Reservoir group flood discharge scheme multi-objective optimization method and system based on digital twinning

Abstract

The invention discloses a reservoir group flood discharge scheme multi-objective optimization method and system based on digital twinning, particularly relates to the technical field of hydraulic engineering digital decision making, and is used for solving the problem of flood discharge scheme decision making failure caused by representation deviation generated by long-term operation of an existing digital twinning model. A model mismatching state is detected and recognized through the deviation of real-time monitoring data and model prediction data, when characterization deviation exists, a flood discharge efficiency transfer chain network is constructed based on historical flood discharge data, vulnerability parameters are calculated, then key nodes are recognized, and the downstream river flood discharge capacity attenuation trend is evaluated. The influence degree of siltation on the reservoir capacity is quantified through a sediment transportation and river evolution coupling model, finally, the reservoir capacity parameters of the digital twinborn model are dynamically corrected according to the influence degree, and multi-target optimization is executed again based on the updated model to generate a flood discharge scheme. And adaptive correction of the digital twin model and dynamic optimization of a flood discharge scheme are realized.

Description

Technical Field

[0001] This invention relates to the field of digital decision-making technology for water conservancy projects, and more specifically, to a multi-objective optimization method and system for flood discharge schemes of reservoir groups based on digital twins. Background Technology

[0002] The application of digital twin technology in flood control scheduling of reservoir groups has formed a typical pattern. It constructs a virtual model synchronized with the physical reservoir group, integrating real-time monitoring data and historical operational patterns to support the simulation and evaluation of flood discharge plans. Existing technical solutions typically rely on a fixed-period model parameter calibration mechanism, combined with multi-objective optimization algorithms to optimize decision variables such as flood discharge allocation and gate opening / closing sequence, in order to achieve a balance between flood control safety, water resource utilization, and ecological protection. This technical approach depends on the digital twin model's ability to continuously map the state of physical entities, and its effectiveness is based on the assumption that model parameters and entity characteristics remain consistent over a long period.

[0003] Existing digital twin models experience a gradual decline in their representational capabilities over long-term operation due to progressive physical changes such as reservoir siltation and facility aging. This mismatch between model aging and actual entity evolution is difficult to detect and correct in a timely manner through periodic calibration mechanisms. When the mapping of the digital twin model to the actual state of the reservoir group exhibits systematic deviations, multi-objective optimization of flood discharge schemes based on the model will suffer from inherent defects. The so-called "optimal solution" generated by the optimization algorithm under the guidance of the distorted model may lead to decision-making risks such as increased flood control risks or misallocation of water resources during actual implementation. This makes it difficult for existing technologies to maintain the effectiveness and reliability of flood discharge scheme optimization in a long-term dynamic operating environment. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, the present invention provides a multi-objective optimization method and system for flood discharge schemes of reservoir groups based on digital twins to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins includes the following steps:

[0007] S1. Obtain real-time reservoir group monitoring data and historical flood discharge plan data from the digital twin model;

[0008] S2. Based on real-time reservoir group monitoring data and prediction data from digital twin models, use a deviation detection algorithm to determine whether there is a representational bias in the digital twin model;

[0009] S3. When the digital twin model has a representation bias, a flood discharge efficiency transmission chain network is constructed by combining historical flood discharge scheme data, and the vulnerability parameters of the flood discharge efficiency transmission chain are calculated by complex network theory.

[0010] S4. Based on the vulnerability parameters of the flood discharge efficiency transmission chain, identify key nodes in the flood discharge efficiency transmission chain network and assess the decline trend of the downstream river channel flood discharge capacity corresponding to the key nodes.

[0011] S5. Based on the decreasing trend of downstream river channel flood discharge capacity, analyze the impact of sedimentation on reservoir capacity using a coupled model of sediment transport and river channel evolution.

[0012] S6. Adjust the reservoir capacity parameters of the digital twin model based on the degree of impact, and re-execute multi-objective optimization based on the adjusted digital twin model to generate a flood discharge plan.

[0013] Furthermore, real-time reservoir group monitoring data and historical flood discharge plan data from the digital twin model are obtained, including:

[0014] Acquire reservoir water level data, inflow data, and gate opening data from real-time reservoir group monitoring data;

[0015] Obtain the historical flood discharge operation command sequence and the corresponding downstream key section water level response sequence from the historical flood discharge plan data.

[0016] Furthermore, the historical flood discharge operation command sequence includes flood discharge flow control commands and gate opening and closing timing commands, and the downstream key section water level response sequence includes flood wave propagation process data corresponding to the time of the historical flood discharge operation commands.

[0017] Furthermore, based on real-time reservoir group monitoring data and the prediction data from the digital twin model, a bias detection algorithm is used to determine whether the digital twin model has representational bias, including:

[0018] Extract reservoir water level data and inflow data from real-time reservoir group monitoring data;

[0019] The reservoir water level data and inflow data are compared point by point with the water level prediction data and flow prediction data at the corresponding time in the digital twin model to generate water level residual sequence and flow residual sequence.

[0020] The sliding window T-test algorithm was applied to the water level residual series and the flow residual series to detect whether there was a statistically significant trend of continuous deviation from zero in the residual series;

[0021] When at least one residual sequence is determined to have a statistically significant trend, it is confirmed that the digital twin model has a representational bias.

[0022] Furthermore, when the digital twin model exhibits representational bias, a flood discharge effectiveness transmission chain network is constructed by combining historical flood discharge scheme data. Vulnerability parameters of the flood discharge effectiveness transmission chain are calculated using complex network theory, including:

[0023] Based on the historical flood discharge operation command sequence and the corresponding downstream key section water level response sequence in the historical flood discharge plan data, the nodes and edges of the flood discharge efficiency transmission chain network are defined.

[0024] Based on the topology of the flood discharge efficiency transmission chain network, the importance index of nodes and the network connectivity index are calculated.

[0025] The vulnerability parameters of the flood discharge efficiency transmission chain are generated by integrating node importance indicators and network connectivity indicators.

[0026] Furthermore, the nodes include reservoir control points and downstream key sections, and the edges are constructed based on the correlation strength between historical flood discharge operation command sequences and downstream key section water level response sequences.

[0027] Furthermore, based on the vulnerability parameters of the flood discharge efficiency transmission chain, key nodes in the flood discharge efficiency transmission chain network are identified, and the decline trend of downstream river channel flood carrying capacity corresponding to the key nodes is assessed, including:

[0028] Based on the node importance index in the vulnerability parameters of the flood discharge efficiency transmission chain, key nodes in the flood discharge efficiency transmission chain network whose node importance index exceeds a preset threshold are selected.

