A reservoir flood control monitoring system and method based on digital twin

Abstract

The present invention discloses a reservoir flood control monitoring system and method based on digital twin, which relates to the technical field of reservoir flood control monitoring. It includes a multi-source data acquisition module. The digital twin module is used to construct a digital twin model according to the real-time monitoring data of the multi-source data acquisition module. The coupling module is used to receive the fusion data set provided by the digital twin module and perform dynamic simulation on the fusion data set to provide dynamic prediction results of dam seepage, deformation and flood evolution. The dynamic warning module includes a dynamic threshold adjustment unit and a warning unit; the warning threshold is dynamically optimized according to the real-time monitoring data, and the warning is triggered in real time. Compared with the traditional warning system with a fixed threshold, this system can adapt to the changes of the dam state and environmental conditions, identify potential abnormal events in advance, provide sufficient time for emergency response and dispatching decision-making for reservoir flood control, thereby effectively reducing the disaster risk and ensuring the safety of reservoir operation.

Description

Technical Field

[0001] The present invention relates to the technical field of reservoir flood control monitoring, and more specifically, to a reservoir flood control monitoring system and method based on digital twin. Background Technique

[0002] A reservoir is one of the flood control engineering facilities. Building a reservoir at an appropriate position in the upper reaches of the flood control area can comprehensively utilize the reservoir to regulate floodwaters. The reservoir storage capacity is used to store floodwaters, reduce the peak flow entering the downstream river channel, and achieve the purpose of reducing flood disasters.

[0003] With the continuous development of reservoir flood control monitoring technology, in terms of the perception layer technology, the existing monitoring system constructs a three-dimensional monitoring network through sensors such as water level gauges, rain gauges, and piezometers, and combines satellite remote sensing and drone inspections to achieve full-basin coverage. The data transmission technology realizes real-time transmission through 4G / 5G, Beidou satellites, etc. Some systems introduce LoRa low-power wide-area networks to enhance the communication capabilities of edge nodes. The data analysis platform integrates GIS maps, BIM models, and Internet of Things data to provide functions such as water level prediction, seepage analysis, and dam deformation monitoring. Some systems introduce machine learning algorithms to optimize the warning threshold.

[0004] However, in the actual use process, existing sensors are difficult to accurately detect hidden cracks inside the dam body (such as concrete stress cracks) and deep-water leakage points. Relying on manual diving operations has low efficiency and high risks. Old sensors are prone to data drift under extreme weather conditions such as heavy rain and freezing, resulting in a water level monitoring error exceeding the critical value of ±5 cm. Rainfall, water level, and geological displacement data still adopt an independent analysis mode, lacking cross-parameter coupling correlation modeling.

[0005] Most systems only implement threshold alarms, lacking dynamic risk prediction based on historical data. When flood control scheduling conflicts with power generation and irrigation demands, the existing systems cannot provide multi-objective collaborative optimization solutions. The link from early warning to plan execution relies on manual decision-making, and the response delay exceeds 30 minutes in extreme cases. Summary of the Invention

[0006] To solve the above problems, the present invention provides a reservoir flood control monitoring system and method based on digital twin.

[0007] The present invention provides a reservoir flood control monitoring system based on digital twin, including the following modules:

[0008] Multi-source data acquisition module, which includes a quantum gyroscope osmotic pressure sensor, a lidar crack detector, and an underwater sonar array, deployed inside the dam body of the reservoir to collect multi-source data of the reservoir. The multi-source data includes the seepage pressure gradient data collected by the quantum gyroscope osmotic pressure sensor, the size data of the cracks inside the dam body obtained by the lidar crack detector, and the topographic data underground in the reservoir. The multi-source data is used as real-time monitoring data and transmitted.

[0009] Digital twin module, which is used to construct a digital twin model based on the real-time monitoring data of the multi-source data acquisition module, and output a fusion data set through the data twin model. The fusion data set includes the seepage pressure field distribution, crack expansion vector, and underwater topographic mutation hotspots.

[0010] Coupling module, which is used to receive the fusion data set provided by the digital twin module, perform dynamic simulation on the fusion data set, and provide dynamic prediction results of dam seepage, deformation, and flood evolution.

[0011] Dynamic warning module, which includes a dynamic threshold adjustment unit and a warning unit. The dynamic threshold adjustment unit is used to dynamically optimize the warning threshold according to the dynamic prediction results of the coupling module.

[0012] The warning unit is used to calculate the flood control risk value of the reservoir, and trigger a warning in real time according to the optimized warning threshold based on the flood control risk value to obtain a warning result.

[0013] Preferably, the specific working steps of the digital twin module are as follows:

[0014] Obtain the real-time monitoring data of the multi-source data acquisition module, and perform spatio-temporal alignment on the real-time monitoring data and the static data in the data floor.

[0015] Train the historical flood evolution data through a Transformer neural network to generate a data-driven model.

[0016] Then perform Kriging interpolation spatial registration on the fusion data to obtain a fusion data set.

[0017] Preferably, the specific working steps of the coupling module are as follows:

[0018] First, divide the dam body into element meshes through finite element discretization;

[0019] Then solve the coupled equation of the seepage field and the stress field;

[0020] Then calculate the cumulative effect of rainfall in a continuous week on the dam displacement based on the rainfall-deformation correlation matrix;

[0021] Finally, the two-dimensional shallow water equation is combined to simulate the flood evolution process, and the dynamic prediction results of dam seepage, deformation and flood evolution are obtained.

