Method and system for monitoring deformation of reservoir dam group based on multi-source fusion of beidou

Abstract

The application provides a kind of based on Beidou multi-source fusion's reservoir dam group deformation monitoring method and system, by receiving the satellite observation signal and sensing data of reservoir dam group monitoring area, obtain multi-source input data, input the preset reservoir dam group digital twin model, carry out physical behavior simulation to reservoir dam group structure, generate global deformation field, input it as initial boundary condition into deformation conduction dynamics model module in reservoir dam group digital twin model, identify the key conduction path of deformation existing in reservoir dam group, based on this, stress-deformation forward calculation and reverse tracing analysis are carried out in reservoir dam group digital twin model, the influence weight of core risk source on reservoir dam group deformation is calculated, and simulation scene of multiple control strategies is constructed in reservoir dam group digital twin model, the deformation trend of reservoir dam group under each simulation scene is simulated and deduced, and deformation early warning information is obtained. The application effectively improves the reliability of early warning response.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to a method and system for monitoring the deformation of reservoir and dam groups based on BeiDou multi-source fusion. Background Technology

[0002] In reservoir-dam group deformation monitoring, continuous tracking of the deformation state of reservoir and dam structures can provide data support for assessing dam stability and providing early warning of potential risks. Traditional reservoir-dam group deformation monitoring methods typically use satellite remote sensing to obtain large-scale deformation trends or deploy ground-based sensors to collect data on local structural strain, seepage, etc. Some technical solutions combine finite element analysis models to process the monitoring data to estimate the possible deformation state inside the reservoir and dam. However, in practical applications, these methods suffer from spatiotemporal differences between satellite observation data and ground-based sensor data, making it difficult to effectively coordinate multi-source data and affecting the accuracy of deformation analysis. At the same time, existing technologies mostly focus on the static description of deformation results and lack dynamic simulation of the deformation transmission law in the reservoir-dam group structure. It is difficult to identify key deformation transmission processes, and thus cannot accurately locate the core risk sources that cause deformation and their degree of impact. This makes the control strategies formulated based on monitoring results insufficiently targeted, and the timeliness and reliability of early warning responses need to be improved. Summary of the Invention

[0003] This invention provides a method and system for monitoring the deformation of reservoir and dam groups based on BeiDou multi-source fusion.

[0004] In a first aspect, embodiments of the present invention provide a method for monitoring the deformation of reservoir-dam clusters based on BeiDou multi-source fusion. The method includes: receiving satellite observation signals from the reservoir-dam cluster monitoring area and sensing data returned by multi-source sensing devices deployed within the monitoring area; unifying the satellite observation signals and the sensing data using a spatiotemporal reference to obtain multi-source input data; inputting the multi-source input data into a preset digital twin model of the reservoir-dam cluster to simulate the physical behavior of the reservoir-dam cluster structure, generating a global deformation field characterizing the surface and internal deformation state of the reservoir-dam cluster; and inputting the global deformation field as initial boundary conditions into the deformation transmission dynamics model module of the digital twin model of the reservoir-dam cluster to simulate... The process of deformation transmission in the reservoir-dam group structure was investigated, and key deformation transmission paths were identified. Based on these key transmission paths, stress-deformation forward calculation and reverse source tracing analysis were performed in the digital twin model of the reservoir-dam group. The stress distribution at each monitoring point along the key deformation transmission path was obtained through forward calculation, and the core risk sources triggering deformation transmission were located through reverse source tracing analysis. The influence weight of the core risk sources on the deformation of the reservoir-dam group was calculated. Based on the influence weight of the core risk sources, simulation scenarios with multiple control strategies were constructed in the digital twin model of the reservoir-dam group. The deformation trend of the reservoir-dam group under each simulation scenario was simulated and deduced to obtain deformation early warning information.

[0005] Secondly, embodiments of the present invention provide a monitoring system, comprising: a memory storing a computer program; and a processor for loading the computer program to implement the above-described method for monitoring reservoir and dam group deformation based on BeiDou multi-source fusion.

[0006] The reservoir-dam group deformation monitoring method based on BeiDou multi-source fusion provided by this invention receives satellite observation signals and sensor data returned from multi-source sensing devices, and unifies the spatiotemporal reference. This transforms heterogeneous data sources into multi-source input data with inherent correlation, effectively eliminating the limitations of single data sources in monitoring range and accuracy, and providing a more collaborative data foundation for subsequent deformation analysis. The multi-source input data is then input into a digital twin model of the reservoir-dam group for physical behavior simulation to generate a global deformation field. This achieves dynamic characterization of the complete deformation state of the reservoir-dam group from the surface to the interior, breaking through the limitations of traditional monitoring that only focuses on surface deformation and improving the comprehensiveness of deformation state description. The global deformation field is then used as the initial boundary condition input into the deformation transmission dynamics model module to simulate deformation propagation. The process identifies key transmission paths, revealing the transmission patterns of deformation in reservoir-dam complex structures and pinpointing core channels for deformation diffusion, providing precise spatial guidance for risk tracing. Based on these key transmission paths, stress-deformation forward and reverse analyses are performed. Forward calculations obtain stress distribution status, while reverse tracing analysis locates core risk sources and calculates their impact weights, achieving a deep correlation between deformation phenomena and the essence of risk, thus improving the accuracy of risk source identification and quantitative assessment capabilities. Simulation scenarios for control strategies are constructed based on the impact weights of core risk sources, and deformation trends are simulated and extrapolated to obtain early warning information. This allows for the development of targeted control plans based on the quantified degree of risk impact, predicting deformation development trends in advance, avoiding the lag of traditional early warning systems, and effectively improving the reliability of early warning responses. Attached Figure Description

