Drainage basin water resource scheduling method and system based on digital twinning

Abstract

The invention relates to the technical field of digital twinning, in particular to a drainage basin water resource scheduling method and system based on digital twinning, and the method comprises the steps: obtaining the spatial topological relation of a plurality of monitoring stations contained in a preset drainage basin, and the real-time water level data of a target monitoring station; according to the historical water level data of the target monitoring station, calculating a corresponding first confidence coefficient; according to the spatial relationship between the target monitoring station and other stations in the spatial topological relationship, calculating a corresponding second confidence coefficient; performing fusion calculation according to the first confidence coefficient and the second confidence coefficient to obtain the target confidence coefficient of the target monitoring station; inputting the target confidence and the real-time water level data into a preset digital twinborn scheduling model, and outputting a water resource scheduling instruction; wherein the water resource scheduling instruction is used for performing water resource scheduling in a preset drainage basin. Therefore, the credibility of the monitoring data can be dynamically evaluated, and the data weight can be adaptively adjusted, so that the reliability and the safety of a water resource scheduling decision are improved.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, specifically to a watershed water resource scheduling method and system based on digital twins for smart water meters. Background Technology

[0002] In the field of watershed water resources management, digital twin technology is increasingly being applied to water resources allocation and decision-making processes. Currently, most watersheds have deployed automated water level monitoring stations, enabling real-time data acquisition and transmission through the Internet of Things (IoT). However, monitoring equipment is affected by environmental interference, equipment malfunctions, and other factors, often resulting in anomalies such as missing, drifting, or sudden changes in key water level data.

[0003] Existing scheduling methods generally use equal weighting or fixed weighting to process data input from different monitoring stations, failing to fully consider the spatial differences in station distribution, the dynamic characteristics of upstream and downstream hydraulic relationships, and the consistent patterns of water level time series changes. This approach leads to the unreasonable amplification of anomalous data in scheduling calculations, causing the results of digital twin models to deviate from actual hydrological processes and reducing the reliability and security of water resource scheduling decisions. Especially under complex hydrological conditions, anomalous fluctuations at isolated stations can easily be misjudged as real hydrological events, leading to erroneous scheduling instructions and posing potential risks to watershed flood control, drought relief, and ecological security.

[0004] Existing technologies lack a mechanism to dynamically assess the reliability of monitoring data, cannot adaptively adjust data weights based on historical water level change patterns and station spatial topology, and cannot guarantee the stability of the digital twin scheduling model in the event of data anomalies. Summary of the Invention

[0005] To address the lack of a mechanism for assessing the reliability of monitoring data, the inability to adaptively adjust data weights based on historical water level change patterns and station spatial topology, and the difficulty in ensuring the stability of digital twin scheduling models under data anomalies, this invention provides a watershed water resource scheduling method and system based on digital twins. This method can dynamically assess the reliability of monitoring data and adaptively adjust data weights, thereby improving the reliability and security of water resource scheduling decisions.

[0006] To achieve the above objectives, firstly, this application proposes a watershed water resources scheduling method based on digital twins, comprising: Acquire the spatial topology of multiple monitoring stations within a preset watershed, as well as the real-time water level data of the target monitoring station; wherein, the target monitoring station is any monitoring station within the preset watershed. Based on the historical water level data of the target monitoring station, the corresponding first confidence level is calculated; whereby the first confidence level is the confidence level determined by the target monitoring station based on the changing pattern of historical water level data; The second confidence level is calculated based on the spatial relationship between the target monitoring station and other stations in the spatial topology; wherein, the second confidence level is the confidence level determined based on the coordination between the target monitoring station and its neighboring stations in the spatial topology. The target confidence level of the target monitoring station is obtained by fusing the first confidence level and the second confidence level. The target confidence level and real-time water level data are input into a preset digital twin scheduling model, and water resource scheduling instructions are output; the water resource scheduling instructions are used to carry out water resource scheduling within a preset watershed.

[0007] In one embodiment, the step of calculating the corresponding first confidence level based on historical water level data of the target monitoring station includes: Acquire historical water level data of the target monitoring station and divide the historical water level data into multiple continuous local time windows; Calculate the water level change characteristic value corresponding to each local time window; whereby the water level change characteristic value is used to characterize the severity of water level fluctuation within the local time window; The local time window in which the water level change characteristic value exceeds the preset anomaly judgment threshold is marked as the first anomaly window; The authenticity of the first abnormal window is verified based on the historical water level data corresponding to the first abnormal window in order to determine the second abnormal window; wherein, the second abnormal window is the time window in the historical water level data in which the abnormality occurs. The cumulative duration of the second abnormal window appearing in multiple local time windows is statistically analyzed, and the ratio between the cumulative duration of the abnormal window and the total monitoring duration is calculated to determine the first confidence level; where the total monitoring duration is the total monitoring duration corresponding to the historical water level data.

[0008] In one embodiment, after the step of acquiring historical water level data of the target monitoring station, the method further includes: Historical water level data is decomposed to obtain trend components, seasonal components, and noise components. The trend component is used to characterize the long-term evolution direction of water level, the seasonal component is used to characterize the periodic fluctuation pattern of water level, and the noise component is used to characterize the random fluctuation of water level. Based on the trend component, the reasonable range of water level changes for the target monitoring station over a long period of time is determined.

[0009] In one embodiment, the step of verifying the authenticity of the first abnormal window based on the historical water level data corresponding to the first abnormal window to determine the second abnormal window includes: Extract the seasonal component segment and the noise component segment that are temporally aligned with the first anomaly window from the seasonal component and the noise component; The actual observed data sequence of the first anomaly window is subtracted point by point from the seasonal component segment and the noise component segment to obtain the current trend data sequence corresponding to the first anomaly window. Perform linear fitting on the current trend data sequence to determine the current trend characteristics corresponding to the first anomaly window; Compare the current trend characteristics with the reasonable range of change; If the current trend characteristics are not within a reasonable range of change, then the first abnormal window is determined as the second abnormal window; wherein, the second abnormal window is the abnormal window after verifying that the first abnormal window is determined to be a real abnormality.

[0010] In one embodiment, the step of calculating the corresponding second confidence level based on the spatial relationship between the target monitoring station and other stations in the spatial topology includes: Identify multiple upstream stations of the target monitoring station within a predefined watershed; Calculate the similarity of water level changes between each upstream station and the target monitoring station; Based on the river slope and river distance between each upstream station and the target monitoring station, the spatial location influence weight of each upstream station is calculated. By combining similarity and spatial location influence weights, the overall influence weight of each upstream station on the target monitoring station is determined; Based on the current water level change characteristic value of each upstream station and its corresponding comprehensive influence weight, calculate the expected water level change characteristic value of the target monitoring station; The second confidence level is calculated based on the actual water level change characteristic value and the expected water level change characteristic value of the target monitoring station.

[0011] In one embodiment, the step of calculating the spatial location influence weight of each upstream station based on the river slope and river distance between each upstream station and the target monitoring station includes: Obtain the river slope and river distance between the upstream station and the target monitoring station; where the upstream station is any station upstream of the target monitoring station. Based on the river slope, the slope influence factor is determined, and the slope influence factor increases with the increase of slope. Based on the river channel distance, a distance influence factor is determined, wherein the distance influence factor decreases as the distance increases; The spatial location influence weight of the target monitoring station is obtained by weighting and summing the slope influence factor and the distance influence factor.

[0012] In one embodiment, the step of calculating the expected water level change characteristic value of the target monitoring station based on the current water level change characteristic value of each upstream station and its corresponding comprehensive influence weight includes: Obtain the current water level change characteristic value of each upstream station during the current monitoring period; The current water level change characteristic value of each upstream station is multiplied by its corresponding comprehensive influence weight to obtain the weighted contribution value of each upstream station. The weighted contribution values ​​of all upstream stations are summed to obtain the expected water level change characteristic value of the target monitoring station.

[0013] In one embodiment, the step of calculating the second confidence level based on the actual water level change characteristic value and the expected water level change characteristic value of the target monitoring station includes: Determine whether the actual water level change characteristic value and the expected water level change characteristic value change in the same direction; If the directions of change are inconsistent, the second confidence level will be set to the preset lowest confidence level value; If the directions of change are consistent, the degree of difference between the actual water level change characteristic value and the expected water level change characteristic value is calculated, and the second confidence level is calculated based on the degree of difference. The greater the degree of difference, the lower the second confidence level.

[0014] In one embodiment, the step of inputting the target confidence level and real-time water level data into a preset digital twin scheduling model and outputting water resource scheduling instructions includes: Based on the target confidence level, real-time water level data is divided into core input data and auxiliary input data; among them, core input data is used to directly drive scheduling decisions, while auxiliary input data is used to correct parameters within the scheduling decision framework. Based on the core input data, the macro-operation interval of at least one reservoir within a preset watershed is determined; wherein, the macro-operation interval is determined according to the real-time water level and preset scheduling rules, and the macro-operation interval includes the flood discharge interval, the water storage interval and the water supply interval. Within a defined macro-operational range, the scheduling operation parameters are optimized and calculated in conjunction with auxiliary input data to generate water resource scheduling instructions; wherein, the scheduling operation parameters include at least one of flood discharge flow, water storage target water level, or water supply flow.