[0022] Preferably, the specific working steps for calculating the cumulative effect of rainfall in a continuous week on the dam displacement based on the rainfall-deformation correlation matrix are as follows:

[0023] According to the formula Calculate to obtain the Sensitivity of deformation within the detection band to rainfall in the area , where is the amount of deformation, sourced from the real-time data of the lidar crack detector, is the rainfall intensity, sourced from the real-time monitoring data of the meteorological station, is the attenuation coefficient, and is taken as the cumulative effect of rainfall in a continuous week on the dam displacement.

[0024] Preferably, the specific working steps for simulating the flood evolution process by combining the two-dimensional shallow water equation are as follows:

[0025] The simulation process is as follows: set the initial water depth and flow velocity, set the inflow, outflow and solid wall boundary conditions, select an appropriate time step according to the CFL condition, and gradually advance the time step through the finite volume method and Godunov format to calculate the flood evolution process;

[0026] Then fuse the real-time data of the underwater sonar with the historical flood pattern library, optimize the initial conditions through the data twin model, and introduce the source term correction driven by the fused data set in the shallow water equation.

[0027] Preferably, the specific working steps of the dynamic warning module are as follows:

[0028] First, train a Bayesian network with historical data to calculate the probability of abnormal events occurring;

[0029] Then predict the future parameter change trend based on time series data.

[0030] Finally, dynamically adjust the abnormal threshold according to the predicted change trend.

[0031] Preferably, the specific steps for dynamically adjusting the abnormal threshold according to the predicted change trend are as follows:

[0032] Set the initial threshold according to historical data or business requirements as the benchmark for adjustment;

[0033] Through the prediction model, obtain the predicted value at the next moment, and compare it with the current measured value Take the difference and find the absolute value to quantify the current data fluctuation range;

[0034] Statistically calculate the standard deviation of historical data as the benchmark volatility for standardizing the current deviation;

[0035] Apply the adjustment factor to the base threshold and finally output the dynamically adjusted threshold;

[0036] Preferably, the specific working steps of the warning unit are as follows:

[0037] According to the formula , calculate and obtain the risk assessment value R, where is the water level overrun value, and are indicator functions;

[0038] If the risk assessment value R is greater than , then trigger a warning.

[0039] Preferably, the specific working steps of the quantum gyroscope osmotic pressure sensor are as follows:

[0040] Measure the phase difference caused by the change of the refractive index of the medium through the optical fiber loop. The phase difference is proportional to the osmotic pressure gradient, so as to realize the accurate measurement of the osmotic pressure;

[0041] The phase difference is calculated according to the formula , where Ω is the angular velocity, A is the area of the optical fiber loop, λ is the wavelength, c is the speed of light, and the phase difference is calculated;

[0042] Convert the phase difference into an osmotic pressure gradient, and correlate the phase difference with the osmotic pressure gradient through a calibration curve or a mathematical model to obtain the osmotic pressure distribution inside the dam body.

[0043] The present invention also proposes a reservoir flood control monitoring method based on digital twin, including the following steps:

[0044] Step 1: Measure the internal seepage pressure gradient of the dam body through a quantum gyroscope osmotic pressure sensor, obtain the three-dimensional size data of the dam body cracks by using a lidar crack detector, and combine the underwater sonar array to real-time scan the reservoir underground terrain data. After all the data are aligned in time and space, they are fused with static databases such as historical flood evolution data and static geological parameters to form a comprehensive data set including seepage pressure field, crack propagation vector and terrain mutation;

[0045] Step 2: Quantify the cumulative impact of continuous rainfall on the dam body displacement based on the rainfall-deformation correlation matrix, simulate the flood evolution process by combining the two-dimensional shallow water equation, predict the future parameter change trend, and dynamically adjust the warning threshold through the formula;

[0046] Step 3: Calculate the flood control risk value and trigger an alarm when R exceeds the dynamically adjusted warning threshold.

[0047] Beneficial effects: By deploying quantum gyroscopic piezometric sensors, lidar crack detectors, and underwater sonar arrays, comprehensive real-time monitoring of the seepage pressure inside the dam body, surface cracks, and underwater terrain is achieved. This multi-dimensional monitoring method ensures the comprehensiveness and real-time nature of the data, enabling the timely detection of potential safety hazards such as excessive seepage pressure, crack expansion, and underwater obstacles. Compared with traditional single-data-source monitoring methods, it provides richer data support and a solid foundation for dam stability assessment and flood control decision-making.

[0048] Through the dynamic threshold adjustment unit and the warning unit, the warning threshold is dynamically optimized according to real-time monitoring data, and an alarm is triggered in real-time. Compared with traditional fixed-threshold warning systems, this system can adapt to changes in the dam body state and environmental conditions, identify potential abnormal events in advance, provide sufficient time for emergency response and scheduling decisions for reservoir flood control, thereby effectively reducing disaster risks and ensuring the safety of reservoir operation. Description of the Drawings

[0049] Figure 1 is a flowchart of the system of the present invention. Detailed Embodiments

[0050] As Figure 1 shown: A reservoir flood control monitoring system based on digital twin includes the following modules:

[0051] Multi-source data acquisition module, the multi-source data acquisition module includes quantum gyroscopic piezometric sensors, lidar crack detectors, and underwater sonar arrays, which are deployed inside the dam body of the reservoir and are used to collect multi-source data of the reservoir. The multi-source data includes seepage pressure gradient data collected by quantum gyroscopic piezometric sensors, size data of cracks inside the dam body obtained by lidar crack detectors, and terrain data underground of the reservoir. The multi-source data is used as real-time monitoring data and transmitted; it should be noted that the quantum gyroscopic piezometric sensor monitors the seepage pressure at different depths and positions inside the dam body in real-time, understands the penetration of water inside the dam body, and provides a basis for evaluating the stability of the dam body. For example, when the seepage pressure is too high, it may lead to safety hazards such as dam body landslides. By measuring the small tilt angle changes of the dam body, the deformation trend of the dam body can be detected in time, so as to take corresponding reinforcement measures to ensure the safety of the dam body.