[0029] For the selected key nodes, extract the corresponding downstream key section water level response sequences from the historical flood discharge plan data;

[0030] By analyzing the trends of flood wave propagation time and flood peak attenuation rate in the water level response sequence of key downstream sections, the trend of decline in the flood discharge capacity of the downstream river channel is generated.

[0031] Furthermore, based on the declining flood discharge capacity trend of downstream river channels, the impact of sedimentation on reservoir capacity is analyzed using a coupled model of sediment transport and river channel evolution, including:

[0032] The trends of flood wave propagation time and flood peak attenuation rate in the decline trend of downstream river channel flood carrying capacity are input into the coupled model of sediment transport and river channel evolution.

[0033] The parameters of river channel sediment transport capacity were calculated based on the trends of flood wave propagation time and flood peak attenuation rate.

[0034] The sedimentation development process in the reservoir area was simulated using a coupled model of sediment transport and channel evolution, under different parameters of sediment transport capacity variation.

[0035] The reservoir capacity loss rate is calculated based on the siltation development process, thus generating the degree of impact of siltation on the reservoir capacity.

[0036] Furthermore, the reservoir capacity parameters of the digital twin model are adjusted based on the degree of impact, and multi-objective optimization is re-executed based on the adjusted digital twin model to generate a flood discharge plan, including:

[0037] Based on the reservoir capacity loss rate in the degree of impact of siltation on reservoir capacity, the parameters of the reservoir capacity curve in the digital twin model are dynamically corrected;

[0038] In the adjusted digital twin model, flood control safety, water resource utilization, and ecological protection objectives are set as optimization objectives for multi-objective optimization.

[0039] Based on the reservoir water level data and inflow data in the real-time reservoir group monitoring data, the optimal solution set that satisfies each optimization objective is solved by a multi-objective evolutionary algorithm;

[0040] Select the flood discharge flow allocation scheme and the gate opening and closing sequence scheme from the optimal solution set to generate the final flood discharge scheme.

[0041] On the other hand, the present invention provides a multi-objective optimization system for flood discharge schemes of reservoir groups based on digital twins, comprising the following modules:

[0042] The data acquisition module is used to acquire real-time reservoir group monitoring data and historical flood discharge plan data from the digital twin model;

[0043] The deviation judgment module is used to determine whether there is a representational deviation in the digital twin model based on real-time reservoir group monitoring data and prediction data from the digital twin model.

[0044] The parameter calculation module is used to construct a flood discharge efficiency transmission chain network by combining historical flood discharge scheme data when there is a representation bias in the digital twin model, and to calculate the vulnerability parameters of the flood discharge efficiency transmission chain through complex network theory.

[0045] The trend assessment module is used to identify key nodes in the flood discharge efficiency transmission chain network based on the vulnerability parameters of the flood discharge efficiency transmission chain, and to assess the decline trend of the downstream river channel flood discharge capacity corresponding to the key nodes.

[0046] The impact analysis module is used to analyze the degree of impact of sedimentation on reservoir capacity based on the decline trend of downstream river flood control capacity, through a coupled model of sediment transport and river evolution.

[0047] The scheme generation module is used to adjust the reservoir capacity parameters of the digital twin model based on the degree of impact, and re-execute multi-objective optimization based on the adjusted digital twin model to generate a flood discharge scheme.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] This invention achieves the self-correction capability of the digital twin model during long-term operation by constructing a dual verification mechanism of flood discharge efficiency transmission chain network and sediment transport coupling analysis. By detecting the deviation between real-time monitoring data and model prediction data, it actively identifies the representational deviation of the digital twin model, overcoming the lag limitation of traditional periodic calibration mechanisms. When model deviation is detected, the flood discharge efficiency transmission chain network constructed by the system based on historical flood discharge scheme data can accurately capture the key vulnerable links within the reservoir group system. By quantitatively analyzing the transmission characteristics of flood discharge efficiency through complex network theory, it can effectively identify the system performance degradation caused by gradual changes such as river siltation and facility aging, providing precise guidance for model parameter correction.

[0050] 2. This invention further transforms the decline trend of downstream river flood discharge capacity into a quantitative assessment of its impact on reservoir capacity by using a coupled model of sediment transport and river evolution. It establishes a technical path from phenomenon monitoring to essential analysis. Based on the assessment results, the reservoir capacity parameters of the digital twin model are dynamically corrected, ensuring the consistency between the virtual model and the physical entity. Multi-objective optimization based on the corrected digital twin model can generate the optimal flood discharge scheme that simultaneously meets the objectives of flood control safety, water resource utilization, and ecological protection. This effectively solves the decision-making bias problem caused by model aging and significantly improves the adaptability and reliability of the flood discharge scheme in long-term operating environments. Attached Figure Description

[0051] Figure 1 This is a flowchart of a multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins, according to the present invention.

[0052] Figure 2 This is a schematic diagram of the structure of a multi-objective optimization system for flood discharge schemes of reservoir groups based on digital twins, according to the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0054] Example 1: Figure 1 This invention presents a multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins, which includes the following steps:

[0055] S1. Obtain real-time reservoir group monitoring data and historical flood discharge plan data from the digital twin model;

[0056] S2. Based on real-time reservoir group monitoring data and prediction data from digital twin models, use a deviation detection algorithm to determine whether there is a representational bias in the digital twin model;

[0057] S3. When the digital twin model has a representation bias, a flood discharge efficiency transmission chain network is constructed by combining historical flood discharge scheme data, and the vulnerability parameters of the flood discharge efficiency transmission chain are calculated by complex network theory.

[0058] S4. Based on the vulnerability parameters of the flood discharge efficiency transmission chain, identify key nodes in the flood discharge efficiency transmission chain network and assess the decline trend of the downstream river channel flood discharge capacity corresponding to the key nodes.

[0059] S5. Based on the decreasing trend of downstream river channel flood discharge capacity, analyze the impact of sedimentation on reservoir capacity using a coupled model of sediment transport and river channel evolution.

[0060] S6. Adjust the reservoir capacity parameters of the digital twin model based on the degree of impact, and re-execute multi-objective optimization based on the adjusted digital twin model to generate a flood discharge plan.