[0015] Secondly, this application proposes a watershed water resources scheduling system based on digital twins, including: The acquisition module is used to acquire the spatial topology of multiple monitoring stations within a preset watershed, as well as the real-time water level data of the target monitoring station; wherein, the target monitoring station is any monitoring station within the preset watershed. The first calculation module is used to calculate the corresponding first confidence level based on the historical water level data of the target monitoring station; wherein, the first confidence level is the confidence level determined by the target monitoring station based on the changing pattern of the historical water level data; The second calculation module is used to calculate the corresponding second confidence level based on the spatial relationship between the target monitoring station and other stations in the spatial topology; wherein, the second confidence level is a confidence level determined based on the coordination between the target monitoring station and its neighboring stations in the spatial topology. The third calculation module is used to perform a fusion calculation based on the first confidence level and the second confidence level to obtain the target confidence level of the target monitoring station; The scheduling module is used to input the target confidence level and real-time water level data into a preset digital twin scheduling model and output water resource scheduling instructions; among which, the water resource scheduling instructions are used to carry out water resource scheduling within a preset watershed.

[0016] One or more technical solutions proposed in this application have, but are not limited to, the following technical effects: This application provides a watershed water resource scheduling method and system based on digital twins and multi-principle sensors. The method acquires the spatial topology of multiple monitoring stations within a preset watershed, as well as real-time water level data of a target monitoring station. The target monitoring station is any monitoring station within the preset watershed. A first confidence level is calculated based on the historical water level data of the target monitoring station, determined by the changing patterns of the historical water level data. A second confidence level is calculated based on the spatial relationship between the target monitoring station and other stations in the spatial topology, determined by the coordination between the target monitoring station and adjacent stations in the spatial topology. The first and second confidence levels are fused to obtain a target confidence level for the target monitoring station. The target confidence level and real-time water level data are input into a preset digital twin scheduling model, and a water resource scheduling instruction is output. This water resource scheduling instruction is used to schedule water resources within the preset watershed. Therefore, by acquiring the spatial topological relationship of multiple monitoring stations within a preset watershed and the real-time water level data of the target monitoring station, the first confidence level and the second confidence level are calculated and fused to obtain the target confidence level. The target confidence level and the real-time water level data are then input into the digital twin scheduling model to output water resource scheduling instructions, thereby dynamically evaluating the credibility of the monitoring data. This method has the ability to dynamically evaluate the credibility of the monitoring data and adaptively adjust the data weights, thereby improving the reliability and security of water resource scheduling decisions. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the first embodiment of the watershed water resources scheduling method based on digital twins in this application; Figure 2 This is a flowchart illustrating the second embodiment of the watershed water resources scheduling method based on digital twins in this application; Figure 3This is a detailed flowchart illustrating the second embodiment of the watershed water resources scheduling method based on digital twins in this application; Figure 4 This is a flowchart illustrating the third embodiment of the watershed water resources scheduling method based on digital twins in this application; Figure 5 This is a flowchart illustrating the fourth embodiment of the watershed water resources scheduling method based on digital twins in this application; Figure 6 This is a schematic diagram of a watershed water resources scheduling system 800 based on digital twins provided in this application. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0019] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0020] To make the technical solutions and advantages of the embodiments of this application clearer, the exemplary embodiments of this application will be described in further detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not an exhaustive list of all embodiments.

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0022] In digital twin-based watershed water resource scheduling applications, water level data from monitoring stations are susceptible to anomalies such as missing data, drift, or sudden changes due to equipment malfunctions and environmental interference. Existing technologies use equal or fixed weights to incorporate data from different monitoring stations into scheduling model simulations, failing to effectively differentiate the actual impacts of spatial distribution differences, upstream-downstream hydraulic relationships, and temporal consistency among stations. Spatial distribution differences reflect the varying hydrological characteristics of the river sections where stations are located; upstream-downstream hydraulic relationships reflect the dynamic constraints of water level changes propagating along the river channel; and temporal consistency characterizes the degree to which water level fluctuations conform to historical patterns. This leads to the unreasonable amplification of abnormal data in scheduling calculations, significantly impacting the stability and simulation reliability of the digital twin scheduling model, thereby reducing the accuracy of water resource scheduling decisions.

[0023] For example, in a digital twin water resources management system for a tributary in the middle reaches of the Yangtze River, during periods of continuous heavy rainfall, monitoring stations located in the river's distributary areas experienced continuous drift in water level sensor readings due to localized siltation, while water level data from upstream stations on the main stream showed a gradual upward trend consistent with hydrodynamic laws. Because existing management methods do not consider the upstream-downstream topological relationships between stations and historical water level change patterns, the drift data from abnormal stations was assigned the same weight as that from normal stations in model simulations. Furthermore, the consistency between this abnormal data and the water level change trends of upstream and downstream stations was not assessed, causing the digital twin system to misjudge local sensor malfunctions as a sudden rise in water levels across the entire basin, thus generating management instructions that deviate from the actual hydrological conditions. Specifically, the time lag in water level response caused by spatial location differences, the data transmission direction determined by upstream-downstream hydraulic relationships, and deviations from historical patterns reflected in temporal consistency were all not incorporated into the dynamic adjustment mechanism of data weights, ultimately leading to distortions in the basis of management decisions.

[0024] If the above problems are not addressed, the digital twin scheduling model will struggle to maintain the continuity and stability of the simulation process when critical monitoring data is abnormal, and the deviation between the simulation results and the actual hydrological process will gradually accumulate. Specifically, the failure to adaptively reduce the weight of abnormal data increases the sensitivity of scheduling calculations to local faults, leading to a continuous decrease in the reliability of the simulation results. Furthermore, the ineffective utilization of spatial topological relationships and historical water level patterns prevents the model from distinguishing between real hydrological changes and equipment malfunctions, potentially triggering misjudgments of reservoir operating status in scheduling commands. The resulting technical consequences include: erroneous execution of flood discharge operations during the impoundment period leading to water waste; interruption of normal water supply during periods of high demand affecting public welfare; and failure to respond promptly to actual water level changes during peak flood periods threatening engineering safety, ultimately impacting the overall efficiency of water resource allocation in the basin and the safety boundaries of system operation.

[0025] Based on this, this application provides a watershed water resources scheduling method based on digital twins. This method is applied to a watershed water resources scheduling system based on digital twins. Please refer to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the watershed water resources scheduling method based on digital twins in this application.

[0026] For ease of understanding, the following explains some key terms in this embodiment: Digital twins refer to the real-time, dynamic, and high-fidelity mapping of physical entities in virtual space. In the field of watershed water resource management, digital twin systems integrate multi-source data, physical models, and artificial intelligence algorithms to construct a virtual copy of the watershed. This is used to simulate hydrological processes, predict future trends, and evaluate management schemes, thereby providing decision-makers with comprehensive information support and decision-making assistance.

[0027] Watershed water resource management refers to the process of unified planning, allocation, and management of water resources within a specific watershed. Its goal is to balance multiple needs such as flood control, water supply, power generation, and ecological protection, and to achieve optimal allocation and utilization of water resources through the regulation of water conservancy facilities such as reservoirs and sluice gates.

[0028] Monitoring stations are observation facilities deployed at specific geographical locations within a watershed, equipped with various sensors such as water level sensors, flow sensors, and water quality sensors. These stations are responsible for collecting hydrological, water quality, and other environmental data in real time and transmitting the data to a central processing system, providing basic data support for watershed management and scheduling.

[0029] Spatial topology describes the geographical location associations and hydraulic connectivity between monitoring stations within a watershed, such as upstream, downstream, and tributary confluence points. Clarifying these topological relationships helps in understanding the propagation paths of water flow and the mutual influence of hydrological events between stations, serving as a crucial basis for analyzing watershed hydrological processes and implementing coordinated management.

[0030] Real-time water level data refers to water level measurements collected by monitoring stations at the current moment or within a very short time delay. This data reflects the immediate state of the water body in the basin and is a key input for real-time simulation and decision-making in digital twin scheduling models.

[0031] Historical water level data refers to water level measurements continuously collected and stored by monitoring stations over a period of time. Analyzing historical water level data can reveal long-term trends, seasonal fluctuations, and the frequency and characteristics of abnormal events, providing a reference benchmark for assessing the reliability of current water level data.

[0032] The digital twin scheduling model is a computational model built on digital twin technology, used to simulate watershed hydrological processes, predict water resource supply and demand balance, and optimize water conservancy project operation strategies. This model can receive real-time data and confidence level information, perform complex simulations, and generate optimal scheduling schemes.

[0033] Water resource dispatching instructions are specific operational suggestions or commands output by a digital twin dispatching model based on the current hydrological conditions of the watershed, forecast results, and preset dispatching objectives. These instructions typically include reservoir discharge volume, target water storage level, and water supply flow rate, and are used to guide the actual operation of water conservancy engineering facilities.

[0034] In this embodiment, the above-mentioned watershed water resource scheduling method based on digital twins includes steps S10 to S50: Step S10: Obtain the spatial topology of multiple monitoring stations within the preset watershed, as well as the real-time water level data of the target monitoring station.

[0035] The target monitoring station is any monitoring station within the preset watershed.

[0036] There are multiple methods that can be used to obtain the spatial topology of multiple monitoring stations within a preset watershed and the real-time water level data of the target monitoring station.