[0007] Figure 1 This is a flowchart of a reservoir-dam group deformation monitoring method based on BeiDou multi-source fusion provided by an embodiment of the present invention.

[0008] Figure 2 This is a schematic diagram of the composition of a monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0009] Please see Figure 1 , Figure 1 A flowchart of a reservoir-dam group deformation monitoring method based on BeiDou multi-source fusion provided in this embodiment of the invention. The method can be executed by a monitoring system and may include the following steps:

[0010] Step S100: Receive satellite observation signals from the reservoir-dam group monitoring area and sensor data returned by multi-source sensing devices deployed within the reservoir-dam group monitoring area, and unify the spatiotemporal reference of the satellite observation signals and sensor data to obtain multi-source input data.

[0011] Satellite observation signals are signals transmitted by satellite systems such as BeiDou and captured by ground receiving equipment. They contain information such as the satellite's position and signal propagation time. Processing this information allows the spatial location of monitoring points to be determined. Multi-source sensing equipment refers to the deployment of various types of sensors within the reservoir-dam group monitoring area, such as strain sensors, seepage pressure sensors, and tilt sensors. These sensors can monitor changes in various physical quantities within the reservoir-dam group in real time and convert them into electrical or digital signals for feedback. Sensor data consists of the physical quantity data about the reservoir-dam group collected by these sensors. For example, strain sensor data reflects the strain of the reservoir-dam structure, seepage pressure data reflects changes in seepage pressure inside the reservoir-dam, and tilt sensor data records changes in the tilt angle of the reservoir-dam.

[0012] Spatiotemporal standardization is the process of unifying satellite observation signals and sensor data in time and space. Since satellite observation signals and sensor data may come from different data sources, and their time references and spatial coordinate systems may be inconsistent, spatiotemporal standardization is necessary. For example, this can be achieved by calibrating the sampling time of different sensors using a unified time standard (such as Coordinated Universal Time, UTC) and simultaneously transforming the spatial coordinates of different sensors to the same coordinate system (such as the WGS-84 coordinate system).

[0013] Step S200: Input multi-source input data into the preset digital twin model of the reservoir-dam group, perform physical behavior simulation on the reservoir-dam group structure, and generate a global deformation field characterizing the surface and internal deformation state of the reservoir-dam group.

[0014] The pre-defined digital twin model of the reservoir-dam group is a virtual model based on a physical model and data-driven approach, capable of simulating the structure and physical behavior of the reservoir-dam group. This model integrates the geometric information, material properties, and boundary conditions of the reservoir-dam group. By inputting multi-source data, it can simulate the physical behavior of the reservoir-dam group under various working conditions. Physical behavior simulation utilizes this digital twin model to simulate the mechanical response and deformation process of the reservoir-dam group under external loads (such as water level changes and seismic forces) and internal factors (such as material aging and seepage). The global deformation field is a physical field describing the deformation state of the reservoir-dam group's surface and interior, containing displacement and strain information at various locations within the reservoir-dam group, comprehensively reflecting its deformation. In practical applications, the digital twin model of the reservoir-dam group can be constructed using finite element analysis software (such as ANSYS and ABAQUS). During the construction process, the geometry of the reservoir-dam group is accurately modeled, the constitutive relations of the materials (such as elasticity and elastoplasticity) are defined, and boundary conditions (such as fixed boundaries and free boundaries) are set. After multi-source input data is fed into the model, the model updates its internal state variables based on the input data. Then, the governing equations of the model are solved using numerical calculation methods (such as the finite element method and the finite volume method) to obtain the stress, strain, and displacement distribution of the reservoir-dam group. Finally, a global deformation field is generated based on these calculation results.

[0015] In one implementation, step S200 may specifically include the following steps S210 to S250:

[0016] Step S210: Perform feature layering on the multi-source input data, extract spatial geometric features, time series features and physical field correlation features from the multi-source input data, and generate a structured feature set containing multiple feature channels.