[0061] S1. Obtain real-time reservoir group monitoring data and historical flood discharge plan data from the digital twin model. The specific implementation is as follows:

[0062] Real-time reservoir group monitoring data is acquired through monitoring equipment deployed at various key locations within the reservoir group. Reservoir water level data is measured using pressure level gauges installed in front of the reservoir dam or at representative locations within the reservoir area. The measurement principle is based on the direct proportionality between hydrostatic pressure and water depth. Measurement data is measured in meters, and the data acquisition frequency is set to once per minute to ensure real-time reflection of dynamic changes in reservoir water levels. Data output is transmitted to a data acquisition unit via analog or digital signals. The data acquisition unit performs analog-to-digital conversion on the signals and adds timestamps. Then, it is uploaded in real-time to the data processing center of the digital twin model via wireless communication networks such as 4G or 5G networks. The data processing center performs preliminary verification of the received reservoir water level data. Verification methods include range checks, i.e., confirming that the water level data is within a reasonable range, such as between the dead water level and the check flood level, and consistency checks, i.e., comparing the rate of change of data at adjacent time points to see if it is within the physically possible range, in order to avoid sensor failure or transmission errors. Inflow data is acquired using ultrasonic flow meters installed at the inflow section of the reservoir. These flow meters employ the time-of-flight measurement principle, calculating the flow velocity by measuring the time difference between upstream and downstream ultrasonic propagation, and then calculating the flow rate by combining this with the cross-sectional area of ​​the river channel. The flow rate is measured in cubic meters per second. The data acquisition frequency is synchronized with the water level data, set to once every minute to ensure data consistency. The output signal of the ultrasonic flow meter is processed by a signal conditioning circuit to eliminate noise interference, and then transmitted to the data acquisition system via wired or wireless means. The data acquisition system stores the flow data in real time and marks the measurement time. The data interface module of the digital twin model periodically calls these data for model updates and deviation detection. Gate opening data is acquired through a rotary encoder or potentiometer-type position sensor in the gate control system. The sensor is installed on the drive shaft of the gate hoist and measures the actual opening position of the gate. The opening data is expressed as a percentage, where 0% corresponds to the gate being fully closed and 100% corresponds to the gate being fully open. The data acquisition frequency is set to once per minute, the same as the water level and flow data, to ensure that all real-time monitoring data are aligned in time. The output signal of the position sensor is acquired by a PLC programmable logic controller. The PLC converts the opening data into a standard communication protocol such as Modbus TCP, and then transmits it to the data server of the digital twin model via industrial Ethernet. The data server unifies the format of the gate opening data and stores it, providing input for subsequent deviation detection.

[0063] Historical flood discharge plan data is obtained from the historical database of the reservoir management information system. The historical database uses a relational database management system such as MySQL or Oracle to store structured data. The data tables include a flood discharge event table, an operation instruction table, and a monitoring response table. The flood discharge event table records the basic information of each flood discharge, such as the event number, start time, and end time. The operation instruction table stores the historical flood discharge operation instruction sequence, and the monitoring response table stores the water level response sequence of downstream key sections. The historical flood discharge operation instruction sequence is obtained by querying the operation instruction table. Each instruction entry in the sequence includes the instruction type, instruction parameters, and instruction timestamp. The instruction type is divided into flood discharge flow control instructions and gate opening and closing sequence instructions. Flood discharge flow control instructions include the target flow value, in cubic meters per second. For example, in a flood discharge event, the flood discharge flow control instruction may specify a target flow of 120 cubic meters per second. Gate opening and closing sequence instructions include the gate number, operation type (open or closed), target opening percentage, and execution time. For example, first, an opening instruction is issued for gate number one with a target opening percentage of 60%, and the execution time is 0 minutes after the start of the event. Then, an opening instruction is issued for gate number two with a target opening percentage of 80%, and the execution time is 15 minutes after the start of the event. The downstream key section water level response sequence is obtained by querying the monitoring response table. Each response entry in the sequence includes the section number, water level value, and water level timestamp. The water level value is in meters. The acquisition frequency is set according to the characteristics of the flood event. For example, during flood discharge, the acquisition frequency is increased to once every 5 minutes to capture flood wave propagation process data. Flood wave propagation process data includes flood wave arrival time, peak water level, and receding time. For example, in a flood discharge event, the downstream key section water level begins to rise 20 minutes after the start of flood discharge, the peak water level reaches 3.5 meters at 50 minutes, and then falls back to the baseline water level at 120 minutes. The correlation between the historical flood discharge operation command sequence and the downstream key section water level response sequence is achieved through time alignment. The data processing module of the digital twin model uses a timestamp matching algorithm to align the execution time of each operation command with the timestamp of the downstream water level response. The alignment method is based on a sliding time window with a window size of 5 minutes. That is, water level data within 5 minutes before and after the command execution time are regarded as the corresponding response, so as to ensure that the causal relationship between the command and the response is accurately mapped, providing a foundation for the subsequent construction of the flood discharge efficiency transmission chain network.

[0064] The flood discharge flow control instructions in the historical flood discharge operation instruction sequence specifically define the flow target to be maintained during the flood discharge process. The flow target is set according to the reservoir scheduling rules and real-time hydrological conditions. For example, in flood control scheduling, the flood discharge flow control instructions may be based on the inflow forecast and the downstream safe discharge calculation. The instruction parameters include the flow value and duration. The flow value is in cubic meters per second, and the duration is in minutes. For example, the flood discharge flow control instruction may be set to maintain a flow of 100 cubic meters per second for 30 minutes. The gate opening and closing sequence instructions specify the order and timing of gate operations. The instruction parameters include the gate identifier, the operation action of opening or closing, the target opening percentage, and the planned execution time. The target opening percentage is calculated by back-calculating the gate flow curve based on the flow control requirements. The gate flow curve is obtained through hydraulic model tests or historical data regression analysis. For example, for an arc gate, an opening of 50% may correspond to a flow of 60 cubic meters per second. The planned execution time is set based on flood forecasts and operation delays to ensure that the instructions are executed at the expected time. The flood wave propagation process data in the downstream key section water level response sequence describes the dynamic characteristics of the flood wave in the downstream river channel. Data parameters include flood wave initiation time, peak time, peak water level, and wave drop time. Flood wave initiation time refers to the moment when the water level begins to rise significantly; peak time refers to the moment when the water level reaches its highest point; peak water level refers to the water level value at the peak; and wave drop time refers to the moment when the water level falls back to the baseflow. These parameters are extracted by analyzing the water level time series curve. The extraction method uses moving average filtering to eliminate noise and then identifies inflection points. For example, a water level change rate exceeding 0.001 meters per second is considered the initiation point, and a water level change rate turning negative is considered the peak point. The historical flood discharge operation command sequence and the downstream key section water level response sequence are stored in JSON or XML structured data formats to facilitate parsing and processing by the digital twin model. Data integrity is ensured through checksum mechanisms, such as adding a CRC cyclic redundancy check code to each data packet. Data transmission is encrypted using security protocols such as HTTPS to prevent data tampering or loss.