[0037] One approach involves professionals using geographic information system (GIS) data and river system maps of the watershed to manually identify and label the spatial locations of each monitoring station and their upstream and downstream connections. Simultaneously, real-time water level data can be collected by deploying basic water level sensors at the monitoring stations and transmitting the data to local data storage devices via wired or wireless communication. Staff can then periodically export and upload the data.

[0038] Another method is to install simple float-type water level gauges at monitoring stations, where patrol personnel periodically read the water level scale and manually record it in paper forms. These records are then compiled and entered into a computer system.

[0039] Step S20: Calculate the corresponding first confidence level based on the historical water level data of the target monitoring station.

[0040] The first confidence level is the confidence level determined by the target monitoring station based on the changing patterns of historical water level data.

[0041] Here, the first confidence level is a quantitative indicator that assesses the reliability of the current real-time water level data based on the historical water level data change patterns of the target monitoring station. The first confidence level reflects the consistency between the current data and the inherent hydrological characteristics of the station, such as whether it conforms to the historical water level range and whether the rate of change is abnormal.

[0042] In one implementation, the daily highest and lowest water levels of the target monitoring station over a past period (e.g., one year) can be statistically analyzed to form a historical water level envelope. When real-time water level data exceeds the range of this envelope, it is considered to be inconsistent with historical patterns, thereby lowering the first confidence level.

[0043] In another implementation, the long-term average value and fluctuation range of historical water level data of the target monitoring station can be calculated, the real-time water level data can be compared with the average value, and the first confidence level can be determined according to the degree of deviation. The greater the deviation, the lower the confidence level.

[0044] In addition, a confidence level can be assigned to the current real-time water level data by experts familiar with the hydrological characteristics of the station, based on their understanding and judgment of historical water level data.

[0045] Step S30: Calculate the corresponding second confidence level based on the spatial relationship between the target monitoring station and other stations in the spatial topology.

[0046] The second confidence level is determined based on the coordination between the target monitoring station and its neighboring stations in the spatial topology.

[0047] Here, the second confidence level is a quantitative indicator that assesses the reliability of current real-time water level data based on the hydrological coordination between the target monitoring station and its neighboring stations in the spatial topology. This confidence level reflects the degree of correlation between the target station's data and the data of surrounding stations in terms of trends and magnitudes of change, such as whether changes in upstream inflow are consistent with changes in downstream water levels.

[0048] In one implementation, different spatial correlation assessment methods can be considered. For example, the direction of water level change at the target monitoring station and its directly upstream neighboring station can be simply compared within the same time period. If the water level at the target station rises while the water level at the upstream station falls, the spatial correlation is considered poor, thus lowering the second confidence level.

[0049] In another implementation, the average water level change of the target monitoring station and all upstream stations within a fixed time window can be calculated, and then the water level change of the target monitoring station can be compared with the average value. The smaller the difference, the higher the second confidence level.

[0050] In addition, human judgment can be used, with hydrological experts subjectively assessing the coordination of water level changes between the target monitoring station and its adjacent stations based on their geographical location, river characteristics, and historical hydrological data, and assigning a confidence level value.

[0051] Step S40: Perform a fusion calculation based on the first confidence level and the second confidence level to obtain the target confidence level of the target monitoring station.

[0052] Among them, the target confidence score is the final reliability assessment value of the real-time water level data of the target monitoring station obtained by combining the first confidence score and the second confidence score. The target confidence score provides a comprehensive judgment on the data quality of the digital twin scheduling model and guides the model on how to make decisions when there is uncertainty in the data.

[0053] As one implementation method, a simple arithmetic average method can be used, which involves directly adding the first confidence level and the second confidence level and then dividing by two to obtain the final target confidence level.

[0054] As another implementation method, the minimum value method can be used, taking the smaller value between the first confidence level and the second confidence level as the target confidence level to reflect the "weakest link effect".

[0055] In addition, fixed weights can be manually set, for example, assigning a weight of 0.6 to the first confidence level and a weight of 0.4 to the second confidence level, and then performing a weighted sum to obtain the target confidence level.

[0056] Step S50: Input the target confidence level and real-time water level data into the preset digital twin scheduling model, and output water resource scheduling instructions.

[0057] Among them, the water resource scheduling instruction is used to schedule water resources within a preset watershed.

[0058] After determining the target confidence level, a corresponding water resource scheduling strategy can be generated based on the target confidence level and real-time water level data, and a water resource scheduling instruction corresponding to the water resource scheduling strategy can be generated.

[0059] In one implementation, the target confidence level can be used as a simple binary judgment parameter. When the target confidence level is higher than a certain preset threshold, the real-time water level data is fully adopted and directly input into the scheduling model; when it is lower than the threshold, the real-time water level data is completely ignored, and the scheduling model instead uses historical averages or predicted values ​​for extrapolation.

[0060] In another implementation, the target confidence level can be used as a multiplier factor, directly multiplied by the real-time water level data to obtain a corrected water level value, which is then input into the digital twin scheduling model. Alternatively, the target confidence level can be used as a soft constraint in the scheduling model. For example, when the confidence level is low, the scheduling model may tend to choose a more conservative scheduling strategy to avoid potential risks arising from data uncertainty.

[0061] This application dynamically assesses the reliability of data from each monitoring station by analyzing its own historical variation patterns (first confidence level) and spatial coordination with adjacent stations (second confidence level), thereby obtaining a comprehensive target confidence level. This target confidence level is then incorporated into the digital twin scheduling model to guide how the model interprets and utilizes real-time data. Unlike existing technologies that apply equal or fixed weights to all monitoring data, this method intelligently identifies and reduces the weight of unreliable data, preventing the unreasonable amplification of abnormal input data. This avoids erroneous scheduling decisions caused by sensor failures or data anomalies, significantly improving the inference stability and simulation credibility of the digital twin scheduling model under data anomaly conditions, and providing more reliable data support for watershed water resource scheduling.

[0062] In some of the above implementation methods, a first confidence level is calculated based on the historical water level data of the target monitoring station. However, in the implementation process, the historical water level data may contain instantaneous fluctuations or unconfirmed outliers. If the confidence level is calculated directly based on these data, it may lead to an inaccurate assessment of the data reliability, thereby affecting the accuracy of water resource scheduling decisions.

[0063] Based on this, a second embodiment of this application is proposed. Please refer to [link / reference]. Figure 2 , Figure 2 This is a flowchart illustrating the second embodiment of the watershed water resources scheduling method based on digital twins in this application.

[0064] As a refinement of step S20 in the first embodiment, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment can be referred to the above description, and will not be repeated hereafter. Based on this, the watershed water resource scheduling method based on digital twins in this application also includes steps S21 to S25: Step S21: Obtain historical water level data of the target monitoring station and divide the historical water level data into multiple continuous local time windows.

[0065] Here, the water level measurement records of the site over a period of time can be retrieved from the data storage system. These water level measurement records can be stored in time series form, including timestamps and corresponding water level values.

[0066] Dividing historical water level data into multiple consecutive local time windows aims to segment long-term continuous data into shorter time periods that are easier to analyze. For example, a year's worth of historical data can be divided into daily, weekly, or hourly windows, each containing a fixed number of data points. This helps in the extraction of local features and the initial identification of anomalies.

[0067] Step S22: Calculate the water level change characteristic value corresponding to each local time window.

[0068] Among them, the water level change characteristic value is used to characterize the severity of water level fluctuations within a local time window.

[0069] Specifically, it can calculate the difference between the maximum and minimum water levels within the window, or calculate the standard deviation and variance of the water level data, or calculate the average absolute value of the rate of change of water level as a characteristic value of water level change.

[0070] It should be noted that water level change characteristic values ​​can be used to intuitively reflect the activity level of water levels in a short period of time.

[0071] Step S23: Mark the local time window where the water level change characteristic value exceeds the preset anomaly judgment threshold as the first anomaly window.

[0072] When the water level change characteristic value exceeds the preset anomaly judgment threshold, the local time window corresponding to the exceeded anomaly judgment threshold is marked as the first anomaly window.

[0073] Here, the anomaly detection threshold can be set based on statistical analysis of historical data, or according to expert experience and watershed characteristics. Furthermore, anomaly detection thresholds can be pre-set, and local time windows exceeding this threshold can be marked as the first anomaly window, which can be used to represent periods that may initially show anomalies.

[0074] In one specific implementation, when setting the anomaly judgment threshold, it can be objectively determined based on the historical statistical distribution of the water level change characteristic value. Here, the water level change characteristic value is the quantified value of the intensity of water level fluctuation within each local time window.

[0075] Furthermore, an anomaly detection threshold can be set after calculating the average and standard deviation of the water level change characteristic values ​​corresponding to all local time windows in history. Specifically, a preset multiple can be selected according to the sensitivity of the watershed to data fluctuations. For example, if k=3, then the anomaly detection threshold can be the average value plus three times the standard deviation.

[0076] Step S24: Verify the authenticity of the first abnormal window based on the historical water level data corresponding to the first abnormal window, so as to determine the second abnormal window.

[0077] The second anomaly window is the time window in historical water level data where anomalies occur.