[0052] The lidar crack detector can accurately detect geometric parameters such as the location, length, and width of cracks on the dam surface, helping staff quickly locate cracks and assess their severity. For example, if it is found that the crack length and width are increasing, it indicates that there may be structural hidden dangers in the dam. Using the high-precision three-dimensional point cloud data obtained by the lidar, a detailed analysis of the crack morphology is carried out, such as the crack trend, depth, and the relationship with the surrounding structure, etc., providing a scientific basis for crack repair and reinforcement;

[0053] The underwater sonar array scans and maps the underwater terrain in the deep water area of the reservoir, obtaining detailed data on the underwater terrain, including water depth, slope, concavity and convexity, etc., providing basic data for the hydrological research and engineering planning of the reservoir. It can detect underwater obstacles such as rocks, shipwrecks, and abandoned objects, avoiding the impact of these obstacles on the operation and maintenance of the reservoir, and at the same time providing safety guarantees for underwater operations;

[0054] Ensure the real-time and accuracy of the digital twin, providing basic data support for subsequent analysis and decision-making;

[0055] The digital twin module is used to construct a digital twin model according to the real-time monitoring data of the multi-source data acquisition module, and output a fusion data set through the data twin model. The fusion data set includes the seepage pressure field distribution, crack propagation vector, and underwater terrain mutation hotspots;

[0056] The coupling module is used to receive the fusion data set provided by the digital twin module, and perform dynamic simulation on the fusion data set, providing dynamic prediction results of dam seepage, deformation, and flood evolution;

[0057] The dynamic warning module includes a dynamic threshold adjustment unit and a warning unit. The dynamic threshold adjustment unit is used to dynamically optimize the warning threshold according to the dynamic prediction results of the coupling module;

[0058] The warning unit is used to calculate the flood control risk value of the reservoir, and trigger a warning in real time according to the optimized warning threshold based on the flood control risk value, obtaining a warning result.

[0059] It should be noted that the multi-source data acquisition module uses a quantum gyroscope piezometric sensor, a lidar crack detector, and an underwater sonar array to collect the internal seepage pressure of the dam, surface crack parameters, and underwater terrain data in real time. This multi-dimensional monitoring method can comprehensively perceive the physical state of the dam and timely discover potential safety hazards, such as excessive seepage pressure, crack propagation, and underwater obstacles;

[0060] The digital twin module constructs a high-precision digital twin model based on the L1-L3 level data floor and the BIM model. This model can reflect the actual state of the dam body in real time, providing reliable basic data support for subsequent dynamic simulation and analysis;

[0061] The coupling module realizes the dynamic prediction of seepage, deformation and flood routing of the dam body through the seepage-stress coupling equation, the rainfall-deformation correlation matrix and the flood routing hydrodynamic equation. This multi-physical field coupling simulation can identify potential risks in advance and provide a scientific basis for flood control decision-making;

[0062] The dynamic warning module, through the dynamic threshold adjustment unit and the warning unit, dynamically optimizes the warning threshold according to the real-time monitoring data and triggers the warning in real time. This intelligent warning mechanism can significantly reduce the false alarm rate, improve the accuracy and timeliness of the warning, and ensure the safety of the reservoir operation;

[0063] The system can support the automatic generation of multi-objective optimization scheduling schemes, providing scientific decision-making support for reservoir flood control. Through dynamic simulation and prediction, the system can formulate emergency plans in advance and improve the emergency response efficiency;

[0064] Traditional monitoring methods usually rely on a single data source and cannot comprehensively perceive the physical state of the dam body. Through the multi-source data acquisition module of this system, combined with quantum gyroscope piezometers, lidar crack detectors and underwater sonar arrays, the system realizes the comprehensive monitoring of the seepage pressure inside the dam body, surface cracks and underwater topography, solving the problem of incomplete data in traditional methods;

[0065] Traditional monitoring systems often have problems of data delay and insufficient accuracy. The digital twin module of this system ensures the real-time and accuracy of the twin model by receiving the monitoring data of the multi-source data acquisition module in real time, providing reliable data support for subsequent analysis and decision-making;

[0066] Traditional monitoring systems usually can only provide static data and cannot perform dynamic simulation and prediction. The coupling module of this system realizes the dynamic prediction of seepage, deformation and flood routing of the dam body through a multi-physical field coupling model, solving the problem that traditional systems cannot identify potential risks in advance;

[0067] Traditional warning systems usually adopt fixed thresholds, which are prone to false alarms or missed alarms. The dynamic warning module of this system dynamically optimizes the warning threshold according to real-time data through the dynamic threshold adjustment unit and the warning unit, significantly reducing the false alarm rate and improving the accuracy and timeliness of the warning.