[0065] S2. Based on real-time reservoir group monitoring data and prediction data from the digital twin model, a deviation detection algorithm is used to determine whether there is a representational bias in the digital twin model. Specifically, the implementation is as follows:

[0066] The process of extracting reservoir water level and inflow data from real-time reservoir monitoring data is completed using a data interface module based on a digital twin model. This module accesses a real-time database that stores raw data collected by monitoring equipment. Reservoir water level data is in meters, and inflow data is in cubic meters per second. The extraction method uses a time range query, with the query condition set to trace back a specified time length from the current point in time, such as 24 hours, to ensure the data covers a complete daily cycle. The extracted reservoir water level and inflow data are arranged in ascending order of timestamps, forming a time series. Each data point in the time series contains a timestamp and a measured value. The timestamp accuracy is 1 second, and the measured value is rounded to three decimal places. After extraction, the data is stored in an in-memory array for subsequent processing. Data integrity is confirmed by checking for missing values. If missing values ​​exist, linear interpolation is used to fill them in. Linear interpolation calculates missing values ​​based on the trend of adjacent data points.

[0067] The reservoir water level data and inflow data are compared point-by-point with the corresponding time-based predicted water level and flow rates from the digital twin model. The digital twin model's predicted data is generated through a simulation module. This module operates a hydraulic model based on historical hydrological data and real-time input, outputting predicted water level and flow rates. Water level predictions are in meters, and flow rates are in cubic meters per second. The timestamps of the predicted data are aligned with the timestamps of the real-time monitoring data using nearest neighbor matching. This means that the timestamp of each real-time data point is matched with the predicted point in the predicted data with the smallest time difference. The maximum allowed time difference for this matching is specified in the model's predictions. The time interval is set to 1 minute. Data points exceeding this threshold are considered invalid and excluded. Point-by-point comparison is achieved by calculating the difference between the real-time measured value and the predicted value. For reservoir water level data, the difference is calculated as the real-time water level value minus the predicted water level value. For inflow data, the difference is calculated as the real-time flow rate value minus the predicted flow rate value. The difference results generate water level residual sequences and flow residual sequences. Each residual point in the water level residual sequence and flow residual sequence contains a timestamp and a residual value. The residual values ​​are expressed in their original units, i.e., the unit for water level residuals is meters and the unit for flow residuals is cubic meters per second. The residual sequences are stored as a list structure and sorted in chronological order to provide input for subsequent statistical analysis.

[0068] A sliding window t-test algorithm is applied to the water level residual series and the flow residual series. Based on the Student's t-distribution principle, the sliding window t-test algorithm is used to test whether the mean of the residual series deviates significantly from zero within the sliding window. The size of the sliding window is set according to the data acquisition frequency and business requirements. For example, when the data acquisition frequency is once per minute, the window size is set to 1440 data points corresponding to a 24-hour period to capture daily-scale trends. The window sliding step is set to 60 data points corresponding to 1 hour to balance computational efficiency and trend detection sensitivity. For each sliding window, the algorithm calculates the sample mean and sample standard deviation of the residual series within the window. The sample mean is calculated as the arithmetic mean of all residual values ​​within the window, and the sample standard deviation is calculated as the ratio of the residual values ​​within the window to the standard deviation. The algorithm takes the square root of the sum of squares of the sample means and then calculates the t-statistic. The t-statistic is equal to the ratio of the sample mean divided by the sample standard deviation to the square root of the window size. The t-statistic is compared with the critical value of the t-distribution, which is determined based on the degrees of freedom and significance level. The degrees of freedom are equal to the window size minus 1, and the significance level is set to 0.05, corresponding to a 95% confidence level. The critical value is obtained by looking up the t-distribution table. For example, when the degrees of freedom are 1439, the critical value for a two-tailed test is approximately 1.96. If the absolute value of the calculated t-statistic exceeds the critical value, it is determined that there is a statistically significant trend in the residual sequence within the window, that is, the residual mean is significantly non-zero. The algorithm traverses the entire residual sequence, performs the above test on all windows, and records the test results for each window.

[0069] When at least one residual sequence is determined to have a statistically significant trend, the digital twin model is confirmed to have a representational bias. The judgment condition is based on the output of the sliding window T-test algorithm. If any one of the water level residual sequences or the flow residual sequences is detected to have a statistically significant trend within any window, the model bias confirmation is triggered, confirming it as a global bias rather than an instantaneous fluctuation. After bias confirmation, a bias flag is generated and stored in the state variables of the digital twin model, triggering subsequent processing procedures, such as model parameter calibration or warning notification. To ensure the reliability of the judgment, the algorithm also considers the persistence of the trend, for example, requiring that a significant trend be detected in multiple consecutive windows before the bias is confirmed. However, in this embodiment, a single-window triggering mechanism is used to simplify the processing. The bias confirmation logic is implemented through conditional statements, that is, checking the test result arrays of the water level residual sequence and the flow residual sequence. If either array contains a true value, a confirmation signal is returned.

[0070] S3. When the digital twin model exhibits representational bias, a flood discharge effectiveness transmission chain network is constructed by combining historical flood discharge scheme data. The vulnerability parameters of the flood discharge effectiveness transmission chain are calculated using complex network theory. The specific implementation is as follows:

[0071] When the digital twin model exhibits representational bias, the process of constructing a flood discharge efficiency transmission chain network by combining historical flood discharge scheme data is initiated. This construction process is based on the historical flood discharge operation command sequence and the corresponding downstream key section water level response sequence from the historical flood discharge scheme data. The historical flood discharge operation command sequence is extracted from the historical database of the reservoir management information system, containing flood discharge flow control commands and gate opening and closing timing commands. The downstream key section water level response sequence is obtained from historical data from downstream hydrological monitoring stations, containing flood wave propagation process data. When defining nodes in the flood discharge efficiency transmission chain network, nodes include reservoir control points and downstream key sections. Reservoir control points are identified by reservoir names or unique codes, such as Reservoir A and Reservoir B. Downstream key sections are identified by section names or geographical coordinates, such as Section X and Section Y. The node set is generated by traversing all participating entities in historical flood discharge events, ensuring that each reservoir control point and each downstream key section appears at least once in the historical data. When defining edges, the edges are based on historical flood discharge operation data. The correlation strength between the command sequence and the downstream key section water level response sequence is constructed. The correlation strength is obtained by calculating the statistical correlation between the two sequences, such as using the Pearson correlation coefficient. The Pearson correlation coefficient is calculated as the product of the covariance of the two sequences divided by their respective standard deviations. The covariance reflects the consistency of the sequence change trend, and the standard deviation measures the dispersion of the sequence. When calculating, the sequence data must be time aligned first. The alignment method adopts sliding time window matching, and the window size is set to 10 minutes. That is, the water level response data within 10 minutes before and after the execution time of the operation command is used for calculation. The correlation strength threshold is set to 0.7. This threshold is determined based on the distribution of correlation coefficients of all correlations in historical data. For example, the upper quartile of the distribution is taken to ensure that only strong correlations are included in the network. The direction of the edge is from the reservoir control point to the downstream key section, which represents the transmission path of the flood discharge impact. After the network is constructed, it is stored as a graph structure, where the node attributes include node type and geographical location, and the edge attributes include correlation strength value and historical frequency of occurrence.