[0078] To ensure the authenticity of the anomaly, the authenticity of the first anomaly window can be verified based on the historical water level data corresponding to the first anomaly window.

[0079] It should be noted that verifying the authenticity of the first anomaly window can distinguish whether there is a real abnormal event in the first anomaly window determined based on the anomaly judgment threshold, and normal instantaneous fluctuations or data noise, such as sensor failure, data transmission error, or actual severe hydrological events.

[0080] Specifically, the data within this window can be compared with data from adjacent time windows to check the persistence of the anomaly; or cross-validation can be performed using other relevant data (such as rainfall, upstream flow, etc.); or machine learning models can be used to identify anomaly patterns. After verification, the window confirmed as a true anomaly is designated as the second anomaly window.

[0081] Step S25: Calculate the cumulative duration of the second abnormal window appearing in multiple local time windows, and calculate the ratio between the cumulative duration of the abnormal window and the total monitoring duration to determine the first confidence level.

[0082] The total monitoring duration refers to the total monitoring duration corresponding to historical water level data.

[0083] After identifying the second anomaly window through the first anomaly window, the cumulative duration of the second anomaly window appearing in multiple local time windows can be statistically analyzed, and the ratio between this cumulative duration and the total monitoring duration can be calculated. The total monitoring duration is the entire time covered by historical water level data. This ratio quantifies the proportion of abnormal events in historical water level data, thereby determining the first confidence level.

[0084] This application employs refined analysis of historical water level data from target monitoring stations to more accurately assess data reliability. First, continuous historical water level data is divided into multiple consecutive local time windows. This allows for the capture of local data characteristics, preventing the neglect of short-term anomalies within long-term series. Next, by calculating the water level change characteristic value within each local time window, the severity of water level fluctuations is quantified, providing a basis for preliminary identification of potential anomalies. When the water level change characteristic value exceeds a preset anomaly judgment threshold, the corresponding local time window is initially marked as the first anomaly window. To avoid misjudgment, this method further verifies the authenticity of these first anomaly windows. This verification process aims to distinguish genuine anomalies from data noise or instantaneous fluctuations, thereby determining a more reliable second anomaly window. In this way, it ensures that the anomalies used for subsequent confidence calculations are verified. Finally, by statistically analyzing the cumulative duration of these genuine anomalies and calculating the ratio to the total monitoring duration, the overall reliability of historical water level data can be objectively quantified, thus determining the first confidence level. This step-by-step verification and quantification method effectively eliminates the interference of spurious anomalies when calculating the first confidence level, ensuring the accuracy and robustness of the confidence assessment. Compared with simple statistics based solely on raw historical data, this method significantly improves the granularity and accuracy of the reliability assessment of historical water level data by introducing local time window analysis and anomaly authenticity verification, providing more reliable input for subsequent digital twin scheduling models and thus optimizing the generation of water resource scheduling instructions.

[0085] In one embodiment, after the step of acquiring historical water level data of the target monitoring station, the method further includes: (1) Decompose the historical water level data to obtain the trend component, seasonal component and noise component.

[0086] Among them, the trend component is used to characterize the long-term evolution direction of water level, the seasonal component is used to characterize the periodic fluctuation pattern of water level, and the noise component is used to characterize the random fluctuation of water level.

[0087] (2) Based on the trend component, determine the reasonable range of water level changes for the target monitoring station over a long period of time.

[0088] Specifically, statistical techniques can be used to break down time-series data in historical water level data into multiple components (e.g., trends, seasons, and noise) to reveal hidden patterns and structures in the data. This allows for the systematic separation of factors affecting water level data at different time scales, enabling independent analysis of each component.

[0089] As an example, an additive or multiplicative model can be used for decomposition, or a more complex STL (Seasonal-Trend decomposition using Loess) approach can be employed.

[0090] Specifically, the trend component represents the overall trend of water level data over a long timescale. It smooths out short-term fluctuations and seasonal effects, and can be used to reflect the long-term impact of factors such as climate change, large-scale water conservancy projects, or geological changes on water levels. This component can be obtained through techniques such as moving averages, regression analysis, or locally weighted scatter smoothing.

[0091] Seasonal components represent patterns in water level data that repeat at fixed periods (such as daily, monthly, or yearly). These patterns are typically related to natural hydrological cycles such as rainfall, snowmelt, and evaporation, as well as seasonal water demand. Seasonal components can be extracted using methods such as the seasonal index method or Fourier analysis.

[0092] The noise component, also known as the residual component, is the random, unpredictable water level fluctuation that remains after removing trend and seasonal effects. It may be caused by measurement errors, sudden events, or unmodeled random factors.

[0093] The reasonable variation range is a preset range determined based on trend components, used to define the normal or acceptable boundary of long-term water level changes at the target monitoring station. This range provides a stable reference benchmark for assessing the long-term reasonableness of water levels and can be set based on historical hydrological statistics (e.g., the mean and standard deviation of trend components) or in combination with hydrological expert experience and watershed management regulations.

[0094] In practical applications, a reasonable range of variation can be determined by fitting the envelope of the trend component. Specifically, the local upper and lower envelopes of the trend component sequence are calculated, and these envelopes can be formed by smoothly connecting the maximum and minimum values ​​within a moving window. Further, linear fitting is performed on the upper and lower envelopes respectively, resulting in two fitted straight lines. The slope of the fitted lower envelope is denoted as... The slope obtained by fitting the upper envelope is denoted as . Finally, based on the two fitting coefficients... and Construct the reasonable variation range, and represent it as the reasonable variation range. This reasonable range of variation represents the reasonable slope range of the long-term water level change trend.

[0095] In this way, a reasonable range of change is objectively and quantitatively established, providing a clear numerical benchmark for subsequent judgments on whether the current trend deviates from the long-term normal range.

[0096] Further, please see Figure 3 , Figure 3 This is a detailed flowchart of the second embodiment. In one specific implementation, step S24, verifying the authenticity of the first abnormal window based on the historical water level data corresponding to the first abnormal window, to determine the second abnormal window specifically includes: Step S241: Extract the seasonal component segment and the noise component segment that are temporally aligned with the first anomaly window from the seasonal component and the noise component.

[0097] The steps of extracting seasonal and noise component segments aim to separate the effects of periodic fluctuations and random disturbances from complex water level data.

[0098] It should be noted that the seasonal component represents the regular changes in water level over a specific time period due to seasonal factors (such as rainfall, snowmelt, and evaporation), while the noise component represents unexplained random fluctuations in the data. By extracting these components that are precisely aligned with the time range of the first anomaly window, background data can be provided for subsequent trend analysis.

[0099] Specifically, after performing time series decomposition on historical water level data, the corresponding time period data can be extracted from the decomposed seasonal and noise components based on the start and end times of the first anomaly window.

[0100] Step S242: Perform point-by-point subtraction operations on the actual observed data sequence of the first anomaly window with the seasonal component segment and the noise component segment to obtain the current trend data sequence corresponding to the first anomaly window.

[0101] It should be noted that the current trend data sequence obtained through point-by-point subtraction is derived by removing the influence of seasonality and randomness from the actual observation data within the first anomaly window, thus highlighting the true trend of water level changes within that window. By subtracting the seasonal and noise components corresponding to the time point from each data point in the actual observation data sequence, a sequence reflecting only the long-term trend of water level changes can be obtained. This effectively eliminates the interference of short-term, non-trend factors in anomaly identification.

[0102] Step S243: Perform linear fitting on the current trend data sequence to determine the current trend characteristics corresponding to the first anomaly window.

[0103] It should also be noted that linear fitting to determine the current trend characteristics can quantify the trend strength and direction of water level changes within the first anomaly window. By performing linear fitting on the current trend data sequence, a geometric mathematical model can be obtained, whose parameters such as slope and intercept can serve as the water level trend characteristics within that window. For example, the slope can represent the rate of rise or fall in water level, and the intercept can represent the baseline value of the trend.

[0104] In addition to linear fitting, multinomial fitting and locally weighted regression can also be used to capture trend characteristics. Comparing the current trend characteristics with a reasonable range of change is crucial for verifying the authenticity of the first anomaly window.

[0105] Step S244: Compare the current trend characteristics with the reasonable range of change.

[0106] Here, the reasonable variation range is determined based on the long-term trend components of historical water level data, representing the reasonable range of long-term water level changes under normal circumstances. By comparing the current trend characteristics of the first anomaly window with this reasonable variation range, it can be determined whether the water level changes within the window exceed the normal fluctuation range of the long-term trend, thereby determining whether the first anomaly window is the second anomaly window that will ultimately confirm a true anomaly.

[0107] Step S245: If the current trend characteristics are not within a reasonable range of change, then the first abnormal window is determined as the second abnormal window.

[0108] The second abnormal window is the abnormal window that is verified after the first abnormal window is determined to be a real abnormality.

[0109] If the comparison results show that the current trend characteristics are not within a reasonable range of change, it indicates that the water level change within the first anomaly window has indeed deviated from the normal trajectory of the long-term trend, thus confirming it as a genuine anomaly event, i.e., the second anomaly window. This ensures that only events with significant deviations from the long-term trend are identified as anomalies, improving the accuracy of anomaly detection.