[0068] As an optional embodiment: The specific working steps of the digital twin module are as follows:

[0069] Obtain the real-time monitoring data of the multi-source data acquisition module, and perform spatio-temporal alignment on the real-time monitoring data and the static data in the data floor; it should be noted that the data floor includes the L1-level data floor: serving as the topographic reference framework;

[0070] L2-level data floor: Integrate the underwater terrain data collected by the underwater sonar array to establish a spatial topological association network;

[0071] L3-level data floor: Embed the structured parameters of the BIM model and associate with historical hydrological archives and geological exploration data;

[0072] Train the historical flood evolution data through the Transformer neural network to generate a data-driven model;

[0073] Then perform Kriging interpolation spatial registration on the fused data to obtain a fused data set. It should be noted that based on the discrete data of seepage monitoring points (such as quantum gyroscope piezometers), use the Kriging interpolation algorithm to construct a continuous pressure field;

[0074] For the seepage pressure field distribution data, rely on the quantum gyroscope piezometer network and use the sliding window method to detect outliers and fill in missing values for the real-time monitoring data;

[0075] For the crack propagation vector data, each crack is composed of several triangular patches, the patch normal vector points to the crack propagation direction, and the vertices are arranged counterclockwise. Use a compact format of 50 bytes per triangular patch, and the header contains 80 bytes of metadata (such as crack ID, scanning time), and mark the crack risk level (such as the red warning area) through the Attributebytecount field;

[0076] For the underwater terrain mutation hotspot data, embed the ModelTransformationTag (affine transformation parameters) and ProjectedCSTypeGeoKey (projection coordinate system type);

[0077] Based on the DEM data, use the gdaldemslope tool in the GDAL library to generate a slope layer and mark the high-risk areas with a slope ≥ 15°;

[0078] It should be noted that the BIM model is used to provide detailed three-dimensional structural information of the dam body, including geometric shape, material properties, and internal structure. The underwater terrain oblique photography data is used to describe the undulation, slope, and obstacle distribution of the underwater terrain. The real-time monitoring parameters include data such as seepage pressure, displacement, settlement, stress and strain of the dam body. The millimeter-level crack dynamic mapping can accurately detect geometric parameters such as the position, length, and width of the cracks on the dam surface, and conduct a detailed analysis of the crack morphology. The three-dimensional visualization technology is used to display the physical state of the dam body and its surrounding environment in an intuitive three-dimensional form, facilitating the management personnel to view and analyze. The three-dimensional dynamic twin can be combined with the multi-physics field coupling model to realize the dynamic prediction of seepage, deformation, and flood evolution of the dam body;

[0079] The BIM model is used to provide detailed three-dimensional structural information of the dam body, including geometric shape, material properties, and internal structure. The underwater terrain oblique photography data is used to describe the undulation, slope, and obstacle distribution of the underwater terrain. The real-time monitoring parameters include data such as seepage pressure, displacement, settlement, stress and strain of the dam body. The millimeter-level crack dynamic mapping can accurately detect geometric parameters such as the position, length, and width of the cracks on the dam surface, and conduct a detailed analysis of the crack morphology;

[0080] It should also be noted that traditional technologies have significant shortcomings in the spatio-temporal alignment, format unification, and transmission stability of multi-source heterogeneous data (such as sensor data, BIM models, and laser point clouds). For example, in industrial scenarios, the acquisition cycles and measurement units of multi-dimensional data (geometric data, time data, physical parameters) are inconsistent, resulting in difficulties in data fusion;

[0081] This technical solution realizes cross-scale data fusion by constructing a multi-level data floor. For example, the L1-level floor integrates low-precision DOM / DEM as the reference framework, the L2-level fuses laser point clouds and underwater terrain data to establish a spatial topological relationship, and the L3-level embeds the BIM parametric model and associates historical data;

[0082] Combined with the Kriging interpolation algorithm and the Transformer neural network, discrete monitoring data (such as seepage pressure, crack propagation vector) is transformed into a continuous pressure field and a dynamic prediction model, improving data utilization and prediction accuracy;

[0083] Traditional digital twin models often fall into the misunderstanding of "being too simple or too complex", resulting in the inability to accurately serve specific industrial problems. This technical solution adopts a digital model fusion-driven method to construct a high-fidelity model by combining industrial mechanisms and multi-source data (such as sensor real-time data, historical fault database).

[0084] As an optional embodiment: The specific working steps of the coupling module are as follows:

[0085] First, the dam body is divided into element meshes of level

[0086] It should be noted that hexahedral meshes are used for the initial dam, and tetrahedral meshes are used for the embankment dam;

[0087] Then, the coupled equations of the seepage field and the stress field are solved; Specifically, it is solved according to the formula

[0088] where is the permeability coefficient tensor, which is obtained from the fusion dataset provided by the digital twin module, is the water head, which comes from the real-time monitoring data of the piezometric sensor, is the storage coefficient, which comes from the geological parameters of the dam body, is the effective stress tensor, which comes from the mechanical model of the dam body, is the coupling coefficient, which comes from experiments or numerical simulations;

[0089] In this embodiment, is the vector differential operator, which is used to represent the gradient operation;

[0090] is the change in water head per unit time divided by the change in time. The water head change rate represents the change in water head per unit time, which is represented by and represents the seepage change caused by stress change, which can accurately describe the flow characteristics of water in the dam body and its interaction with the stress field through operation;

[0091] It should be noted that in this embodiment, the water head is an important concept in hydraulics, which represents the mechanical energy possessed by a unit weight of liquid, including the position head, pressure head, and velocity head. The unit of the water head is usually meters (m). Specifically, the position head (z): represents the height of the liquid relative to a certain reference plane, reflecting the potential energy of the liquid. The pressure head (p / γ): represents the value obtained by converting the pressure at a certain point of the liquid into the height of a liquid column, where p is the pressure and γ is the unit weight of the liquid. The velocity head (v² / 2g): represents the energy corresponding to the flow velocity of the liquid, where v is the flow velocity and g is the acceleration due to gravity;

[0092] The initial value of the value of

[0093] is the permeability coefficient tensor. In this embodiment, for homogeneous materials (such as concrete or homogeneous soil), the permeability coefficient is usually a scalar value, which can be directly measured through laboratory tests. In this embodiment, the scalar permeability coefficient (homogeneous material) measured in the laboratory is integrated with the material parameter library in the digital twin module, and the permeability coefficient tensor is obtained by dynamically adjusting it through the Kalman filter algorithm;

[0094] Then, based on the rainfall-deformation correlation matrix, calculate the cumulative effect of rainfall in a continuous week on the dam displacement;

[0095] Finally, combine the two-dimensional shallow water equation to simulate the flood evolution process, and obtain the dynamic prediction results of the dam seepage, deformation and flood evolution.