[0072] Nodes include reservoir control points and downstream key sections. Reservoir control points correspond to actual reservoir engineering projects, such as gravity dams or arch dams. Downstream key sections correspond to fixed monitoring locations on the river channel, such as bridges or hydrological stations. Node definition must ensure uniqueness and avoid duplication, for example, through geographic information system coordinate verification. Edges are constructed based on the correlation strength between historical flood discharge operation command sequences and downstream key section water level response sequences. The correlation strength calculation uses a multi-index fusion method, considering not only the Pearson correlation coefficient but also the mutual information value of the sequences. The mutual information value measures the amount of shared information between two sequences. The probability distribution based on sequence values ​​is calculated, and the joint correlation is estimated using histograms. The probability and marginal probability, and the final value of the association strength are calculated as Pearson correlation coefficient × weight 1 + normalized mutual information × weight 2. Weight 1 and weight 2 are set according to the reliability of historical data. For example, weight 1 is 0.6 and weight 2 is 0.4. The weight values ​​are determined by expert experience and based on the importance assessment of different indicators in the reservoir's operation history. When constructing edges, an edge is added between corresponding nodes only when the association strength value exceeds the threshold of 0.7. The weight of the edge is set to the association strength value for subsequent network analysis. After the network is constructed, an integrity check is performed to ensure that all nodes are connected by at least one edge to avoid isolated nodes affecting network analysis.

[0073] Based on the topology of the flood discharge efficiency transmission chain network, node importance and network connectivity indices are calculated. Node importance is assessed using degree centrality and betweenness centrality. Degree centrality is calculated by dividing the number of edges connected to a node by the total number of edges in the network, reflecting the node's direct influence. Betweenness centrality is calculated by the frequency of a node's appearance in all shortest paths, reflecting the node's control over network traffic. Shortest paths are calculated using Dijkstra's algorithm, based on edge weights (i.e., association strength). Path length is defined as the sum of the reciprocals of the weights, with higher association strength resulting in shorter paths. When calculating node importance, degree centrality and betweenness centrality are normalized to the range of 0 to 1. The normalization method is to divide the current value by the maximum value in the network. The network connectivity index uses the average path length and the clustering coefficient. The average path length is calculated as the average of the shortest path lengths between all node pairs, reflecting the overall efficiency of the network. The clustering coefficient is calculated as the ratio of the actual number of edges between a node's neighbors to the number of possible edges, reflecting the local compactness of the network. When calculating the clustering coefficient, for each node, its set of neighbor nodes includes all directly connected nodes. The actual number of edges is obtained by counting the number of edges between neighbor nodes. The number of possible edges is the number of combinations of neighbor nodes, i.e., n×(n-1) / 2, where n is the number of neighbor nodes. The network connectivity index is also normalized to the range of 0 to 1 to facilitate subsequent fusion.

[0074] The vulnerability parameters of the flood discharge efficiency transmission chain are generated by fusing node importance indicators and network connectivity indicators. The fusion method adopts a weighted linear combination, with the weight of the node importance indicator set to 0.6 and the weight of the network connectivity indicator set to 0.4. The weight values ​​are set based on the operating characteristics of the reservoir group. Historical event analysis shows that node importance contributes more to vulnerability. The vulnerability parameter is calculated as the weighted sum of node importance indicators × the weighted sum of network connectivity indicators. The weighted sum of node importance indicators = degree centrality × 0.5 + betweenness centrality × 0.5. The weighted sum of network connectivity indicators = 0.5 × 1 - 0.5 × average path length + 0.5 × clustering coefficient. The average path length is taken as the reciprocal to maintain a positive correlation with vulnerability, that is, the shorter the path, the higher the vulnerability. The clustering coefficient is directly used to reflect the enhancement of vulnerability by local compactness. The vulnerability parameter value ranges from 0 to 1. The higher the value, the more vulnerable the flood discharge efficiency transmission chain. After the parameter is calculated, it is stored as a network attribute and used for subsequent risk assessment. The fusion process ensures that all indicators have consistent dimensions, and unit differences are eliminated through normalization.

[0075] S4. Based on the vulnerability parameters of the flood discharge efficiency transmission chain, identify key nodes in the flood discharge efficiency transmission chain network, and assess the decline trend of the downstream river channel flood carrying capacity corresponding to the key nodes. The specific implementation is as follows:

[0076] Based on the node importance index in the vulnerability parameters of the flood discharge efficiency transmission chain, key nodes in the flood discharge efficiency transmission chain network whose node importance index exceeds a preset threshold are screened. The node importance index is derived from the calculation results of the vulnerability parameters of the flood discharge efficiency transmission chain, including degree centrality and betweenness centrality. Degree centrality reflects the proportion of nodes with direct connections, while betweenness centrality reflects the mediating role of nodes in the shortest path of the network. The preset threshold is set based on the statistical distribution of the node importance index values ​​in historical data. For example, the 80th percentile of each value of degree centrality and betweenness centrality is taken as the threshold. The 80th percentile is calculated by taking the value of the 80th percentile after sorting all node index values ​​in ascending order, ensuring that only nodes with higher importance are selected as key nodes. The screening process is implemented by traversing all nodes in the flood discharge efficiency transmission chain network. For each node, it is checked whether at least one of its degree centrality and betweenness centrality exceeds the corresponding threshold. If the condition is met, the node is marked as a key node. The list of key nodes is stored as an array structure, containing node identifiers and corresponding node importance index values, for subsequent processing.

[0077] For the selected key nodes, the corresponding downstream key section water level response sequences were extracted from the historical flood discharge plan data. The historical flood discharge plan data was obtained from the historical database of the reservoir management information system and includes historical flood discharge operation instruction sequences and downstream key section water level response sequences. Key nodes correspond to downstream key sections. Therefore, during extraction, the section number in the historical data was matched according to the identifier of the key node. The matching method adopted was exact string comparison or geographic coordinate matching to ensure that each key node corresponds to a unique downstream key section water level response sequence. The water level response sequence data includes timestamps and water level values. The water level values ​​are in meters, and the timestamp accuracy is 1 minute. The extracted data covers multiple historical flood discharge events to ensure the integrity and representativeness of the sequence. After extraction, the data is stored in time series format. Each sequence is arranged in chronological order and labeled with the corresponding flood discharge event number to provide a basis for subsequent trend analysis.

