary data verification processes.

[0213] For example, some intelligent monitoring platforms use time series analysis combined with expert experience rules to dynamically assess the current hydrological conditions in order to decide when to shut down additional data verification steps.

[0214] This approach not only saves on computing costs but also helps extend the lifespan of equipment and reduce maintenance frequency.

[0215] In this application, stopping data cross-validation can optimize resource allocation, reduce invalid operations, and improve the overall system efficiency.

[0216] S72 restores the digital twin to normal operating mode and records the identity information of all abnormal sensors and the generation log of compensatory status data during this emergency response.

[0217] Restoring the digital twin to normal operating mode involves several steps. First, the operating parameters of each component are reset to the settings required for daily monitoring. Second, based on previously recorded anomalies, problematic sensor nodes are identified and repaired one by one.

[0218] During this process, it is crucial to meticulously record the time and location of each failure, along with the specific corrective measures taken. This is invaluable for subsequent maintenance. Furthermore, special attention should be paid to compensatory status data provided in emergency situations. These data are often based on model predictions rather than direct measurements and must be clearly labeled for future reference.

[0219] For example, a water conservancy project team developed an automated management system that can automatically generate log files containing all the above information and archive them for long-term tracking and analysis.

[0220] This not only helps improve the reliability and stability of the system, but also provides valuable lessons for similar events that may occur in the future.

[0221] In this application, restoring the normal operating mode and recording exception logs can enhance system robustness, accumulate an operational knowledge base, and promote continuous improvement.

[0222] In some practical implementations, operators have encountered situations where the target watershed is under a red alert for severe rainstorms, a key hydrological monitoring section located in a canyon experiences a complete communication link disruption due to a landslide, and two nearby rain gauges generate persistently high readings due to strong winds causing foliage to obstruct the view. Under these combined anomaly conditions, the digital twin faces the dual risks of partially missing and contaminated input data.

[0223] If the system fails to identify abnormal rain gauges and uses their data for status updates, it will result in an inflated rainfall input field and an earlier-than-expected flood peak. If linear interpolation compensation is used for the interrupted section, the contraction effect of the canyon section is ignored, the actual water level is underestimated, and thus the reservoir discharge is insufficient.

[0224] Therefore, for this scenario, technicians take the following steps to minimize risk: Step 1.1: In response to receiving a red rainstorm warning signal, activate the digital twin corresponding to the target watershed and initiate cross-validation of data from all online hydrological sensors; delineate a linear validation cluster for the water level gauge with an upstream-downstream spacing of no more than 8 kilometers along the river channel, and construct a Thiessen polygon validation cluster for the rain gauge based on micro-topographic units; The spatial scale of the verification cluster mentioned in this section can be set according to the actual needs of relevant technical personnel and in combination with the hydrological response characteristics of the watershed; this application does not impose any restrictions on it. Step 1.2: Calculate the average real-time data of similar sensors within each validation cluster, and compare the relative deviation between the real-time data of each sensor and the average of its cluster. If the relative deviation does not exceed the preset tolerance, return to the normal data fusion process and continue to use the sensor data to update the twin; if the relative deviation exceeds 15%, proceed to step 1.3. The relative deviation tolerance mentioned in this section can be set according to the actual needs of relevant technical personnel, combined with the sensor type and installation environment; this application does not impose any limitations on it. Step 1.3: Mark the sensor with excessive deviation as a sensor to be verified and start continuous monitoring; if the deviation continues to exceed the limit in the next 3 consecutive sampling periods, each period is 5 minutes, then it is confirmed as an abnormal sensor and proceed to step 1.4. If the deviation recovers to within the tolerance within any period, the verification mark is canceled and the process returns to the normal data fusion workflow; Step 1.4: Write the confirmed abnormal sensor identity information into the data filtering blacklist, block their data contribution when the digital twin status is updated, and generate status correction data based on the fusion of the remaining non-abnormal sensor data; Step 1.5: Detect whether there is an interruption in the data source of the main control monitoring section in the digital twin; If there is no interruption, the compensation process is skipped and the simulation scheduling phase is entered directly. If there is an interruption, proceed to step 1.6; Step 1.6: Retrieve the topographic data and riverbed roughness data of the river section associated with the interrupted section, and calculate the expected water level of the section based on the measured flow and water level of the upstream section using the Muskingen-Kangi method. Step 1.7: Based on the measured water level at the downstream section, verify whether the expected water level calculated in Step 1.6 satisfies the condition that the hydraulic gradient is monotonically decreasing and the ratio of the upstream and downstream slopes does not exceed 3 times. If the verification passes, proceed to step 1.8; If the verification fails, return to step 1.6, adjust the Manning coefficient or introduce an interval inflow correction term to recalculate; The slope ratio threshold mentioned in this section can be set according to the actual needs of relevant technical personnel and in combination with the stability of the river longitudinal profile; this application does not impose any limitation on it. Step 1.8: Mark the verified expected water level as compensatory state data, attach the "data source=compensation" metadata tag, and then inject it into the digital twin; Step 1.9: Check if an emergency dispatch instruction for the specified protection target has been received; if not, maintain a uniform low-resolution simulation across the entire watershed. If received, proceed to step 1.10; Step 1.10: Analyze the geographical range of the protection target specified in the instruction, delineate the core protection area and non-core area, configure a 50-meter grid resolution for the core protection area and a 300-meter grid resolution for the non-core area, and construct a high / low resolution nested simulation subtask; Step 1.11: Allocate no less than 70% of the total computing resources to prioritize the execution of high-resolution simulation subtasks in the core protected area; Step 1.12: Monitor the flood risk situation output by the simulation in the core area in real time; If the risk situation does not reach the warning level, continue the simulation until it is completed; if it reaches the warning level, proceed to step 1.13. Step 1.13: Extract the current water storage status of the upstream reservoir group, solve the multi-objective optimization model based on the flood risk situation and reservoir status, generate and issue joint flood discharge instructions for the reservoir group.