[0096] It should be noted that traditional methods usually analyze the seepage field or stress field independently, ignoring the dynamic interaction between the two (for example, the change of seepage pressure leads to dam deformation, and the deformation further affects the seepage path). This solution divides the dam into parameters such as hexahedrons (initial dam) and tetrahedral stacked dams through finite element discretization, realizes the dynamic coupling calculation of the two fields, and solves the problem of prediction distortion of a single physical field model.

[0097] Moreover, environmental factors such as continuous rainfall and flood impact have the characteristics of time lag and nonlinearity on the damage of the dam. The module uses the rainfall-deformation correlation matrix to conduct correlation analysis on data such as rainfall intensity and duration within a week and the dam displacement. For example, the sliding window algorithm is used to calculate the cumulative effect of rainfall gradient on pore water pressure, and the viscoelastic constitutive model is combined to predict the lag deformation amount, solving the problem of lag in risk assessment under the action of environmental dynamic loads.

[0098] As an optional embodiment: The specific working steps of calculating the cumulative effect of rainfall in a continuous week on the dam displacement based on the rainfall-deformation correlation matrix are as follows:

[0099] According to the formula , calculate to obtain the Detect the sensitivity of the deformation within the detection band to the rainfall in the area , is the deformation amount, which comes from the real-time data of the lidar crack detector, is the rainfall intensity, which comes from the real-time monitoring data of the weather station, is the attenuation coefficient, and is used as the cumulative effect of rainfall in a continuous week on the dam displacement. It should be noted that the attenuation coefficient Determine the optimal value of the attenuation coefficient through the least squares method or other fitting methods. The value range is from 0.01 to 1. In this embodiment, the initial value is 0.5;

[0100] It should be noted that by simulating the hysteresis of rainfall infiltration and constructing an objective function based on historical monitoring data, the prior art does not consider the time lag of rainfall infiltration, and this technical solution can comprehensively address the issue of the time lag of rainfall infiltration;

[0101] It should be noted that it reflects the instantaneous influence degree of the change in rainfall intensity in the j-th area on the deformation of the i-th monitoring point. In the formula, is the partial derivative of deformation with respect to rainfall, representing the change in deformation caused by a unit change in rainfall intensity;

[0102] represents the infinitesimal change in the deformation of the

[0103] -th monitoring point. This change reflects the deformation condition of the dam body at this monitoring point and is usually obtained through real-time monitoring by devices such as lidar crack detectors; the infinitesimal change in the rainfall intensity in the

[0104] -th area at time t. This change reflects the dynamic change of rainfall intensity and is usually provided by the real-time monitoring data of meteorological stations; is the attenuation coefficient, which describes the time attenuation characteristic of the rainfall cumulative effect through the exponential function . The larger the attenuation coefficient

[0105] , the faster the cumulative influence of short-term rainfall on deformation attenuates; By using the sensitivity

[0106] , the non-linear relationship between rainfall intensity and dam body deformation is identified. In the initial stage of heavy rainfall, the influence of the attenuation coefficient is weak, and the sensitivity is mainly driven by the instantaneous rainfall intensity; during long-term continuous rainfall, the attenuation effect dominates and the sensitivity gradually decreases;

[0107] reveals the response differences of dam body deformation under different rainfall patterns (such as heavy rain, continuous light rain), providing a quantitative basis for flood control strategies; The optimized value of the attenuation coefficient

[0108] (such as obtaining λ = 0.3 through fitting) can reflect the infiltration characteristics of the dam body material; The partial derivative

[0109] itself can reflect the sensitivity of dam body deformation to changes in rainfall intensity, and this sensitivity may change over time or with changes in rainfall intensity, thus capturing the non-linear relationship. The time attenuation factor

[0110] Comprehensively consider the cumulative effect: By summing over all time steps, the formula can comprehensively consider the cumulative effect of rainfall, further capture the complex nonlinear relationship between rainfall intensity and dam deformation, and can evaluate the cumulative amount of dam displacement under specific rainfall scenarios (such as continuous rainfall for a week) and predict the critical safety threshold.

[0111] As an optional embodiment: The specific working steps of simulating the flood evolution process in combination with the two-dimensional shallow water equation are as follows:

[0112] The simulation process is as follows: Set the initial water depth and flow velocity, set the inflow, outflow, and solid wall boundary conditions, select an appropriate time step according to the CFL condition, and gradually advance the time step through the finite volume method and Godunov format to calculate the flood evolution process;

[0113] It should be noted that in this embodiment, the simulation is carried out according to the formula where is the conserved variable, including water depth (h), horizontal velocity (hu), and vertical velocity (hv), which are derived from the real-time monitoring data of the underwater sonar array, is the flux term, which describes the flow of water in the horizontal direction, is the source term, including terrain slope and frictional resistance;

[0114] The first part of the equation represents the rate of change of the conserved variable with time, ensuring that the total amount of flood remains conserved during the simulation;

[0115] The second part of the equation represents the divergence of the flux term, which describes the change of flood momentum in space, ensuring the conservation and correct propagation of momentum, Similarly, it is for gradient calculation;

[0116] The source term considers influencing factors such as terrain slope and frictional resistance, enabling the model to more accurately simulate the evolution of floods in complex terrains;

[0117] Visualize the simulation results to show the flood propagation path, flow velocity, and water depth changes;

[0118] Through the simulation, the flood propagation path, flow velocity, and water depth distribution at different time steps can be obtained. These results can be used to evaluate flood risks, optimize flood control measures, and guide emergency responses;

[0119] Then fuse the real-time underwater sonar data with the historical flood pattern library, optimize the initial conditions through the data twin model to output the fused data set, and introduce the source term correction driven by the fused data set in the shallow water equation.