[0078] The trends of flood wave propagation time and peak attenuation rate in the water level response sequence of downstream key sections were analyzed. Flood wave propagation time was calculated as the time difference between the execution of the flood discharge operation command and the moment when the water level at the downstream key section began to rise significantly, in minutes. The moment when the water level began to rise significantly was determined by analyzing the derivative of the water level sequence, i.e., the moment when the rate of change in water level exceeded 0.001 meters per minute. Peak attenuation rate was calculated as the ratio of the peak water level to the initial water level, where the initial water level was the stable water level before the start of flood discharge, and the peak water level was the maximum value in the water level sequence. The trend of propagation time variation was obtained by calculating the slope of linear regression of propagation time in multiple historical events. The linear regression used the least squares method, with the independent variable being the event occurrence time and the dependent variable being the propagation time. The slope value reflects the increasing or decreasing trend of propagation time over time. The trend of flood peak attenuation rate variation was also calculated using linear regression, with the independent variable being the event occurrence time and the dependent variable being the flood peak attenuation rate. The slope value reflects the change of attenuation rate over time. In the trend analysis, data preprocessing included removing outliers, which were defined as data points exceeding three standard deviations of the mean, to ensure the reliability of the regression results.

[0079] The generation of downstream river channel flood capacity decay trend is based on a comprehensive assessment of the trends in flood wave propagation time and peak attenuation rate. The flood capacity decay trend is defined as the weighted sum of the slopes of the propagation time and attenuation rate trends. The weights are set according to the relative importance of the two trends to flood capacity: the propagation time trend has a weight of 0.6, and the attenuation rate trend has a weight of 0.4. These weights are determined through expert consultation and historical event analysis. Based on the negative impact of prolonged propagation time and increased peak attenuation on flood capacity, the weighted sum is calculated as the propagation time trend slope multiplied by 0.6 plus the peak attenuation rate trend slope multiplied by 0.4. The resulting value is called the flood capacity decay index. A positive index value indicates an intensified flood capacity decay trend, while a negative value indicates improvement. The final output of the flood capacity decay trend is time series data, including the decay index and corresponding confidence interval for each key node. The confidence interval is calculated using the standard error of regression analysis, with a confidence level set at 95% to assess the statistical significance of the trend. The generated data is stored in a structured format for subsequent risk assessment and model adjustment.

[0080] S5. Based on the declining trend of downstream river channel flood control capacity, analyze the impact of sedimentation on reservoir capacity using a coupled model of sediment transport and river channel evolution. The specific implementation is as follows:

[0081] The trends of flood wave propagation time and flood peak attenuation rate in the downstream river channel's flood capacity decay trend are input into the sediment transport and river channel evolution coupling model. The flood wave propagation time trend is derived from the downstream river channel flood capacity decay trend analysis results and is expressed as a slope representing the rate of change of propagation time over years, in minutes per year. The flood peak attenuation rate trend is also expressed as a slope representing the rate of change of attenuation rate over years, in dimensionless years. The input process is implemented through the data interface module, which converts the trend data into a model-readable format, such as CSV or JSON, ensuring consistency of the time base during conversion. The Gregorian calendar year is used as the time axis. The model input parameters include the trend slope and initial conditions. The initial conditions are set based on the current river channel topography and hydrological data, such as the initial riverbed elevation and sediment particle size distribution. The sediment transport and river channel evolution coupling model uses the control volume method to discretize the river channel, dividing the river channel into multiple calculation units. The length of each unit is set according to the river channel morphology, for example, the unit length of a straight section is 100 meters, and the unit length of a curved section is 50 meters. The model running time step is set to 1 day to balance calculation accuracy and efficiency.

[0082] The parameters for the river channel's sediment transport capacity were calculated based on the trends in flood wave propagation time and flood peak attenuation rate. Empirical formulas were used in the calculations. The slope of the flood wave propagation time trend is correlated with the sediment-carrying capacity coefficient, which reflects the water flow's ability to carry sediment. These formulas were established through regression analysis of historical data. For example, for every 0.1 minute increase in the slope of the propagation time trend per year, the sediment-carrying capacity coefficient decreases by 0.05. The slope of the flood peak attenuation rate trend is correlated with the sediment settling rate, which reflects the sediment's settling in still water. The velocity and relationship are calibrated using laboratory test data. For example, for every 0.01 increase in the slope of the decay rate trend per year, the sediment deposition rate increases by 0.001 meters per second. The parameters for the change in river sediment transport capacity are ultimately calculated as a weighted combination of the flow sediment carrying capacity coefficient and the sediment deposition rate. The weights are set according to the contribution of the parameters to the transport capacity. The flow sediment carrying capacity coefficient has a weight of 0.7, and the sediment deposition rate has a weight of 0.3. The weight values ​​are determined through sediment dynamics principles and expert experience. After the parameters are calculated, they are used as input variables for the model to adjust the sediment transport equation in the model.

[0083] A coupled model of sediment transport and channel evolution was used to simulate the sedimentation development process in a reservoir area under different sediment transport capacity variation parameters. The simulation process was based on solving the sediment continuity equation and the motion equation using the finite difference method. The sediment continuity equation describes the conservation of sediment mass, while the motion equation describes the sediment transport rate. The initial conditions used current reservoir topographic survey data, such as the reservoir bottom elevation distribution obtained through a multibeam echo sounder. The boundary conditions included inflow sediment concentration and flow conditions. The inflow sediment concentration was obtained from upstream hydrological station monitoring data, in kilograms per cubic meter (kg / m³), and the flow conditions used the design flood hydrograph. The simulation duration was set to 10 years to cover medium- and long-term sedimentation effects. For each sediment transport capacity variation parameter scenario, the model calculated the change in sedimentation thickness at each point in the reservoir area over time. The sedimentation thickness was obtained by dividing the sediment deposition amount by the dry density of the sediment, which was taken as 1600 kg / m³. The simulation output included a sedimentation thickness distribution map and a sedimentation volume time series. The sedimentation volume was calculated by integrating the sedimentation thickness of the reservoir area, in cubic meters.

[0084] The reservoir capacity loss rate is calculated based on the siltation development process to generate the degree of impact of siltation on the reservoir capacity. The reservoir capacity loss rate is calculated as the ratio of siltation volume to the original reservoir capacity. The original reservoir capacity is obtained from reservoir design data, such as the capacity value at the corresponding water level obtained from the reservoir characteristic curve. The siltation volume is extracted from the simulation output. The same water level benchmark is used in the calculation, such as the normal storage level. The capacity loss rate is expressed as a percentage. The degree of impact of siltation on the reservoir capacity is generated based on the time change trend of the capacity loss rate. The degree of impact is divided into levels. For example, a loss rate of less than 5% is a slight impact, 5% to 10% is a moderate impact, and greater than 10% is a severe impact. The level thresholds are determined by reservoir operation specifications and historical cases. The impact degree output includes numerical and descriptive ratings for subsequent decision support. The calculation process is automated through scripts to ensure that the results are repeatable and verifiable.