[0110] For example, the specific process for calculating the first confidence level is as follows: For the target monitoring station (station k), the water level change curve obtained by the water level sensor can be acquired. (Real-time water level data), with Based on this, multiple consecutive local time windows are constructed at preset intervals (e.g., 2 days). For any local time window Calculate the characteristic value of water level change , specifically, The calculation formula is as follows:

[0111] in, For the u-th local time window Water level change characteristic value, This represents the i-th water level value within the local time window. This represents the (i-1)th water level value within this local time window.

[0112] Before determining the water level change curve Each local time window After obtaining the corresponding water level change characteristic value Q, the water level change characteristic values ​​for each local time window can be statistically analyzed. If the water level change characteristic value Q exceeds the historical normal water level change characteristic threshold, it indicates that there is an anomaly in the water level change within the time window, and the time window is determined as the first abnormal window.

[0113] Furthermore, the water level change curve can be analyzed. Perform STL decomposition to determine the trend components of the water level change curve. Seasonal portion and noise components , for trend components By determining the upper and lower envelopes and performing linear fitting, the range of evolution trend coefficients for long-term water level changes can be obtained. (Reasonable range of variation), among which, The fitting coefficients for the lower envelope are... The fitting coefficient of the upper envelope is used as the coefficient range of the long-term evolution trend of the water level as the benchmark for judging the reasonable trend range.

[0114] Furthermore, for each first anomaly window (preliminary anomaly window), the corresponding water level change curve within that time window is extracted. and the corresponding seasonal components within that time window. (Seasonal component segment) and noise classification (Noise component segment). In determining... , and Then, the actual observation data sequence of the first anomaly window (the corresponding water level change curve within that time window) can be used. ), and seasonal weighting and noise component segment Perform point-by-point subtraction to obtain the current trend data sequence corresponding to the first anomaly window. The specific calculation formula is as follows:

[0115] Furthermore, in determining the current trend data sequence Then, the current trend data sequence can be analyzed. Perform linear fitting to determine the current trend characteristics corresponding to the first anomaly window. If the water level changes within the trend coefficient range (reasonable change range) are within the range of historical normal evolution, then the water level changes within the initial abnormal window are still within the range of historical normal evolution. Therefore, the window is determined to be a normal window and does not constitute a real abnormality. This means that the water level curve of the current time window is normal. Otherwise, there is an abnormality, and the first abnormal window is determined to be the second abnormal window.

[0116] Furthermore, by statistically analyzing multiple local time windows in the water level change curve, the cumulative anomaly time of the second anomaly window can be used to identify the current trend characteristics. With respect to the reasonable range of change determined from historical trend components Compare the current trend characteristics. satisfy That is, the current trend characteristics It falls within the range of evolutionary trend coefficients (a reasonable range of change). If the water level changes within the initial abnormal window are within the range of historical normal evolution, then the window is determined to be a normal window and does not constitute a real anomaly. This means that the water level curve of the current time window is normal. Otherwise, there is an anomaly, and the first abnormal window is determined to be the second abnormal window.

[0117] Furthermore, the water level change curve can be statistically analyzed. The cumulative duration of the anomaly in the second anomaly window, which is identified within multiple local time windows, is determined to contain the anomaly. And further calculate the cumulative duration of the anomalies. Between and the total monitoring duration The ratio of the two values ​​is used to obtain the first confidence level. .

[0118] Here, it should be noted that, It can intuitively reflect the proportion of abnormal situations in the historical data of this monitoring station. The higher the value, the more frequently the site has generated unreliable and abnormal data in the past, and the worse the overall reliability of its historical data.

[0119] In some of the embodiments described above in this application, a second confidence level is calculated based on the spatial relationship between the target monitoring station and other stations in the spatial topology. However, in actual watershed water resource scheduling, relying solely on general spatial relationships may not accurately capture the specific impact mechanism of upstream stations on the water level changes of the target monitoring station, resulting in a lack of refinement and accuracy in the calculated second confidence level, thereby affecting the reliability of subsequent water resource scheduling instructions.

[0120] Based on this, a third embodiment of this application is further proposed; please refer to [link / reference]. Figure 4 , Figure 4 This is a flowchart illustrating the third embodiment of the watershed water resources scheduling method based on digital twins in this application.

[0121] As a refinement of step S30 in the first embodiment, in the first and second embodiments of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description and will not be repeated hereafter. Based on this, the watershed water resource scheduling method based on digital twins in this application also includes steps S31 to S36: Step S31: Determine multiple upstream stations of the target monitoring station within the preset watershed.

[0122] Among them, the multiple upstream stations within the preset watershed that determine the target monitoring station can refer to all monitoring stations identified in the watershed system that are located upstream of the target monitoring station in the direction of water flow and have a direct or indirect impact on the water level changes of the target monitoring station.

[0123] Specifically, this can be achieved by analyzing the watershed's river system map, digital elevation model data, and hydrological connectivity model to identify all upstream tributaries and main streams into which water flows flow to the target monitoring station. Alternatively, geographic information system (GIS) tools can be used, combined with the watershed's hydrological and geographical features, to automatically or semi-automatically identify and mark all relevant monitoring stations upstream of the target monitoring station.

[0124] Step S32: Calculate the similarity of water level changes between each upstream station and the target monitoring station.

[0125] Calculating the similarity between the water level changes at each upstream station and the target monitoring station can quantify the synchronous or lagging impact of water level changes at upstream stations on water level changes at the target monitoring station.

[0126] As one approach, time series analysis methods, such as cross-correlation functions, dynamic time warping algorithms, or Pearson correlation coefficients, can be used to analyze historical water level data to assess the synchronicity or lag correlation of water level change trends at two stations.

[0127] As another approach, machine learning models, such as recurrent neural networks or long short-term memory networks, can be used to learn the nonlinear relationship between water level changes at upstream stations and water level changes at target monitoring stations, and output a similarity index.

[0128] Step S33: Calculate the spatial location influence weight of each upstream station based on the river slope and river distance between each upstream station and the target monitoring station.

[0129] Here, the spatial location influence weight can be used to represent the relative importance of upstream stations in influencing water level changes at target monitoring stations due to their geographical location (e.g., river slope, river distance). Physical geographical factors can be incorporated into the assessment to more accurately evaluate the hydrological impact of upstream stations on downstream stations; for example, closer distances, steeper slopes, and faster water flow can lead to more significant impacts.

[0130] One approach is to establish empirical formulas to convert river slope, which affects water flow velocity and transmission time, and river distance, which affects water flow attenuation and transmission path, into weighting factors. For example, a function can be designed such that the greater the slope and the closer the distance, the higher the weight value.

[0131] Another approach is to utilize hydrological geographic information system data, combined with hydrological models, to simulate the transmission characteristics of water flow under different river conditions. Based on the simulation results, the contribution of different upstream stations to the water level changes of the target station can be analyzed, thereby determining the spatial location influence weight.

[0132] In one specific implementation, the step of calculating the spatial location influence weight of each upstream station based on the river slope and river distance between each upstream station and the target monitoring station includes: (1) Obtain the river slope and river distance between the upstream station of the target and the target monitoring station; wherein, the upstream station of the target is any upstream station of the target monitoring station; (2) Determine the slope influencing factor based on the river slope, wherein the slope influencing factor increases with the increase of slope; (3) Determine the distance influence factor based on the river channel distance, wherein the distance influence factor decreases as the distance increases; (4) The slope influence factor and the distance influence factor are weighted and summed to obtain the spatial location influence weight of the target monitoring station.

[0133] Among them, the river slope reflects the gravitational potential energy and velocity potential of the water flow. The greater the slope, the faster the water flow usually is, and the more rapidly the impact is transmitted downstream. The river distance represents the path length that the water flow needs to take from the upstream station to the target monitoring station. The longer the distance, the more significant the attenuation and delay effects may be during the water flow transmission process.

[0134] The steeper the river channel slope, the stronger the kinetic energy and transport capacity of the water flow, and the greater its instantaneous impact on downstream water levels. Therefore, the slope influence factor should increase with the slope to reflect this positive correlation. This factor can be mapped to the slope value using a linear function or a piecewise linear function. For example, a baseline slope value can be set, and when the actual slope is higher than the baseline value, the slope influence factor increases by a certain proportion; when it is lower than the baseline value, it decreases by another proportion. Alternatively, nonlinear functions, such as exponential or logarithmic functions, can be used to better simulate the complexity of the slope's influence on water flow. For example, when the slope is small, the influence factor increases slowly; when the slope reaches a certain threshold, the growth rate of the influence factor accelerates to reflect the significant impact of rapid currents on downstream areas.

[0135] The longer the river channel, the greater the impact of factors such as resistance, evaporation, and seepage on the water flow during transmission, leading to flow attenuation and lag, and weakening the instantaneous impact on downstream water levels. Therefore, the distance influence factor should decrease with increasing distance to reflect this negative correlation. This factor can be determined using an inverse proportional function or a negative exponential function. Alternatively, a piecewise function can be used, employing different attenuation rates in different distance intervals—for example, slower attenuation over shorter distances and faster at longer distances—to simulate the actual physical process of water flow transmission.

[0136] The step of weighted summation of the slope and distance influence factors aims to comprehensively consider the contributions of river slope and distance to the spatial location of upstream stations. By weighting and summing, two independent physical influence factors are merged into a unified spatial location influence weight. This spatial location influence weight can be used to reflect the comprehensive geographical characteristics of the impact of upstream stations on the hydrological conditions of the target monitoring station.