[0120] By The source term is multiplied by the data-driven correction and the correction coefficient to obtain a new source term. , and the new source term is incorporated into the shallow water equations to simulate the flood propagation process;

[0121] Specifically, the data-driven correction is the correction amount generated by a residual network trained with historical flood data, which is used to compensate for the error of the source term.

[0122] The specific acquisition method is to input historical data into the shallow water equations (such as the S_phys term) to obtain the physical model prediction results, and the historical data is the historical fusion dataset;

[0123] The residual block structure is adopted. Each residual block contains a convolutional layer, batch normalization, and a ReLU activation function, and is obtained by learning the residual mapping between the input and output through cross-layer connections;

[0124] The role of the correction coefficient is to balance the contributions of the physical model and the data-driven correction term, and its value is 0.5 in this embodiment.

[0125] As an optional embodiment: The specific working steps of the dynamic warning module are as follows:

[0126] First, a Bayesian network is trained with historical data to calculate the probability of an abnormal event occurring;

[0127] In this embodiment, according to the formula , it is calculated that, where A is the abnormal event and B is the combination of detection parameters;

[0128] Then, based on time series data, the future parameter change trend is predicted.

[0129] It should be noted that in this embodiment, specifically: ;

[0130] Finally, the abnormal threshold is dynamically adjusted according to the predicted change trend. It should be noted that by training a Bayesian network with historical data, calculating the probability of an abnormal event occurring, and predicting the future parameter change trend based on time series data, the dynamic adjustment of the abnormal threshold is realized. This design has the following significant beneficial effects:

[0131] Through the Bayesian network, the dynamic warning module can calculate the probability of an abnormal event occurring based on historical data and real-time monitoring parameters, significantly improving the accuracy and reliability of the warning. This method can effectively identify potential safety hazards and avoid false alarms and missed alarms;

[0132] Predicting the future parameter change trend based on time series data, the dynamic warning module can adjust the anomaly threshold in real time according to the prediction results. This dynamic adjustment mechanism ensures that the warning threshold can adapt to the changes in the dam state and environmental conditions, improving the flexibility and adaptability of the warning;

[0133] By predicting the future parameter change trend, the dynamic warning module can identify potential abnormal events in advance, providing sufficient time for the reservoir flood control to carry out emergency response and dispatching decisions, thereby reducing the disaster risk;

[0134] The prediction and adjustment capabilities of the dynamic warning module provide a scientific basis for the reservoir flood control and support intelligent decision-making. By dynamically optimizing the warning threshold, the system can more accurately evaluate the health status of the dam, providing reliable data support for flood control dispatching and emergency response;

[0135] In summary, through the combination of Bayesian network and time series prediction, the dynamic warning module realizes accurate warning and dynamic adjustment of abnormal events, significantly improving the intelligent level and warning ability of the reservoir flood control monitoring system, and providing a solid guarantee for the safe operation of the reservoir.

[0136] As an optional embodiment: The specific steps for dynamically adjusting the anomaly threshold according to the predicted change trend are as follows:

[0137] Set an initial threshold according to historical data or business requirements as the benchmark for adjustment;

[0138] Through the prediction model, obtain the predicted value at the next moment , and take the absolute value of the difference from the current measured value to quantify the current data fluctuation range;

[0139] Statistical standard deviation of historical data is used as the benchmark volatility to standardize the current deviation;

[0140] Apply the adjustment factor to the base threshold and finally output the dynamically adjusted threshold.

[0141] In this embodiment, the specific calculation steps for adjustment are as follows:

[0142] According to the formula , calculate and obtain the adjusted threshold;

[0143] where is the initially set threshold, is the standard deviation of historical data. It should be noted that by combining the standard deviation of historical data and the changing trend of prediction parameters, the anomaly threshold is dynamically adjusted. This mechanism can flexibly optimize the warning threshold according to the actual operation status and environmental conditions of the dam body, ensuring the accuracy and adaptability of the warning. By reducing the false alarm rate and missed alarm rate, the system can identify potential risks more timely and provide more reliable warning support for reservoir flood control;

[0144] It should be noted that traditional methods rely on fixed thresholds and are difficult to cope with the dynamic changes of system load and environmental conditions (such as rainfall and floods). For example, during the rainy season, the seepage flow of the dam body may surge due to continuous rainfall, and fixed thresholds are prone to missed alarms; while normal fluctuations in the dry season may falsely trigger warnings;

[0145] This solution quantifies the data fluctuation range by introducing the standard deviation of historical data and dynamically adjusts the threshold in combination with the prediction trend, enabling the threshold range to flexibly expand and contract with the actual changes of parameters such as seepage pressure and deformation rate;

[0146] For example, when the prediction model detects that the rainfall in the next 72 hours exceeds the historical average, the sensitivity of the seepage pressure threshold is automatically reduced to capture potential risks in advance;