[0225] Therefore, by introducing the above-mentioned operating procedures, this application can dynamically select the data processing path in extreme scenarios where sensor anomalies and data interruptions coexist, avoid invalid calculations, and ensure high-fidelity state reconstruction and accurate scheduling decisions are completed within a limited emergency window.

[0226] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0227] Based on the same inventive concept, this application also provides a digital twin-based water conservancy digital management system. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the digital twin-based water conservancy digital management system provided below can be found in the limitations of the digital twin-based water conservancy digital management method described above, and will not be repeated here.

[0228] In one exemplary embodiment, a digital twin-based water conservancy digital management system is provided, comprising: The twin activation module is configured to activate the digital twin corresponding to the target watershed in response to receiving an early warning signal characterizing an extreme hydrological event, and to initiate cross-validation of data from all online hydrological sensors within the target watershed. The state correction module is configured to identify anomalous sensors whose data deviation exceeds the tolerance based on the data cross-validation results, and generate state correction data that excludes the contribution of anomalous sensor data to update the digital twin. The state compensation module is configured to respond to an interruption in the data source of any master monitoring section in the digital twin, call the real-time data of the upstream section, the real-time data of the downstream section, and the inherent physical properties of the river channel of the master monitoring section, deduce and generate compensatory state data of the master monitoring section, and inject the compensatory state data into the digital twin. The simulation partitioning module is configured to, in response to receiving an emergency dispatch command for a specified protection target, divide the whole-basin flood evolution simulation task based on digital twins into a high-resolution simulation subtask for the core protection area and a low-resolution simulation subtask for the non-core area. The instruction generation module is configured to prioritize scheduling computing resources to execute high-resolution simulation subtasks in the core protected area, and generate scheduling instructions to regulate the discharge of upstream reservoirs based on the flood risk situation it outputs.

[0229] The various modules in the aforementioned digital twin-based water conservancy digital management system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0230] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 3As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a digital twin-based digital water conservancy management method. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0231] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0232] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0233] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0234] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0235] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0236] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0237] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0238] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A digital management method for water conservancy based on digital twins, characterized in that, include: In response to receiving an early warning signal characterizing an extreme hydrological event, the digital twin corresponding to the target watershed is activated, and cross-validation of data from all online hydrological sensors within the target watershed is initiated. Based on the results of the cross-validation of the data, abnormal sensors whose data deviation exceeds the tolerance are identified, and state correction data excluding the contribution of the abnormal sensor data is generated to update the digital twin. In response to an interruption in the data source of any main control monitoring section in the digital twin, the system calls up the real-time data of the upstream section, the real-time data of the downstream section, and the inherent physical properties of the river channel to deduce and generate compensatory state data of the main control monitoring section, and injects the compensatory state data into the digital twin. In response to receiving an emergency dispatch instruction for a designated protection target, the whole-basin flood evolution simulation task based on the digital twin is divided into a high-resolution simulation subtask for the core protection area and a low-resolution simulation subtask for the non-core area. Prioritize scheduling computing resources to execute high-resolution simulation subtasks in the core protected area, and generate scheduling instructions to regulate the flood discharge of the upstream reservoir group based on the flood risk situation output by the high-resolution simulation subtasks.

2. The water conservancy digital management method based on digital twin as described in claim 1, characterized in that, The initiation of cross-validation of data from all online hydrological sensors within the target watershed includes: Define spatially adjacent validation clusters for each type of hydrological sensor; Calculate the average real-time data of similar hydrological sensors within each validation cluster; The real-time data of each hydrological sensor is compared with the average real-time data of its respective validation cluster to determine the deviation. Hydrological sensors whose deviation comparison results exceed the preset tolerance are marked as sensors to be verified.