[0085] S6. Adjust the reservoir capacity parameters of the digital twin model based on the degree of impact, and re-execute multi-objective optimization based on the adjusted digital twin model to generate a flood discharge plan. The specific implementation is as follows:

[0086] Based on the reservoir capacity loss rate, which is part of the impact of siltation on reservoir capacity, the parameters of the reservoir capacity curve in the digital twin model are dynamically corrected. The reservoir capacity loss rate is derived from the analysis results of the impact of siltation on reservoir capacity and is expressed as a percentage of the reduction in current capacity relative to the original capacity. The reservoir capacity curve parameters define the relationship between the reservoir water level and the corresponding capacity, and are represented by a polynomial function, such as a quadratic function. The parameters include a constant term, a linear coefficient, and a quadratic coefficient. The dynamic correction adjusts the reservoir capacity curve parameters based on the reservoir capacity loss rate. For example, when the reservoir capacity loss rate is 5%, the constant term parameter is reduced by 5% proportionally, and the linear and quadratic coefficients are adjusted accordingly to maintain the curve shape. The adjustment ratio is determined through regression analysis of historical data to ensure that the corrected reservoir capacity curve accurately reflects the current siltation status. The correction process is implemented through the parameter update module of the digital twin model. The parameter update module reads the reservoir capacity loss rate data, applies the correction algorithm to calculate the new parameter values, and writes the new parameter values ​​into the configuration file of the digital twin model, thus completing the dynamic update of the reservoir capacity curve parameters.

[0087] In the adjusted digital twin model, flood control safety, water resource utilization, and ecological protection objectives are set as optimization objectives for multi-objective optimization. The flood control safety objective is defined as the water level at key cross-sections of the downstream river channel not exceeding the safe water level threshold, which is obtained from river design specifications, such as determining it based on the elevation of the river embankment. The water resource utilization objective is defined as maintaining the reservoir water level within the suitable water supply range, which is determined by the reservoir scheduling diagram, such as setting upper and lower limits based on historical water supply data. The ecological protection objective is defined as the downstream discharge flow not being less than the ecological base flow, which is determined through hydrological and ecological studies, such as using 30% of the multi-year average flow as the minimum ecological requirement. The optimization objectives are expressed in the form of mathematical functions. The flood control safety objective function is calculated as the sum of squares of the deviations between the downstream water level and the safe water level threshold. The water resource utilization objective function is calculated as the absolute value of the deviation between the reservoir water level and the median of the suitable water supply range. The ecological protection objective function is calculated as the negative deviation penalty between the downstream discharge flow and the ecological base flow. Each objective function is normalized to the range of 0 to 1 to facilitate processing by the multi-objective optimization algorithm.

[0088] Based on real-time reservoir group monitoring data, including reservoir water level and inflow data, a multi-objective evolutionary algorithm is used to find the optimal solution set that satisfies each optimization objective. The real-time reservoir group monitoring data is obtained from the data interface of the digital twin model. The reservoir water level data is in meters, and the inflow data is in cubic meters per second. The multi-objective evolutionary algorithm adopts a non-dominated sorting genetic algorithm, which searches for the optimal solution by simulating the biological evolution process. The initial population size is set to 100 individuals, each individual representing a flood discharge scheme, including the flood discharge flow value and gate operation sequence. The number of algorithm iterations is set to 1000 generations to ensure convergence to the Pareto front. The crossover probability is set to 0.9, and the mutation probability is set to 0.1. Population diversity is achieved by randomly exchanging individual genes. The fitness function is calculated based on the values ​​of each optimization objective function. Non-dominated sorting and crowding distance are used to evaluate the quality of individuals. The final output is a non-dominated solution set, i.e., the optimal solution set, where each solution corresponds to a flood discharge scheme that satisfies the trade-offs of each optimization objective.

[0089] The flood discharge allocation scheme and the gate opening and closing sequence scheme are selected from the set of optimal solutions to generate the final flood discharge scheme. The selection process is based on a multi-attribute decision-making method, such as the ideal point method. The ideal point method sorts each solution by calculating the Euclidean distance between it and the ideal solution. The ideal solution consists of the optimal value of each objective function, such as minimizing the objective function value of flood control safety, water resource utilization, and ecological protection. The Euclidean distance is calculated as the square root of the sum of the squares of the differences between each objective function value and the ideal value. The solution with the smallest distance is selected as the final scheme. The flood discharge allocation scheme includes the flood discharge value of each reservoir, in cubic meters per second. The gate opening and closing sequence scheme includes the opening time, closing time, and opening percentage of each gate, with the time precision set to minutes. After the final flood discharge scheme is generated, it is exported as structured data in a digital twin model output interface for actual dispatch and command.

[0090] Example 2: Figure 2 A schematic diagram of a multi-objective optimization system for flood discharge schemes of reservoir groups based on digital twins is provided. This system includes the following modules:

[0091] The data acquisition module is used to acquire real-time reservoir group monitoring data and historical flood discharge plan data from the digital twin model;

[0092] The deviation judgment module is used to determine whether there is a representational deviation in the digital twin model based on real-time reservoir group monitoring data and prediction data from the digital twin model.

[0093] The parameter calculation module is used to construct a flood discharge efficiency transmission chain network by combining historical flood discharge scheme data when there is a representation bias in the digital twin model, and to calculate the vulnerability parameters of the flood discharge efficiency transmission chain through complex network theory.

[0094] The trend assessment module is used to identify key nodes in the flood discharge efficiency transmission chain network based on the vulnerability parameters of the flood discharge efficiency transmission chain, and to assess the decline trend of the downstream river channel flood discharge capacity corresponding to the key nodes.

[0095] The impact analysis module is used to analyze the degree of impact of sedimentation on reservoir capacity based on the decline trend of downstream river flood control capacity, through a coupled model of sediment transport and river evolution.

[0096] The scheme generation module is used to adjust the reservoir capacity parameters of the digital twin model based on the degree of impact, and re-execute multi-objective optimization based on the adjusted digital twin model to generate a flood discharge scheme.