[0147] Moreover, the anomalies of parameters such as dam body deformation and seepage have hysteresis (such as displacement response occurring 1 - 3 days after heavy rainfall), and traditional methods are difficult to give timely warnings due to ignoring the cumulative effect of time series;

[0148] This solution can alleviate the above problems by modifying the threshold through the changing trend of parameters (such as the derivative of seepage rate ΔH / Δt);

[0149] By capturing the short - term fluctuation characteristics through prediction deviation, the dynamic response of the threshold to environmental changes is realized, which is superior to the fixed - threshold strategy, and the current deviation is converted into a ratio relative to the historical fluctuation level;

[0150] As an optional embodiment: The specific working steps of the warning unit are as follows:

[0151] According to the formula , the risk assessment value R is calculated and obtained, where is the water level over - limit value, and are indicator functions; It should be noted that is the baseline value, representing the initial risk threshold in the ideal state, is the water level over - limit value, representing the value by which the current water level exceeds the safety water level. If the water level is higher, the value is larger, and the contribution to the risk is greater. Therefore, the coefficient before in the formula is negative, indicating that the water level over - limit will reduce the R value and increase the risk;

[0152] Indicates that when a flooding event occurs, during the flooding the value is 1, otherwise it is 0; the flooding event significantly increases the risk, so the coefficient before I flooding in the formula is negative, indicating that flooding reduces the R value;

[0153] Indicates that when the power generation system is operating normally, indicates that the value of power generation is 1, otherwise it is 0; the normal operation of the power generation system helps to reduce the risk, so the coefficient before I power generation in the formula is positive, indicating that normal power generation increases the R value;

[0154] If the risk assessment value R is greater than , then an early warning is triggered. It should be noted that it can quickly quantify the risk, ensure that the system responds in a timely manner when potential threats appear, and improve the decision-making efficiency and system security.

[0155] As an optional embodiment: The specific working steps of the quantum gyroscope osmotic pressure sensor are as follows:

[0156] Measure the phase difference caused by the change of the refractive index of the medium through the optical fiber loop. The phase difference is proportional to the osmotic pressure gradient, so as to realize the accurate measurement of the osmotic pressure;

[0157] The phase difference is calculated according to the formula , where Ω is the angular velocity, A is the area of the optical fiber loop, λ is the wavelength, c is the speed of light, and the phase difference is calculated; this phase difference reflects the change of the osmotic pressure gradient;

[0158] Convert the phase difference into the osmotic pressure gradient, and correlate the phase difference with the osmotic pressure gradient through a calibration curve or a mathematical model, so as to obtain the osmotic pressure distribution inside the dam body.

[0159] The present invention also proposes a reservoir flood control monitoring method based on digital twin, including the following steps:

[0160] Step 1: Measure the internal seepage pressure gradient of the dam body through a quantum gyroscope osmotic pressure sensor, obtain the three-dimensional size data of the dam body cracks by using a lidar crack detector, and combine the underwater sonar array to scan the reservoir underground terrain data in real time. After all the data are aligned in time and space, they are fused with static databases such as historical flood evolution data and static geological parameters to form a comprehensive data set including the seepage pressure field, crack propagation vector and terrain mutation;

[0161] Step 2: Based on the rainfall-deformation correlation matrix, quantify the cumulative impact of continuous rainfall on the dam body displacement, combine the two-dimensional shallow water equation to simulate the flood evolution process, predict the future parameter change trend, and dynamically adjust the early warning threshold through the formula;

[0162] Step 3: Calculate the flood control risk value and trigger an alarm when R exceeds the dynamically adjusted warning threshold.

[0163] By deploying quantum gyroscope piezometric sensors, lidar crack detectors, and underwater sonar arrays, comprehensive real-time monitoring of seepage pressure inside the dam body, surface cracks, and underwater topography has been achieved. This multi-dimensional monitoring method ensures the comprehensiveness and real-time nature of the data, enabling the timely detection of potential safety hazards such as excessive seepage pressure, crack expansion, and underwater obstacles. Compared with traditional single-data-source monitoring methods, it provides richer data support and a solid foundation for dam stability assessment and flood control decision-making.

[0164] Through the dynamic threshold adjustment unit and the warning unit, the warning threshold is dynamically optimized according to real-time monitoring data, and the warning is triggered in real time. Compared with traditional fixed-threshold warning systems, this system can adapt to changes in the dam body state and environmental conditions, identify potential abnormal events in advance, provide sufficient time for emergency response and dispatching decisions for reservoir flood control, thereby effectively reducing disaster risks and ensuring the safety of reservoir operation.

[0165] The above is only the preferred implementation mode of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the concept of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, several improvements and refinements made without departing from the principle of the present invention should also be regarded as within the protection scope of this template.