[0017] Feature hierarchy is the process of classifying and extracting multi-source input data according to different feature types. This method decomposes complex multi-source data into features with different physical meanings. Spatial geometric features describe the geometric shape and positional changes of a reservoir-dam group in space, such as the changes in the three-dimensional coordinates of monitoring points, the magnitude of these changes, and the direction angle. Time series features reflect the changing patterns of multi-source input data over time. Time series analysis of strain sensing data, seepage pressure data, and tilt sensing data yields trend components, periodic components, and random disturbance components. Physical field correlation features describe the relationships between different physical fields. The deformation process of a reservoir-dam group involves the interaction of multiple physical fields, such as elasticity and fluid dynamics. By analyzing the cross-correlation between spatial geometric features and time series features, physical field correlation features such as elastic modulus correlation features, Poisson's ratio influence features, and permeability coefficient coupling features can be obtained. A structured feature set is a feature set obtained by organizing and integrating extracted spatial geometric features, time series features, and physical field correlation features according to certain rules. It contains multiple feature channels, each channel corresponding to a feature type. This structured approach can better represent the feature information of multi-source input data.

[0018] In one implementation, step S210 may include the following steps S211 to S215:

[0019] Step S211: Perform mode separation on the multi-source input data, decomposing the multi-source input data into satellite observation mode data and sensor device mode data. The satellite observation mode data includes carrier phase observation values ​​and pseudorange observation values, and the sensor device mode data includes strain sensing data, seepage pressure data and tilt sensing data.

[0020] Modality separation is the process of separating multi-source input data according to their source and properties. Since satellite observation signals and data collected by multi-source sensing devices have different physical meanings and characteristics, separating them facilitates subsequent processing and analysis. Satellite observation modal data is obtained from satellite observation signal processing. Carrier phase observations are obtained by measuring the carrier phase of the satellite signal, possessing high accuracy and suitable for high-precision positioning and deformation monitoring. Pseudorange observations are obtained by measuring the propagation time of the satellite signal, with relatively lower accuracy, but can be used for rapid positioning and gross error detection. Sensing device modal data is data collected by multi-source sensing devices. Strain sensing data reflects the strain of the dam structure, seepage pressure data reflects changes in seepage pressure within the dam, and tilt angle sensing data records changes in the dam's tilt angle. Specifically, modality separation can be performed on multi-source input data based on data identification information or acquisition methods. For satellite observation signals, the carrier phase and pseudorange can be measured and recorded using the receiver's signal processing module to obtain carrier phase observations and pseudorange observations. For multi-source sensing devices, the collected data can be classified and stored according to the type and number of the sensors to obtain strain sensing data, seepage pressure data and tilt angle sensing data.

[0021] Step S212: Extract spatial geometric features from satellite observation modal data, calculate the three-dimensional coordinate changes of monitoring points through carrier phase differential processing, and construct a spatial coordinate time series by combining BeiDou satellite ephemeris features. The spatial geometric features include the amplitude and orientation angle features of the coordinate changes.

[0022] Carrier phase differential processing compares the differences between satellite carrier phase observations received by two or more receivers to eliminate or reduce the influence of common errors such as satellite orbital errors, ionospheric delay, and tropospheric delay, thereby obtaining high-precision relative positioning results. In satellite observation modal data, differential processing of carrier phase observations allows calculation of the three-dimensional coordinate change of the monitoring point relative to a reference point. BeiDou satellite ephemeris features refer to the orbital and clock information of BeiDou satellites, which can be obtained by receiving satellite broadcast ephemeris or precise ephemeris. Combining BeiDou satellite ephemeris features, a spatial coordinate time series can be constructed, recording the three-dimensional coordinate information of the monitoring point at different times. The magnitude of the coordinate change in the spatial geometric features represents the magnitude of the coordinate change of the monitoring point within a certain time period, while the orientation angle feature represents the direction of the coordinate change. Specifically, a stable reference point needs to be selected first, whose coordinates are known and remain unchanged during monitoring. Then, the carrier phase observations of the monitoring point and the reference point are differentially processed to obtain differential carrier phase observations. By resolving the ambiguity of the differential carrier phase observations, the three-dimensional coordinate change of the monitoring point relative to the reference point can be obtained. At the same time, based on the received BeiDou satellite ephemeris information and the coordinate changes of the monitoring points, a spatial coordinate time series can be constructed.

[0023] Step S213: Extract time series features from the modal data of the sensing device, perform time series decomposition on the strain sensing data, seepage pressure data and tilt angle sensing data, separate the trend component, periodic component and random disturbance component, and generate a multi-dimensional time series feature vector.

[0024] Time series decomposition is the process of breaking down time series data into different components. This decomposition allows for a better understanding of the changing patterns and characteristics of time series data. Strain sensing data, seepage pressure data, and tilt angle sensing data in sensor modal data are all time series data that change over time, containing different components such as trend components, periodic components, and random disturbance components. The trend component reflects the long-term changing trend of the data over time; for example, the deformation of a reservoir-dam group may gradually increase over time, which is a trend change. The periodic component reflects the periodic changing pattern of the data; the seepage pressure of a reservoir-dam group may exhibit periodic fluctuations with seasonal changes. The random disturbance component represents fluctuations in the data caused by random factors, such as measurement errors and environmental noise. A multi-dimensional time series feature vector is a vector obtained by combining and representing the separated trend components, periodic components, and random disturbance components according to certain rules, which can comprehensively reflect the time series characteristics of sensor modal data. Specifically, different time series decomposition methods can be used to decompose strain sensing data, seepage pressure data, and tilt angle sensing data. Seasonal decomposition methods can decompose time series data into trend, seasonal, and residual terms, while wavelet decomposition methods can decompose time series data into components of different scales.