[0137] Step S34: Combine similarity and spatial location influence weights to determine the comprehensive influence weight of each upstream station on the target monitoring station.

[0138] The comprehensive influence weight can be used to calculate the expected water level changes at the target monitoring stations.

[0139] One approach is to perform a weighted average or product operation on the similarity and spatial location influence weights. For example, the overall influence weight can be expressed as a linear combination of the similarity and spatial location influence weights, where the coefficients can be adjusted according to the actual situation.

[0140] Another approach is to construct a multi-factor fusion model, such as fuzzy comprehensive evaluation or analytic hierarchy process, to comprehensively evaluate similarity, spatial location influence weights, and other potential influencing factors (e.g., catchment area, vegetation cover, etc.) to obtain a comprehensive influence weight.

[0141] Step S35: Calculate the expected water level change characteristic value of the target monitoring station based on the current water level change characteristic value of each upstream station and its corresponding comprehensive influence weight.

[0142] Here, the expected water level change characteristic value can provide a benchmark for judging whether the actual water level change of the target monitoring station is abnormal, that is, the water level change that the target station should have under normal hydrological connectivity.

[0143] One approach is to multiply the current water level change characteristic value of each upstream station with its corresponding comprehensive influence weight, and then sum the products of all upstream stations to obtain the expected water level change characteristic value of the target monitoring station.

[0144] Another approach is to use machine learning-based prediction models, such as support vector regression or gradient boosting trees, to train the model using the current water level change features and comprehensive impact weights of the upstream stations as input features to predict the expected water level change features of the target monitoring stations.

[0145] In one specific implementation, the step of calculating the expected water level change characteristic value of the target monitoring station based on the current water level change characteristic value of each upstream station and its corresponding comprehensive influence weight includes: (1) Obtain the current water level change characteristic value of each upstream station during the current monitoring period.

[0146] The current water level change characteristic value can be used to represent the water level fluctuation at upstream stations within a specific time period. The current water level change characteristic value can characterize the trend or severity of water level changes during the current monitoring period.

[0147] For example, it can be calculated as the difference between the maximum and minimum water levels during the current monitoring period, the absolute value of the rate of change of water level, or the standard deviation of the water level series.

[0148] (2) Multiply the current water level change characteristic value of each upstream station by its corresponding comprehensive influence weight to obtain the weighted contribution value of each upstream station.

[0149] The weighted contribution value can be used to quantify the specific impact of each upstream station on the water level changes of the target monitoring station.

[0150] The comprehensive impact weight is a weight calculated by combining factors such as the similarity of water level changes between upstream stations and target monitoring stations, river slope, and river distance. The comprehensive impact weight can be used to reflect the intensity of the hydrological impact of upstream stations on target monitoring stations.

[0151] By multiplying the current water level change characteristic value of the upstream station with this weight, the specific "contribution" or "influence share" of the upstream station to the expected water level change of the target monitoring station can be obtained.

[0152] (3) Sum the weighted contribution values ​​of all upstream stations to obtain the expected water level change characteristic value of the target monitoring station.

[0153] The expected water level change characteristic value can be used to synthesize the weighted contributions of all relevant upstream stations, thereby predicting the trend or magnitude of water level changes at the target monitoring station. The expected water level change characteristic value is a reasonable inference about the water level changes at the target station based on the upstream hydrological conditions.

[0154] For example, if multiple upstream stations show a significant rise in water level, and these stations have a high overall weight in their influence on the target station, then the expected water level change characteristic value of the target station will also show an upward trend or a large change.

[0155] Step S36: Calculate the second confidence level based on the actual water level change characteristic value and the expected water level change characteristic value of the target monitoring station.

[0156] The second confidence level can be used to reflect the consistency or reasonableness between the actual water level change at the target monitoring station and the expected water level change based on upstream influence prediction, in order to assess the spatial reliability of the target station data.

[0157] One approach is to determine whether the actual water level change characteristic value and the expected water level change characteristic value are consistent in direction. If they are inconsistent, the confidence level is low; if they are consistent, the confidence level is calculated based on the degree of difference between the two. The greater the difference, the lower the confidence level.

[0158] Another approach is to construct a confidence function that takes the deviation between the actual and expected values ​​as input and outputs a confidence value between 0 and 1. For example, a Gaussian radial basis function or a Sigmoid function can be used for mapping, so that the confidence is high when the deviation is within a certain range and drops rapidly when it exceeds the range.

[0159] In one specific implementation, the step of calculating the second confidence level based on the actual water level change characteristic value and the expected water level change characteristic value of the target monitoring station includes: (1) Determine whether the actual water level change characteristic value and the expected water level change characteristic value are consistent in their direction of change.

[0160] It should be noted that directional indicators that clearly reflect the overall trend of water level rise and fall can be used to determine whether the direction of change of the actual water level change value and the expected water level change characteristic value are consistent. Here, the directional indicator can be the magnitude relationship between the actual water level change characteristic value and the expected water level change characteristic value.

[0161] Here, when judging whether the actual water level change characteristic value and the expected water level change characteristic value are consistent in their direction of change, the direction of change refers to whether the water level reflected by the water level change characteristic value is on an upward trend, a downward trend, or remains relatively stable.

[0162] To determine consistency, one can compare the signs of two feature values. For example, a positive value indicates an increase, a negative value indicates a decrease, and zero indicates stability. If the signs are the same, the directions are considered consistent; if the signs are different, the directions are considered inconsistent. Alternatively, a change threshold can be set. When a feature value exceeds this threshold, it is determined to be either an increase or a decrease, and then the increase / decrease states of the two values ​​can be compared to see if they are the same.

[0163] (2) If the direction of change is inconsistent, the second confidence level is set to the preset minimum confidence level value.

[0164] If the directions of change are inconsistent, the second confidence level will be set to the preset minimum confidence level value.

[0165] When the actual water level change characteristic value is inconsistent with the expected water level change characteristic value, it indicates that the water level change at the target monitoring station and the combined influence of the upstream station are significantly inconsistent, which usually indicates that there may be an anomaly at the target monitoring station.

[0166] Therefore, setting the second confidence level to a preset minimum confidence value is intended to immediately reflect this high level of uncertainty or anomaly. The minimum confidence value can be a preset fixed value, such as 0 or a very small positive number close to 0, such as 0.01, to ensure that the confidence level of the site is significantly reduced when the directions are inconsistent.

[0167] (3) If the direction of change is consistent, calculate the degree of difference between the actual water level change characteristic value and the expected water level change characteristic value, and calculate the second confidence level based on the degree of difference. The greater the degree of difference, the lower the second confidence level.

[0168] The greater the degree of difference, the lower the second confidence level.

[0169] When the actual water level change characteristic value and the expected water level change characteristic value change in the same direction, it indicates that the water level change trend at the target monitoring station is coordinated with the overall influence trend at the upstream station. At this point, it is necessary to further quantify the degree of this coordination.

[0170] Here, the degree of difference can be determined by calculating the absolute difference, relative difference, or mean squared error between the two. Specifically, the second confidence level can be calculated based on the degree of difference using methods such as inverse proportional functions, exponential decay functions, or piecewise linear functions. For example, a function can be defined such that the confidence level is highest when the degree of difference is zero, and gradually decreases as the degree of difference increases until the degree of difference reaches a certain threshold, at which point the confidence level drops to its lowest point.

[0171] This application first identifies multiple upstream stations that influence the target monitoring station and then constructs a comprehensive influence weight by comprehensively considering the similarity of historical water level changes between these upstream stations and the target monitoring station, as well as physical geographical factors (such as river slope and distance). This multi-dimensional and refined weight calculation method enables the system to more accurately quantify the actual hydrological impact of each upstream station on the target monitoring station. Subsequently, using these comprehensive influence weights and the current water level change characteristic values ​​of the upstream stations, the expected water level change characteristic value of the target monitoring station is calculated. This expected value represents the expected water level change trend of the target station under normal hydrological connectivity. Finally, by comparing the actual water level change characteristic value of the target monitoring station with the expected water level change characteristic value, the spatial rationality of the target station data can be more accurately assessed, thereby calculating a more reliable second confidence level. This mechanism avoids the limitations of judging solely based on general spatial relationships, making the calculation of the second confidence level closer to the actual hydrological situation of the watershed and providing high-quality input data for subsequent digital twin scheduling models.

[0172] For example, the specific process for calculating the second confidence level is as follows: Target monitoring sites It is possible to identify target monitoring sites. Water level change curve And further identify target monitoring stations. upstream site set (Multiple upstream sites), targeting a set of upstream sites Any target upstream station Determine the upstream site of the target Water level change curve And calculate the upstream site of the target Water level change curve With target monitoring stations Water level change curve similarity between The specific calculation formula is as follows:

[0173] in, The function is used to determine the dynamic event warping method. and The degree of difference between them, when and The smaller the distance between the two curves, the smaller the difference between them and the higher their similarity.

[0174] Furthermore, the upstream sites of the target can be identified. With target monitoring stations River slope between and river distance And based on the river slope and river distance Calculate the weight of spatial location influence The specific calculation formula is as follows:

[0175] in, The closer it is to 1, the steeper the slope and the faster the water flow, and the higher the probability that the upstream station will flow into the current target station; This represents the longest river channel distance from the target station to the upstream station. This indicates that the longer the river channel, the smaller the impact of the upstream water level on the target station's water level.