Claims

1. A reservoir flood control monitoring system based on digital twin, characterized in that, It includes the following modules: The multi-source data acquisition module, including a quantum gyroscope osmotic pressure sensor, a lidar crack detector, and an underwater sonar array, is deployed inside the dam body of the reservoir to collect multi-source data of the reservoir. The multi-source data includes the seepage pressure gradient data collected by the quantum gyroscope osmotic pressure sensor, the size data of the cracks inside the dam body obtained by the lidar crack detector, and the topographic data underground in the reservoir. The multi-source data is used as real-time monitoring data and transmitted; The digital twin module is used to construct a digital twin model based on the real-time monitoring data of the multi-source data acquisition module, and output a fusion data set through the data twin model. The fusion data set includes the seepage pressure field distribution, crack expansion vector, and underwater topographic mutation hotspots; The coupling module is used to receive the fusion data set provided by the digital twin module, perform dynamic simulation on the fusion data set, and provide dynamic prediction results of dam seepage, deformation, and flood evolution; The dynamic warning module includes a dynamic threshold adjustment unit and a warning unit. The dynamic threshold adjustment unit is used to dynamically optimize the warning threshold according to the dynamic prediction results used by the coupling module; It is used to calculate the flood control risk value of the reservoir, and trigger a warning in real time according to the optimized warning threshold based on the flood control risk value to obtain a warning result; The specific working steps of the coupling module are as follows: First, the dam body is divided into element meshes of Then solve the coupled equation of the seepage field and the stress field; Then calculate the cumulative effect of rainfall in a continuous week on the dam displacement based on the rainfall-deformation correlation matrix; Finally, combine the two-dimensional shallow water equation to simulate the flood evolution process to obtain dynamic prediction results of dam seepage, deformation, and flood evolution; The specific working steps of calculating the cumulative effect of rainfall in a continuous week on the dam displacement based on the rainfall-deformation correlation matrix are as follows: According to the formula , the deformation within the detection band is calculated for the sensitivity of the rainfall in the area . is the amount of deformation, sourced from the real-time data of the lidar crack detector, is the rainfall intensity, sourced from the real-time monitoring data of the weather station, is the attenuation coefficient; is the partial derivative of deformation with respect to rainfall, representing the change in deformation caused by a unit change in rainfall intensity; the exponential function describes the time decay characteristics of the cumulative effect of rainfall.

2. The reservoir flood control monitoring system based on digital twin according to claim 1, characterized in that, The specific working steps of the digital twin module are as follows: Obtain the real-time monitoring data of the multi-source data acquisition module, and perform spatio-temporal alignment on the real-time monitoring data and the static data in the data floor; Train the historical flood evolution data through a Transformer neural network to generate a data-driven model; Then perform Kriging interpolation spatial registration on the fusion data to obtain a fusion data set.

3. A reservoir flood control monitoring system based on digital twin according to claim 1, characterized in that, The specific working steps of combining the two-dimensional shallow water equation to simulate the flood evolution process are as follows: The simulation process is as follows: set the initial water depth and flow velocity, set the inflow, outflow, and solid wall boundary conditions, select the time step according to the CFL condition, and gradually advance the time step through the finite volume method and the Godunov format to calculate the flood evolution process; Then fuse the real-time underwater sonar data with the historical flood pattern library, output the fusion data set through the data twin model to optimize the initial conditions, and introduce a source term correction driven by the fusion data set in the shallow water equation.

4. A reservoir flood control monitoring system based on digital twin according to claim 1, characterized in that, The specific working steps of the dynamic warning module are as follows: First, train a Bayesian network through historical data to calculate the probability of abnormal events occurring; Then predict the future parameter change trend based on time series data; Finally, dynamically adjust the abnormal threshold according to the predicted change trend.

5. The reservoir flood control monitoring system based on digital twin according to claim 4, wherein The specific steps of dynamically adjusting the abnormal threshold according to the predicted change trend are as follows: Set an initial threshold based on historical data or business requirements as the benchmark for adjustment; Obtain the predicted value at the next moment through the prediction model , and make a difference with the current measured value to calculate the absolute value of the difference and quantify the current data fluctuation range; Statistically calculate the standard deviation of historical data as the benchmark volatility for normalizing the current deviation; Apply the adjustment factor to the base threshold and finally output the dynamically adjusted threshold.

6. The reservoir flood control monitoring system based on digital twin according to claim 5, characterized in that The specific working steps of the warning unit are as follows: According to the formula , the risk assessment value R is calculated and obtained, where is the water level overrun value, and are indicator functions; If the risk assessment value R is greater than , a warning is triggered; the is the threshold value after dynamic adjustment.

7. The reservoir flood control monitoring system based on digital twin according to claim 1, characterized in that, The specific working steps of the quantum gyroscope osmotic pressure sensor are as follows: Measure the phase difference caused by the change in the refractive index of the medium through the fiber optic loop. The phase difference is proportional to the osmotic pressure gradient, thereby realizing the accurate measurement of the osmotic pressure; Phase difference formula is calculated, where Ω is the angular velocity, A is the area of the optical fiber loop, λ is the wavelength, and c is the speed of light, and the phase difference is calculated; Convert the phase difference into a seepage pressure gradient, and correlate the phase difference with the seepage pressure gradient through a calibration curve or a mathematical model, so as to obtain the seepage pressure distribution inside the dam body.

8. A reservoir flood control monitoring method based on digital twin, applicable to a reservoir flood control monitoring system according to any one of claims 1 to 7, characterized in that, Include the following steps: Step 1: Measure the internal seepage pressure gradient of the dam body through the quantum gyroscope osmotic pressure sensor, obtain the three-dimensional dimension data of the dam body cracks using the lidar crack detector, and combine the underwater sonar array to real-time scan the reservoir underground terrain data. After all the data are aligned in time and space, they are fused with the historical flood evolution data and static geological parameters to form a comprehensive data set containing the seepage pressure field, crack propagation vector, and terrain mutation; Step 2: Based on the rainfall-deformation correlation matrix, quantify the cumulative impact of continuous rainfall on the dam body displacement, combine the two-dimensional shallow water equation to simulate the flood evolution process, predict the future parameter change trend, and dynamically adjust the warning threshold through the formula; Step 3: Calculate the flood control risk value, and trigger a warning when R exceeds the dynamically adjusted warning threshold.

Images