[0025] Step S214: Obtain the preset physical field correlation feature extraction rules, and based on the material property characteristics and geological conditions of the reservoir-dam group structure, obtain the cross-correlation degree between spatial geometric features and time series features, and generate a physical field correlation feature matrix that includes elastic modulus correlation features, Poisson's ratio influence features and permeability coefficient coupling features.

[0026] The predefined physical field correlation feature extraction rules are predefined rules used to extract the correlation between spatial geometric features and time series features. These rules can be established based on physical theories, empirical formulas, or machine learning algorithms. The material properties of the reservoir-dam group structure include the elastic modulus, Poisson's ratio, and density of the materials, while geological conditions include the geological structure and soil properties of the reservoir-dam group. The cross-correlation degree between spatial geometric features and time series features reflects the degree of interaction and influence between different physical fields. The deformation of the reservoir-dam group is not only related to changes in spatial location but also to changes in physical quantities over time (such as changes in seepage pressure and strain). The elastic modulus correlation feature describes the influence of the material's elastic modulus on the deformation of the reservoir-dam group; the Poisson's ratio influence feature reflects the effect of Poisson's ratio on the deformation of the reservoir-dam group; and the permeability coefficient coupling feature reflects the coupling relationship between the permeability coefficient and other physical quantities. The physical field correlation feature matrix is ​​a matrix obtained by organizing and arranging the extracted elastic modulus correlation features, Poisson's ratio influence features, and permeability coefficient coupling features according to certain rules. It can comprehensively represent the correlation between spatial geometric features and time series features.

[0027] Step S215: Concatenate the spatial geometric features, time series feature vectors, and physical field correlation feature matrix according to the channel dimension to generate a structured feature set with three-dimensional feature dimensions of space, time, and physical field. The product of the number of channels in the structured feature set and the feature dimension is equal to the sum of the number of channels of each individual feature.

[0028] The three-dimensional features of space, time, and physical field correspond to spatial geometric features, time series features, and physical field correlation features, respectively, describing the deformation characteristics of the reservoir-dam group from different perspectives. The number of channels in the structured feature set represents the number of channels contained in the feature set, and the feature dimension represents the number of features contained in each channel. The product of the number of channels in the structured feature set and the feature dimension is equal to the sum of the number of channels of each individual feature. This ensures that feature information is not lost during the stitching process.

[0029] Step S220: Input the structured feature set into the multi-scale physical field coupling module of the digital twin model of the reservoir-dam group, and solve the deformation response characteristics of the reservoir-dam group structure under the action of multiple physical fields by solving the coupling relationship between elasticity and fluid dynamics.

[0030] The multi-scale physics coupling module of the reservoir-dam group digital twin model is specifically designed to handle multi-physics interaction problems, considering the coupling effects of multiple physics fields such as elasticity and fluid dynamics. Elasticity studies the elastic deformation of an object under stress, while fluid dynamics studies the flow and mechanical properties of fluids. During the deformation of the reservoir-dam group, the deformation of the reservoir-dam structure is affected not only by the elastic mechanical properties of its own materials but also by the fluid dynamics of seepage within the reservoir-dam. By solving the coupling relationship between elasticity and fluid dynamics, the deformation response characteristics of the reservoir-dam group structure under the action of multiple physics fields can be obtained. The deformation response characteristics describe the deformation of the reservoir-dam group structure under the action of multiple physics fields, such as displacement, strain, and stress. Specifically, the structured feature set is first input into the multi-scale physics coupling module, which then establishes a coupled model of elasticity and fluid dynamics based on the spatial geometric features, time series features, and physics field correlation features in the structured feature set. Numerical methods such as the finite element method or the finite volume method can be used to solve the coupled model. The solution process requires consideration of the boundary and initial conditions of the reservoir-dam group structure, such as the contact boundary conditions between the dam bottom and bedrock, and the contact boundary conditions between the dam surface and water. Through iterative calculations, the deformation state of the reservoir-dam group structure is continuously updated until the preset convergence conditions are met. Finally, the deformation response characteristics of the reservoir-dam group structure under multiphysics field conditions are obtained.

[0031] In one implementation, step S220 may include the following steps S221 to S225:

[0032] Step S221: Input the structured feature set into the feature allocation layer of the multi-scale physical field coupling module, and allocate the structured feature set to the elasticity calculation branch and the fluid dynamics calculation branch according to the feature dimension. The elasticity calculation branch receives the solid mechanics features in the spatial geometric features and the physical field associated feature matrix, and the fluid dynamics calculation branch receives the seepage features in the time series feature vector and the physical field associated feature matrix.