[0097] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0098] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0099] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0100] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0101] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0102] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0103] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins, characterized in that, Includes the following steps: S1. Obtain real-time reservoir group monitoring data and historical flood discharge plan data from the digital twin model; S2. Based on real-time reservoir group monitoring data and prediction data from digital twin models, use a deviation detection algorithm to determine whether there is a representational bias in the digital twin model; S3. When the digital twin model has a representation bias, a flood discharge efficiency transmission chain network is constructed by combining historical flood discharge scheme data, and the vulnerability parameters of the flood discharge efficiency transmission chain are calculated by complex network theory. S4. Based on the vulnerability parameters of the flood discharge efficiency transmission chain, identify key nodes in the flood discharge efficiency transmission chain network and assess the decline trend of the downstream river channel flood discharge capacity corresponding to the key nodes. S5. Based on the decreasing trend of downstream river channel flood discharge capacity, analyze the impact of sedimentation on reservoir capacity using a coupled model of sediment transport and river channel evolution. S6. Adjust the reservoir capacity parameters of the digital twin model based on the degree of impact, and re-execute multi-objective optimization based on the adjusted digital twin model to generate a flood discharge plan.

2. The multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins according to claim 1, characterized in that, Acquire real-time reservoir group monitoring data and historical flood discharge plan data from the digital twin model, including: Acquire reservoir water level data, inflow data, and gate opening data from real-time reservoir group monitoring data; Obtain the historical flood discharge operation command sequence and the corresponding downstream key section water level response sequence from the historical flood discharge plan data.

3. The multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins according to claim 2, characterized in that, in, The historical flood discharge operation command sequence includes flood discharge flow control commands and gate opening and closing timing commands. The downstream key section water level response sequence includes flood wave propagation process data corresponding to the time of the historical flood discharge operation commands.

4. The multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins according to claim 1, characterized in that, Based on real-time reservoir group monitoring data and prediction data from digital twin models, a bias detection algorithm is used to determine whether the digital twin model exhibits representational bias, including: Extract reservoir water level data and inflow data from real-time reservoir group monitoring data; The reservoir water level data and inflow data are compared point by point with the water level prediction data and flow prediction data at the corresponding time in the digital twin model to generate water level residual sequence and flow residual sequence. The sliding window T-test algorithm was applied to the water level residual series and the flow residual series to detect whether there was a statistically significant trend of continuous deviation from zero in the residual series; When at least one residual sequence is determined to have a statistically significant trend, it is confirmed that the digital twin model has a representational bias.

5. The multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins according to claim 1, characterized in that, When digital twin models exhibit representational bias, a flood discharge effectiveness transmission chain network is constructed by combining historical flood discharge scheme data. Vulnerability parameters of the flood discharge effectiveness transmission chain are calculated using complex network theory, including: Based on the historical flood discharge operation command sequence and the corresponding downstream key section water level response sequence in the historical flood discharge plan data, the nodes and edges of the flood discharge efficiency transmission chain network are defined. Based on the topology of the flood discharge efficiency transmission chain network, the importance index of nodes and the network connectivity index are calculated. The vulnerability parameters of the flood discharge efficiency transmission chain are generated by integrating node importance indicators and network connectivity indicators.

6. The multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins according to claim 5, characterized in that, The nodes include reservoir control points and downstream key sections, and the edges are constructed based on the correlation strength between historical flood discharge operation command sequences and downstream key section water level response sequences.

7. The multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins according to claim 1, characterized in that, Based on the vulnerability parameters of the flood discharge efficiency transmission chain, key nodes in the flood discharge efficiency transmission chain network are identified, and the decline trend of downstream river channel flood carrying capacity corresponding to the key nodes is assessed, including: Based on the node importance index in the vulnerability parameters of the flood discharge efficiency transmission chain, key nodes in the flood discharge efficiency transmission chain network whose node importance index exceeds a preset threshold are selected. For the selected key nodes, extract the corresponding downstream key section water level response sequences from the historical flood discharge plan data; By analyzing the trends of flood wave propagation time and flood peak attenuation rate in the water level response sequence of key downstream sections, the trend of decline in the flood discharge capacity of the downstream river channel is generated.

8. The multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins according to claim 1, characterized in that, Based on the declining flood discharge capacity trend of downstream river channels, the impact of sedimentation on reservoir capacity is analyzed using a coupled model of sediment transport and river channel evolution, including: The trends of flood wave propagation time and flood peak attenuation rate in the decline trend of downstream river channel flood carrying capacity are input into the coupled model of sediment transport and river channel evolution. The parameters of river channel sediment transport capacity were calculated based on the trends of flood wave propagation time and flood peak attenuation rate. The sedimentation development process in the reservoir area was simulated using a coupled model of sediment transport and channel evolution, under different parameters of sediment transport capacity variation. The reservoir capacity loss rate is calculated based on the siltation development process, thus generating the degree of impact of siltation on the reservoir capacity.

9. A multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins according to claim 1, characterized in that, The reservoir capacity parameters of the digital twin model are adjusted based on the degree of impact, and multi-objective optimization is re-executed based on the adjusted digital twin model to generate a flood discharge plan, including: Based on the reservoir capacity loss rate in the degree of impact of siltation on reservoir capacity, the parameters of the reservoir capacity curve in the digital twin model are dynamically corrected; In the adjusted digital twin model, flood control safety, water resource utilization, and ecological protection objectives are set as optimization objectives for multi-objective optimization. Based on the reservoir water level data and inflow data in the real-time reservoir group monitoring data, the optimal solution set that satisfies each optimization objective is solved by a multi-objective evolutionary algorithm; Select the flood discharge flow allocation scheme and the gate opening and closing sequence scheme from the optimal solution set to generate the final flood discharge scheme.

10. A multi-objective optimization system for flood discharge schemes of reservoir groups based on digital twins, used to implement the multi-objective optimization method for flood discharge schemes of reservoir groups based on digital twins as described in any one of claims 1-9, characterized in that, Includes the following modules: The data acquisition module is used to acquire real-time reservoir group monitoring data and historical flood discharge plan data from the digital twin model; The deviation judgment module is used to determine whether there is a representational deviation in the digital twin model based on real-time reservoir group monitoring data and prediction data from the digital twin model. The parameter calculation module is used to construct a flood discharge efficiency transmission chain network by combining historical flood discharge scheme data when there is a representation bias in the digital twin model, and to calculate the vulnerability parameters of the flood discharge efficiency transmission chain through complex network theory. The trend assessment module is used to identify key nodes in the flood discharge efficiency transmission chain network based on the vulnerability parameters of the flood discharge efficiency transmission chain, and to assess the decline trend of the downstream river channel flood discharge capacity corresponding to the key nodes. The impact analysis module is used to analyze the degree of impact of sedimentation on reservoir capacity based on the decline trend of downstream river flood control capacity, through a coupled model of sediment transport and river evolution. The scheme generation module is used to adjust the reservoir capacity parameters of the digital twin model based on the degree of impact, and re-execute multi-objective optimization based on the adjusted digital twin model to generate a flood discharge scheme.

Images