[0176] After identifying target monitoring sites Water level change curve similarity between And spatial location influence weight Then, the upstream site of the target can be determined. For target monitoring stations water level impact (Comprehensive influence weight), the specific calculation formula is as follows:

[0177] in, The function is a normalization function, and its influence on spatial location is weighted. Normalization was performed. It should be noted that this was done when determining the similarity of water level changes between the target monitoring station and any of its upstream stations. And spatial location influence weight This allows us to calculate the overall influence weight of the upstream site on the target site. Among them, spatial location influences weight. This only reflects the physical impact potential of a single site based on slope and distance. To make a fair comparison among all upstream sites for the same target site and to quantify the proportional relationship of each site's relative influence, it is necessary to analyze the physical impact potential of all upstream sites. The values ​​are normalized.

[0178] Furthermore, using the most recent time window... Based on this, obtain the characteristic values ​​of water level changes at each upstream station at the current moment. The set, that is, And the water level change characteristic value of the target monitoring station k itself. Determine the current target station based on changes in upstream water levels. Expected water level change characteristic value The specific calculation formula is as follows:

[0179] Among them, for Characteristic values ​​of water level changes at each upstream station The degree of influence of water level Multiply by these values ​​to obtain the weighted contribution value for each upstream station, and then... The weighted contribution values ​​of each upstream station are summed to obtain the characteristic value of the expected water level change. .

[0180] Determining the characteristic values ​​of expected water level changes Then, the characteristic values ​​of water level changes at the current target station can be compared. Characteristic values ​​of expected water level changes The sign of the two values ​​is the same. When the signs are the same, it indicates that the current water level change at the target station is in the same direction as the expected trend. Therefore, based on the water level change characteristic value of the target station... Characteristic values ​​of expected water level changes The differences were used to determine the confidence level of spatial water flow changes at the current target station. (Second confidence level), calculated using the following formula:

[0181] in, The characteristic value of water level change Characteristic values ​​of expected water level changes The difference value indicates a lower confidence level; the larger the value, the lower the confidence level. Here, q is not zero. It should be noted that when comparing the actual water level changes at the target monitoring stations... Characteristic values ​​of expected water level changes based on upstream relationships If the two change in the same direction, then the relative difference value can be calculated. This difference value reflects the relative magnitude and direction of the deviation of the actual change from the expected value (a positive value indicates that the actual change is greater than expected, and a negative value indicates the opposite). In order to evaluate the difference of a single station within the context of the overall difference of all monitored stations at the current scheduling time, normalization processing is required.

[0182] When the signs of the two are different, it indicates that there is an anomaly at the target station, and the confidence level (second confidence level) of the water flow change between the current target stations is determined. .

[0183] In another implementation, after determining the first confidence level... Second confidence level Then, the target confidence level of the target monitoring station can be obtained by fusing the first and second confidence levels. The specific calculation formula is as follows:

[0184] in, The function is a normalized function.

[0185] After calculating the confidence level of historical water level changes for each monitoring station. confidence level of spatial water flow variation Then, the intermediate fusion value of the site can be obtained through multiplication. This value comprehensively reflects the reliability of a station in both temporal and spatial dimensions. To place the reliability assessment of each station within a unified decision-making framework across the entire watershed, and to ensure its comparability and operability (e.g., for subsequent data tiering and scheduling), an intermediate fusion value from all stations is required. Normalization is performed.

[0186] Based on the above embodiments, a fourth embodiment of this application is proposed. Please refer to [link to embodiment]. Figure 5 , Figure 5 This is a flowchart illustrating the fourth embodiment of the watershed water resources scheduling method based on digital twins in this application.

[0187] As a refinement of step S50 in the first embodiment, the same or similar content in the above embodiments of this application can be referred to the above description, and will not be repeated hereafter. Based on this, the watershed water resource scheduling method based on digital twins in this application also includes steps S51~S53: Step S51: Based on the target confidence level, divide the real-time water level data into core input data and auxiliary input data.

[0188] The core input data is used to directly drive scheduling decisions, while the auxiliary input data is used to correct parameters within the scheduling decision framework.

[0189] Here, real-time water level data forms the basis for water resource allocation decisions, but its reliability can vary due to factors such as the monitoring environment and equipment malfunctions. Target confidence reflects the reliability of real-time water level data.

[0190] It should be noted that dividing real-time water level data into core input data and auxiliary input data aims to classify and process them according to their reliability (reflected by target confidence level) in order to improve the accuracy and robustness of scheduling decisions.

[0191] Core input data consists of data with high confidence levels and considered highly reliable, which can be directly used to drive the main scheduling decision logic. Auxiliary input data, on the other hand, consists of data with relatively lower confidence levels or used to supplement information, which can be used to fine-tune or correct parameters within the main decision framework to cope with uncertainty.

[0192] For example, a confidence threshold can be set. When the target confidence level is higher than the threshold, the corresponding real-time water level data is classified as core input data; when the target confidence level is lower than the threshold, it is classified as auxiliary input data. Alternatively, fuzzy logic or machine learning models can be used to dynamically allocate the weights or roles of real-time water level data in core and auxiliary inputs based on the continuous values ​​of the target confidence level, rather than a simple binary partition.

[0193] Step S52: Based on the core input data, determine the macroscopic operating range of at least one reservoir within the preset watershed.

[0194] The macro-operation range is determined based on real-time water levels and preset scheduling rules. The macro-operation range includes flood discharge range, water storage range, and water supply range.

[0195] Here, the macro-operational range refers to the overall operational status of the reservoir under specific real-time water level conditions.

[0196] It should be noted that by using core input data, the current water level of the reservoir can be accurately determined, and combined with preset scheduling rules, the reservoir's operational status can be divided into different macro-level intervals. These intervals guide the main scheduling strategies that the reservoir should adopt in the current state. For example, when the water level is too high, it enters the flood discharge interval; when the water level is moderate, it enters the water storage interval; and when the water level is low, it enters the water supply interval. This can be achieved, for example, through table lookup or conditional statements. Different macro-level operational intervals corresponding to different water level thresholds are predefined. For example, if the real-time water level is higher than the flood control limit, it is determined to be in the flood discharge interval; if the real-time water level is between the normal water storage level and the flood control limit, it is determined to be in the water storage interval; and if the real-time water level is lower than the normal water storage level but higher than the dead water level, it is determined to be in the water supply interval. Alternatively, a rule-based expert system or decision tree model can be used, taking complex scheduling rules and real-time water level data as input, to automatically infer and determine the macro-level operational interval of the reservoir.

[0197] Step S53: Within the determined macro-operation range, the scheduling operation parameters are optimized and calculated in conjunction with auxiliary input data to generate water resource scheduling instructions.

[0198] The scheduling operation parameters include at least one of the following: flood discharge flow, target water level for water storage, or water supply flow.

[0199] After determining the macroscopic operating range of the reservoir, specific scheduling operations can be further refined.

[0200] While auxiliary input data may have relatively low confidence levels, they provide additional, supplementary information that can be used to fine-tune and optimize scheduling parameters within a macro-level context. Scheduling parameters are key indicators that directly affect reservoir operation; for example, specific discharge flows need to be determined in the flood discharge zone, target water levels may need to be adjusted in the storage zone, and water supply flows need to be determined in the supply zone. By combining auxiliary input data with optimization calculations, scheduling instructions can be made more precise and flexible to cope with various uncertainties and complexities in actual operation.

[0201] For example, multi-objective optimization algorithms, such as genetic algorithms and particle swarm optimization algorithms, can be used to calculate the optimal scheduling operation parameters by taking the constraints of the macroscopic operating range, auxiliary input data, and scheduling objectives as optimization objectives or constraints. Alternatively, model-based predictive control methods can be used to simulate the reservoir response under different scheduling operation parameters using hydrological and hydrodynamic models combined with auxiliary input data, and then select the parameter combination that meets the requirements of the macroscopic operating range and achieves the best results.

[0202] After acquiring the target confidence level and real-time water level data from the target monitoring stations, this application does not directly input all data into the digital twin scheduling model. Instead, it first intelligently classifies the real-time water level data based on the target confidence level, dividing it into core input data and auxiliary input data. This hierarchical processing allows the highly reliable core input data to directly drive the judgment of the reservoir's macro-operational range, ensuring the accuracy of scheduling decisions. Specifically, based on the core input data, the system can accurately identify the reservoir's current flood discharge range, water storage range, or water supply range, thus setting clear boundaries and targets for subsequent refined scheduling. Within this macro framework, the system further combines auxiliary input data to optimize specific scheduling operation parameters, such as flood discharge flow, water storage target level, or water supply flow. Although the auxiliary input data may have a lower confidence level, as supplementary information, it corrects and fine-tunes the parameters within the macro range, ensuring that the final generated water resource scheduling instructions not only conform to the macro strategy but also adapt more precisely to the actual situation, improving the flexibility and adaptability of scheduling decisions. By employing this hierarchical processing and layer-by-layer optimization approach, the proposed solution effectively addresses the contradiction between data reliability and decision-making precision in complex water resource scheduling, ensuring the scientific validity and effectiveness of scheduling instructions.