[0033] The feature allocation layer of the multi-scale physics coupling module is used to allocate and manage the structured feature set. Based on the feature's dimension and properties, different features in the structured feature set are assigned to different computational branches. The elasticity calculation branch mainly handles problems related to solid mechanics. Spatial geometric features describe the spatial location and deformation of the reservoir-dam group structure, while solid mechanics features in the physics correlation feature matrix, such as elastic modulus correlation features and Poisson's ratio influence features, reflect the mechanical properties of the reservoir-dam materials. The fluid dynamics calculation branch mainly handles problems related to fluid mechanics. Seepage pressure data in the time series feature vector reflects the fluid flow inside the reservoir-dam, while seepage features in the physics correlation feature matrix, such as permeability coefficient coupling features, describe the interaction between the fluid and solid. Specifically, the feature allocation layer analyzes and identifies the structured feature set, assigning it to the elasticity calculation branch and the fluid dynamics calculation branch based on the feature's dimension and type. A rule-matching method can be used, pre-defining that spatial geometric features and solid mechanics features belong to the elasticity calculation branch, while time series feature vectors and seepage features belong to the fluid dynamics calculation branch. When the structured feature set is input into the feature allocation layer, the layer will allocate the corresponding features to the corresponding computation branches according to these rules.

[0034] Step S222: In the elasticity calculation branch, based on the three-dimensional elastic constitutive relation constructed by Hooke's law, the three-dimensional coordinate changes in the spatial geometric features are used as displacement boundary conditions. The three-dimensional elastic constitutive relation is input for finite element solution to generate the structural stress distribution characteristic tensor.

[0035] The three-dimensional coordinate changes in the spatial geometric features represent the displacement changes of the reservoir-dam group structure over a certain time period. Inputting these changes as displacement boundary conditions into the three-dimensional elastic constitutive relation allows determination of the deformation state of the reservoir-dam group structure. The finite element method (FEM) discretizes the continuous elastic body into a finite number of elements. By solving the equilibrium equations of each element, the stress and strain distribution of the entire elastic body is obtained. The structural stress distribution characteristic tensor describes the stress distribution within the reservoir-dam group structure, containing information on the magnitude, direction, and distribution of stress. Specifically, a three-dimensional elastic constitutive relation is first established based on the geometry and material properties of the reservoir-dam group. Then, the three-dimensional coordinate changes in the spatial geometric features are applied as displacement boundary conditions to the finite element model of the reservoir-dam group. The finite element model discretizes the reservoir-dam group structure into multiple elements, each element having displacement degrees of freedom at its nodes. By solving the equilibrium equations of the finite element model, the stress and strain distribution of each element can be obtained. Finally, the stress distribution information of all elements is integrated to generate the structural stress distribution characteristic tensor.

[0036] Step S223: In the fluid dynamics calculation branch, based on the seepage field control relationship constructed by Darcy's law, the seepage pressure data in the time series feature vector is used as the initial condition, and the seepage field control relationship is solved by the finite volume method to generate the pore water pressure distribution feature matrix.

[0037] Based on Darcy's law, a seepage field control relationship can be constructed, describing the relationship between pore water pressure and seepage velocity. The seepage pressure data in the time-series feature vector records the changes in seepage pressure at different locations within the reservoir / dam over time. Using this data as initial conditions input into the seepage field control relationship determines the initial state of the seepage field. The finite volume method is a numerical method for solving partial differential equations. It divides the computational domain into a finite number of control volumes. By integrating over each control volume, discrete control equations are obtained, and then these equations are solved to obtain the distribution of the seepage field. The pore water pressure distribution feature matrix describes the pore water pressure distribution within the reservoir / dam, containing pore water pressure values ​​at different locations. Specifically, firstly, based on the geological conditions and material properties of the reservoir / dam group, a seepage field control relationship based on Darcy's law is established. Then, the seepage pressure data in the time-series feature vector is applied as initial conditions to the finite volume model of the seepage field. The finite volume model divides the seepage region within the reservoir / dam into multiple control volumes, each with a degree of freedom regarding pore water pressure. By solving the governing equations of the finite volume model, the pore water pressure distribution of each control volume can be obtained. Finally, the pore water pressure distribution information of all control volumes is integrated to generate a pore water pressure distribution feature matrix.

[0038] Step S224: Construct a fluid-structure interaction interface processing layer, exchange interface data between the structural stress distribution characteristic tensor and the pore water pressure distribution characteristic matrix, and update the bidirectional coupling relationship between structural deformation and seepage field distribution through iterative calculation until the preset convergence condition is met.

[0039] The fluid-structure interaction (FSI) interface processing layer is used to handle the interaction between fluid and solid. It is responsible for exchanging interface data between the structural stress distribution characteristic tensor and the pore water pressure distribution characteristic matrix, achieving bidirectional coupling between structural deformation and seepage field distribution. The structural stress distribution characteristic tensor describes the stress distribution within the reservoir-dam complex structure, while the pore water pressure distribution characteristic matrix describes the pore water pressure distribution within the reservoir-dam. These two characteristics interact and influence each other. During FSI, structural deformation affects pore water flow and pressure distribution, while changes in pore water pressure also affect structural stress and deformation. Iterative calculation is a method that gradually approximates the solution by continuously updating the structural deformation and seepage field distribution until a preset convergence condition is met. The preset convergence condition could be that the changes in structural deformation and seepage field distribution are less than a certain threshold, or that the error of the calculation result is within an acceptable range. Specifically, the FSI interface processing layer matches and maps the structural stress distribution characteristic tensor and the pore water pressure distribution characteristic matrix to determine the nodes and elements on the FSI interface. Then, in each iteration step, the stress information in the structural stress distribution feature tensor is transferred to the pore water pressure distribution feature matrix to update the pore water pressure distribution; simultaneously, the pressure information in the pore water pressure distribution feature matrix is ​​transferred to the structural stress distribution feature tensor to update the structural deformation. This process continues iterating until a preset convergence condition is met.