3. The water conservancy digital management method based on digital twin as described in claim 2, characterized in that, The process of identifying anomalous sensors whose data deviation exceeds the tolerance and generating state correction data that excludes the contribution of data from the anomalous sensors includes: If the deviation comparison result of any sensor to be verified continues to exceed a preset time, the sensor to be verified is identified as an abnormal sensor. When updating the hydrological status of the digital twin, real-time data from all confirmed abnormal sensors are masked. The state correction data is generated by fusing real-time data from all non-abnormal sensors.

4. The water conservancy digital management method based on digital twin as described in claim 3, characterized in that, The deduction generates compensatory state data for the main control monitoring section, including: Retrieve the stored river cross-section topographic data and riverbed roughness data associated with the main control monitoring section; Based on the real-time flow and water level of the upstream section, combined with the topographic data and riverbed roughness data of the river section, the expected water level of the main control monitoring section is calculated through hydraulic continuity logic. The rationality of the expected water level is verified by reverse analysis based on the real-time water level at the downstream section. The expected water level obtained through reverse verification will be used as the compensating state data.

5. The water conservancy digital management method based on digital twin as described in claim 4, characterized in that, The task of simulating the evolution of the entire basin flood system is divided into a high-resolution simulation sub-task for the core protected area and a low-resolution simulation sub-task for the non-core area, including: Parse the emergency dispatch command and extract the geographical range of the specified protection target; Water conservancy elements within the geographical area are designated as core protection zones, and areas outside the geographical area are designated as non-core zones. A first grid resolution is configured for the core protected area, and a second grid resolution is configured for the non-core area, wherein the first grid resolution is higher than the second grid resolution; Based on the first grid resolution and the second grid resolution, the high-resolution simulation subtask and the low-resolution simulation subtask are constructed respectively.

6. The water conservancy digital management method based on digital twin as described in claim 5, characterized in that, The prioritized computing resources execute the high-resolution simulation subtask of the core protected area and generate scheduling instructions for regulating the flood discharge of the upstream reservoir group, including: Allocate no less than a preset proportion of total computing resources to the high-resolution simulation subtasks of the core protected area; Real-time monitoring of the flood risk situation output by the high-resolution simulation subtask of the core protected area; In response to the flood risk situation reaching the warning level, the current water storage status of the upstream reservoir group is extracted; Based on the flood risk situation and the current water storage status of the upstream reservoir group, a joint flood discharge command for the reservoir group is generated to reduce the flood risk of the core protected area.

7. The water conservancy digital management method based on digital twin as described in claim 6, characterized in that, The method further includes: In response to the completion of the high-resolution simulation subtask of the core protected area, the low-resolution simulation subtask of the non-core area is initiated. The output results of the low-resolution simulation subtask in the non-core area are coupled with the boundary output results of the high-resolution simulation subtask in the core protection area to generate a comprehensive flood evolution status map of the entire basin.

8. The water conservancy digital management method based on digital twin as described in claim 7, characterized in that, The method further includes: In response to the end of the extreme hydrological event, cross-validation of data from all online hydrological sensors is stopped. The digital twin is restored to its normal operating mode, and the identity information of all abnormal sensors and the generation log of compensatory status data during this emergency response are recorded.

9. The water conservancy digital management method based on digital twin as described in claim 8, characterized in that, The inherent physical properties of the river channel include at least the river channel cross-sectional topographic data and the riverbed roughness data.

10. A digital management system for water conservancy based on digital twins, using the method described in any one of claims 1 to 9, characterized in that, include: The twin activation module is configured to activate the digital twin corresponding to the target watershed in response to receiving an early warning signal characterizing an extreme hydrological event, and to initiate cross-validation of data from all online hydrological sensors within the target watershed. The state correction module is configured to identify abnormal sensors whose data deviation exceeds the tolerance based on the results of the data cross-validation, and generate state correction data that excludes the contribution of the abnormal sensor data to update the digital twin. The state compensation module is configured to respond to an interruption in the data source of any master monitoring section in the digital twin, call the real-time data of the upstream section, the real-time data of the downstream section, and the inherent physical properties of the river channel of the master monitoring section, deduce and generate compensatory state data of the master monitoring section, and inject the compensatory state data into the digital twin. The simulation partitioning module is configured to, in response to receiving an emergency dispatch command for a specified protection target, divide the whole-basin flood evolution simulation task based on the digital twin into a high-resolution simulation subtask for the core protection area and a low-resolution simulation subtask for the non-core area. The instruction generation module is configured to prioritize scheduling computing resources to execute high-resolution simulation subtasks of the core protected area, and generate scheduling instructions to regulate the flood discharge of the upstream reservoir group based on the flood risk situation output by the high-resolution simulation subtasks.

Images