[0203] Please see Figure 6 , Figure 6A schematic diagram of a digital twin-based watershed water resources scheduling system 60 provided in this application, the digital twin-based watershed water resources scheduling system 60 includes: The acquisition module 61 is used to acquire the spatial topology of multiple monitoring stations within a preset watershed, as well as the real-time water level data of the target monitoring station; wherein, the target monitoring station is any monitoring station within the preset watershed. The first calculation module 62 is used to calculate the corresponding first confidence level based on the historical water level data of the target monitoring station; wherein, the first confidence level is the confidence level determined by the target monitoring station based on the changing pattern of the historical water level data; The second calculation module 63 is used to calculate the corresponding second confidence level based on the spatial relationship between the target monitoring station and other stations in the spatial topology; wherein, the second confidence level is a confidence level determined based on the coordination between the target monitoring station and its neighboring stations in the spatial topology. The third calculation module 64 is used to perform a fusion calculation based on the first confidence level and the second confidence level to obtain the target confidence level of the target monitoring station; The scheduling module 65 is used to input the target confidence level and real-time water level data into a preset digital twin scheduling model and output water resource scheduling instructions; wherein, the water resource scheduling instructions are used to carry out water resource scheduling within a preset watershed.

[0204] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A watershed water resources scheduling method based on digital twins, characterized in that, The method includes: The spatial topology of multiple monitoring stations within a preset watershed and the real-time water level data of a target monitoring station are obtained; wherein, the target monitoring station is any monitoring station within the preset watershed. Based on the historical water level data of the target monitoring station, a corresponding first confidence level is calculated; wherein, the first confidence level is the confidence level determined by the target monitoring station based on the changing pattern of historical water level data; A second confidence level is calculated based on the spatial relationship between the target monitoring station and other stations in the spatial topology; wherein, the second confidence level is a confidence level determined based on the coordination between the target monitoring station and its neighboring stations in the spatial topology. The target confidence level of the target monitoring station is obtained by fusing the first confidence level and the second confidence level. The target confidence level and the real-time water level data are input into a preset digital twin scheduling model, and a water resource scheduling instruction is output; wherein, the water resource scheduling instruction is used to schedule water resources within the preset watershed.

2. The watershed water resources scheduling method based on digital twins according to claim 1, characterized in that, The step of calculating the corresponding first confidence level based on the historical water level data of the target monitoring station includes: The historical water level data of the target monitoring station is obtained, and the historical water level data is divided into multiple consecutive local time windows; Calculate the water level change characteristic value corresponding to each local time window; wherein, the water level change characteristic value is used to characterize the intensity of water level fluctuation within the local time window; The local time window in which the water level change characteristic value exceeds the preset anomaly judgment threshold is marked as the first anomaly window; The authenticity of the first abnormal window is verified based on the historical water level data corresponding to the first abnormal window to determine the second abnormal window; wherein, the second abnormal window is the time window in the historical water level data in which the abnormality occurs. The cumulative duration of the second abnormal window occurring in multiple local time windows is statistically analyzed, and the ratio between the cumulative duration of the abnormality and the total monitoring duration is calculated to determine the first confidence level; wherein, the total monitoring duration is the total monitoring duration corresponding to the historical water level data.

3. The watershed water resources scheduling method based on digital twins according to claim 2, characterized in that, After the step of acquiring the historical water level data of the target monitoring station, the method further includes: The historical water level data is decomposed to obtain trend components, seasonal components, and noise components; wherein, the trend component is used to characterize the long-term evolution direction of the water level, the seasonal component is used to characterize the periodic fluctuation pattern of the water level, and the noise component is used to characterize the random fluctuation of the water level. Based on the trend components, a reasonable range of water level changes for the target monitoring station over a long period is determined.

4. A watershed water resources scheduling method based on digital twins according to claim 3, characterized in that, The step of verifying the authenticity of the first abnormal window based on the historical water level data corresponding to the first abnormal window to determine the second abnormal window includes: Extract the seasonal component segment and the noise component segment that are temporally aligned with the first anomaly window from the seasonal component and the noise component; The actual observed data sequence of the first anomaly window is subtracted point by point from the seasonal component segment and the noise component segment to obtain the current trend data sequence corresponding to the first anomaly window. Perform linear fitting on the current trend data sequence to determine the current trend characteristics corresponding to the first anomaly window; Compare the current trend characteristics with the reasonable range of change; If the current trend feature is not within the reasonable change range, then the first abnormal window is determined as the second abnormal window; wherein, the second abnormal window is the abnormal window after verifying that the first abnormal window is determined to be a real abnormality.

5. A watershed water resources scheduling method based on digital twins according to claim 1, characterized in that, The step of calculating the corresponding second confidence level based on the spatial relationship between the target monitoring station and other stations in the spatial topology includes: Identify multiple upstream stations of the target monitoring station within the preset watershed; Calculate the similarity of water level changes between each of the upstream stations and the target monitoring station; Based on the river slope and river distance between each upstream station and the target monitoring station, the spatial location influence weight of each upstream station is calculated. By combining the similarity and the spatial location influence weight, the comprehensive influence weight of each upstream station on the target monitoring station is determined; Based on the current water level change characteristic value of each upstream station and its corresponding comprehensive influence weight, the expected water level change characteristic value of the target monitoring station is calculated. The second confidence level is calculated based on the actual water level change characteristic value of the target monitoring station and the expected water level change characteristic value.

6. A watershed water resources scheduling method based on digital twins according to claim 5, characterized in that, The step of calculating the spatial location influence weight of each upstream station based on the river slope and river distance between each upstream station and the target monitoring station includes: Obtain the river slope and river distance between the upstream station and the target monitoring station; wherein, the upstream station is any upstream station of the target monitoring station; Based on the river slope, a slope influence factor is determined, wherein the slope influence factor increases as the slope increases; Based on the river channel distance, a distance influence factor is determined, wherein the distance influence factor decreases as the distance increases; The spatial location influence weight of the target monitoring station is obtained by weighted summation of the slope influence factor and the distance influence factor.

7. A watershed water resources scheduling method based on digital twins according to claim 5, characterized in that, The step of calculating the expected water level change characteristic value of the target monitoring station based on the current water level change characteristic value of each upstream station and its corresponding comprehensive influence weight includes: Obtain the current water level change characteristic value of each of the upstream stations during the current monitoring period; The current water level change characteristic value of each upstream station is multiplied by its corresponding comprehensive influence weight to obtain the weighted contribution value of each upstream station. The weighted contribution values ​​of all upstream stations are summed to obtain the expected water level change characteristic value of the target monitoring station.

8. A watershed water resources scheduling method based on digital twins according to claim 5, characterized in that, The step of calculating the second confidence level based on the actual water level change characteristic value of the target monitoring station and the expected water level change characteristic value includes: Determine whether the actual water level change characteristic value and the expected water level change characteristic value change in the same direction; If the directions of change are inconsistent, the second confidence level is set to the preset minimum confidence level value; If the directions of change are consistent, the degree of difference between the actual water level change characteristic value and the expected water level change characteristic value is calculated, and the second confidence level is calculated based on the degree of difference, wherein the greater the degree of difference, the lower the second confidence level.

9. A watershed water resources scheduling method based on digital twins according to any one of claims 1 to 8, characterized in that, The step of inputting the target confidence level and the real-time water level data into a preset digital twin scheduling model and outputting water resource scheduling instructions includes: Based on the target confidence level, the real-time water level data is divided into core input data and auxiliary input data; wherein, the core input data is used to directly drive scheduling decisions, and the auxiliary input data is used to correct parameters within the scheduling decision framework; Based on the core input data, the macro-operational range of at least one reservoir within the preset watershed is determined; wherein, the macro-operational range is determined according to the real-time water level and preset scheduling rules, and the macro-operational range includes a flood discharge range, a water storage range and a water supply range; Within the defined macroscopic operating range, the scheduling operation parameters are optimized and calculated in conjunction with the auxiliary input data to generate the water resource scheduling instruction; wherein, the scheduling operation parameters include at least one of flood discharge flow, water storage target water level, or water supply flow.

10. A watershed water resources scheduling system based on digital twins, characterized in that, The system includes: The acquisition module is used to acquire the spatial topology of multiple monitoring stations within a preset watershed, as well as the real-time water level data of a target monitoring station; wherein, the target monitoring station is any monitoring station within the preset watershed. The first calculation module is used to calculate a corresponding first confidence level based on the historical water level data of the target monitoring station; wherein, the first confidence level is the confidence level determined by the target monitoring station based on the changing pattern of the historical water level data; The second calculation module is used to calculate a corresponding second confidence level based on the spatial relationship between the target monitoring station and other stations in the spatial topology; wherein, the second confidence level is a confidence level determined based on the coordination between the target monitoring station and adjacent stations in the spatial topology; The third calculation module is used to perform a fusion calculation based on the first confidence level and the second confidence level to obtain the target confidence level of the target monitoring station; The scheduling module is used to input the target confidence level and the real-time water level data into a preset digital twin scheduling model and output water resource scheduling instructions; wherein, the water resource scheduling instructions are used to carry out water resource scheduling within the preset watershed.

Images