[0040] Step S225: Based on the converged structural stress distribution characteristic tensor and pore water pressure distribution characteristic matrix, the comprehensive deformation response characteristics of the reservoir-dam group structure under the action of multiple physical fields are obtained. The dimension of the deformation response characteristics is matched with the number of feature channels of the structured feature set.

[0041] The converged structural stress distribution feature tensor and pore water pressure distribution feature matrix represent the stress distribution and pore water pressure distribution of the reservoir-dam group structure obtained after the convergence of fluid-structure interaction calculations. The comprehensive deformation response feature is the deformation characteristic of the reservoir-dam group structure obtained after considering the interaction of multiple physical fields such as elasticity and fluid dynamics, including information on displacement, strain, and stress. The dimension of the deformation response feature matches the number of feature channels in the structured feature set, ensuring consistency between the deformation response feature and the input structured feature set. Specifically, based on the converged structural stress distribution feature tensor and pore water pressure distribution feature matrix, the deformation response characteristics of the reservoir-dam group structure, such as displacement, strain, and stress, can be calculated. The strain distribution can be calculated from the stress distribution feature tensor using the stress-strain relationship and geometric equations, and then the displacement distribution can be obtained through integration. Simultaneously, the influence of pore water pressure on structural deformation can be considered by combining the pore water pressure distribution feature matrix. The final comprehensive deformation response feature has the same dimension as the number of feature channels in the structured feature set, allowing for correlation and comparative analysis between the deformation response feature and the structured feature set.

[0042] Step S230: Based on the deformation response characteristics, adaptive mesh refinement is performed on the reservoir-dam group structure in the finite element mesh generation module of the digital twin model of the reservoir-dam group to generate a layered mesh model containing different precision levels.

[0043] The finite element mesh generation module is used in the digital twin model of the reservoir-dam complex to mesh the structure. Adaptive mesh refinement is a method that automatically adjusts the mesh density based on deformation response characteristics. Deformation response characteristics reflect the deformation of the reservoir-dam complex structure under multi-physics conditions. In areas with large or drastic deformation, a denser mesh is needed to accurately describe the stress and strain distribution of the structure; while in areas with small or gradual deformation, a sparser mesh can be used to reduce computational load. The layered mesh model contains meshes of different precision levels, each corresponding to a different mesh density. This layered approach improves computational efficiency while maintaining accuracy.

[0044] Specifically, the finite element mesh generation module determines which areas of the reservoir-dam complex structure require mesh refinement based on deformation response characteristics. The areas for mesh refinement can be determined using indicators such as displacement gradient and strain gradient in the deformation response characteristics. For areas requiring refinement, the module subdivides the original mesh to generate a denser mesh; for areas not requiring refinement, the original mesh density remains unchanged. Through multiple iterations and adjustments, a layered mesh model containing different levels of precision is ultimately generated.

[0045] Step S240: Perform physical behavior simulation on the surface and internal structure of the reservoir and dam group using a hierarchical mesh model, collect the displacement vector and strain tensor of each mesh node, and construct a deformation state dataset covering the entire reservoir and dam group.

[0046] The hierarchical mesh model has been used to mesh the reservoir and dam group structure, with different regions having different levels of precision. Simulating the physical behavior of the reservoir and dam group's surface and internal structure using this model involves simulating the mechanical response and deformation process of the reservoir and dam group under external loads and internal factors. The displacement vector of each mesh node represents the magnitude and direction of that node's displacement in space, while the strain tensor describes the strain state of the element containing that node. The deformation state dataset is obtained by integrating the collected displacement vectors and strain tensors of each mesh node, covering the entire reservoir and dam group and comprehensively reflecting its deformation state.

[0047] Specifically, the hierarchical mesh model is input into the simulation module of the digital twin model of the reservoir-dam group. This module simulates the physical behavior of the reservoir-dam group based on its material properties, boundary conditions, and load conditions. During the simulation, the displacement vector and strain tensor of each mesh node are calculated and recorded. Finally, the displacement vectors and strain tensors of all mesh nodes are integrated to construct a deformation state dataset covering the entire reservoir-dam group.

[0048] Step S250: Perform spatiotemporal interpolation on the deformation state dataset to generate a continuously distributed global deformation field. The spatial resolution of the global deformation field is the same as the highest accuracy level of the layered mesh model.

[0049] The deformation state dataset is discrete data composed of displacement vectors and strain tensors of each grid node. Spatiotemporal interpolation can be used to obtain deformation information at any location on the surface and inside the reservoir-dam group. The global deformation field is a continuous physical field describing the deformation state of the reservoir-dam group's surface and interior, capable of comprehensively and accurately reflecting the deformation situation. The spatial resolution of the global deformation field is the same as the highest accuracy level of the layered grid model. This means that in the global deformation field, deformation information can be described using the highest accuracy level of the grid density in the layered grid model, ensuring the accuracy and precision of the global deformation field. Specifically, different spatiotemporal interpolation methods can be used to interpolate the deformation state dataset. For spatial interpolation, methods such as Kriging interpolation and spline interpolation can be used to predict deformation values ​​at other locations based on the known deformation information of grid nodes. For temporal interpolation, methods such as linear interpolation and spline interpolation can be used to obtain deformation information at any time based on deformation state data at different times. Through spatiotemporal interpolation, the discrete deformation state dataset is transformed into a continuously distributed global deformation field.

[0050] Step S300: Input the global deformation field as the initial boundary condition into the deformation transmission dynamics model module in the digital twin model of the reservoir-dam group, simulate the transmission process of deformation in the reservoir-dam group structure, and identify the key deformation transmission paths in the reservoir-dam group.

[0051] The global deformation field describes the deformation state of the reservoir-dam group's surface and interior. Inputting this as the initial boundary condition into the deformation transmission dynamics model module determines the initial deformation state of the reservoir-dam group structure. This module is specifically designed to simulate the deformation transmission process within the reservoir-dam group structure. It considers factors such as the material properties, structural characteristics, and boundary conditions of the reservoir-dam group, simulating the deformation transmission process by solving the dynamic equations. The critical deformation transmission path refers to the path within the reservoir-dam group structure that has a high deformation transmission efficiency and a significant impact on the overall deformation of the reservoir-dam group.

[0052] In one implementation, step S300 may include the following steps S310 to S350:

[0053] Step S310: Extract boundary conditions for the global deformation field, determine the displacement constraint boundary and force boundary conditions of the reservoir-dam group structure, and use the displacement value of the displacement constraint boundary and the stress value of the force boundary conditions as initial input features.

[0054] The global deformation field contains deformation information of the reservoir-dam group's surface and interior. Extracting boundary conditions from this field determines the boundary conditions of the reservoir-dam group structure. Displacement-constrained boundaries refer to the boundary regions within the reservoir-dam group structure that are subject to displacement restrictions. The bottom region of the reservoir-dam group, in contact with bedrock, is typically a fixed boundary where displacement is restricted. Force boundary conditions refer to the boundary regions within the reservoir-dam group structure that are subjected to external forces. The area of ​​the reservoir-dam surface in contact with water is subjected to water pressure. The displacement values ​​of the displacement-constrained boundaries represent the magnitude and direction of displacement at each node on that boundary, while the stress values ​​of the force boundary conditions represent the magnitude and direction of stress at each node on that boundary. The initial input features are obtained by integrating the displacement values ​​of the displacement-constrained boundaries and the stress values ​​of the force boundary conditions. These features serve as input data for the deformation transmission dynamics model module, used to determine the boundary state of the reservoir-dam group structure at the initial moment.

[0055] In one implementation, step S310 may include the following steps S311 to S315:

[0056] Step S311: Extract isosurfaces from the global deformation field to generate a set of deformation isosurfaces representing the same degree of deformation. The set of deformation isosurfaces includes multiple nested isosurfaces extending from the surface of the reservoir-dam group into the interior.

[0057] Isosurface extraction is a method for extracting surfaces with the same numerical values ​​from a three-dimensional data field. Extracting isosurfaces from a global deformation field means extracting surfaces with the same degree of deformation. A set of deformation isosurfaces is a collection of multiple isosurfaces, each representing a region within a reservoir / dam group with the same degree of deformation. Multi-layered nested isosurfaces refer to isosurfaces extending inward from the surface of the reservoir / dam group and nested within each other, with the degree of deformation gradually changing from the outside to the inside. Specifically, the Marching Cubes algorithm can be used to extract isosurfaces. This algorithm discretizes the global deformation field into a three-dimensional cubic mesh. By determining the relationship between the vertex values ​​of each cube and the isosurface values, the intersection of the cube and the isosurface is determined. Based on the intersection, isosurface fragments are generated inside the cube. These fragments are then stitched together to obtain a complete isosurface. By setting different isosurface values, multiple isosurfaces can be obtained, thus generating a set of deformation isosurfaces.

[0058] Step S312: Identify the fixed boundary region and free boundary region of the reservoir-dam group structure in the set of deformation isosurfaces. The fixed boundary region is the bottom region of the reservoir-dam in contact with the bedrock, and the free boundary region is the region of the reservoir-dam surface in contact with the water.

[0059] In the set of deformation isosurfaces, different isosurfaces reflect the degree of deformation at different locations within the reservoir-dam group. Fixed boundary regions refer to areas within the reservoir-dam group structure that are subject to displacement constraints. The bottom region of the reservoir-dam, in contact with bedrock, is typically a fixed boundary, its displacement constrained by the bedrock, resulting in relatively small deformation. Free boundary regions refer to areas within the reservoir-dam group structure that can deform freely. The areas of the reservoir-dam surface in contact with water are not subject to significant displacement constraints and are affected by water pressure and flow, resulting in relatively large deformation. Specifically, fixed boundary regions and free boundary regions are identified based on the distribution of the deformation isosurface set and the structural characteristics of the reservoir-dam group. By analyzing the shape and location of the isosurfaces, it can be determined which areas are in contact with bedrock and which are in contact with water. For the bottom region of the reservoir-dam in contact with bedrock, its isosurfaces are usually relatively flat and the degree of deformation is small, thus it is identified as a fixed boundary region; for the areas of the reservoir-dam surface in contact with water, its isosurfaces are usually more complex and the degree of deformation is large, thus it is identified as a free boundary region.

[0060] Step S313: Extract the displacement constraint boundary in the fixed boundary region, collect the three-dimensional displacement vector of each node on the displacement constraint boundary, obtain the magnitude and direction cosine of the three-dimensional displacement vector, and generate the displacement constraint boundary condition feature set.

[0061] Displacement constraint boundaries are specific boundaries within a fixed boundary region. Extracting displacement constraint boundaries from this region determines which nodes' displacements are restricted. A three-dimensional displacement vector represents the magnitude and direction of the displacement of each node on the displacement constraint boundary in three-dimensional space; its magnitude represents the magnitude, and its direction cosine represents the direction. The displacement constraint boundary condition feature set is obtained by integrating the magnitude and direction cosine of the collected three-dimensional displacement vectors of each node, describing the displacement characteristics of the displacement constraint boundary. Specifically, first, the nodes on the displacement constraint boundary are determined based on the location and shape of the fixed boundary region. Then, the three-dimensional displacement vectors of these nodes are collected from the global deformation field. The magnitude and direction cosine of each three-dimensional displacement vector are calculated and used as eigenvalues. Finally, the eigenvalues ​​of all nodes are organized and stored to generate the displacement constraint boundary condition feature set.

[0062] Step S314: Extract force boundary conditions in the free boundary region. Based on the hydrological and meteorological characteristics of the reservoir and dam group, obtain the distribution force vectors of static water pressure, dynamic water pressure and wind load on the free boundary region, and generate a force boundary condition feature set.

[0063] The free boundary region is the area within the reservoir-dam group structure subjected to external forces. Extracting force boundary conditions within this region determines the external forces acting on it. The hydrological environment of the reservoir-dam group includes water level, flow velocity, and water temperature, while meteorological characteristics include wind speed, wind direction, and air temperature. Static pressure is the pressure generated by the gravity of the water body, hydrodynamic pressure is the pressure generated by the flow of water, and wind load is the load generated by the action of wind. The distributed force vector represents the distribution of these external forces in the free boundary region, including the magnitude and direction of the forces. The force boundary condition feature set is a set of features obtained by integrating the distributed force vectors, describing the force boundary conditions of the free boundary region. Specifically, based on the hydrological and meteorological characteristics of the reservoir-dam group, appropriate formulas and methods are used to calculate the static pressure, hydrodynamic pressure, and wind load. For static pressure, the pressure values ​​at different depths can be calculated based on hydrostatic principles; for hydrodynamic pressure, fluid mechanics formulas can be used to calculate the pressure values ​​based on flow velocity and water density; for wind load, the wind force can be calculated based on meteorological data and the shape of the structure. The calculated distributed force vectors are allocated and organized according to the nodes on the free boundary region to generate a force boundary condition feature set.

[0064] Step S315: Spatiotemporally align the displacement constraint boundary condition feature set with the force boundary condition feature set so that all features have a unified timestamp and spatial coordinate system, and generate the initial input feature sequence of the deformation transmission dynamics model module. The time interval of the initial input feature sequence is the same as the time resolution of the global deformation field.

[0065] The displacement constraint boundary condition feature set and the force boundary condition feature set describe the displacement constraint boundary and force boundary conditions of the reservoir-dam group structure, respectively. However, they may originate from different data sources or acquisition times, thus requiring spatiotemporal alignment. Spatiotemporal alignment is the process of unifying the timestamps and spatial coordinate systems of different feature sets, ensuring that all features have the same temporal and spatial reference. The initial input feature sequence is the feature sequence obtained after spatiotemporally aligning the displacement constraint boundary condition feature set and the force boundary condition feature set. It serves as the input data for the deformation transmission dynamics model module, used to determine the boundary state of the reservoir-dam group structure at the initial moment. The time interval of the initial input feature sequence is the same as the temporal resolution of the global deformation field, ensuring temporal consistency between the input features and the global deformation field. Specifically, firstly, the timestamps of the displacement constraint boundary condition feature set and the force boundary condition feature set are calibrated to have the same time reference. Interpolation or extrapolation methods can be used to transform feature data acquired at different times to the same time point. Then, the spatial coordinate system of the feature sets is transformed to have the same spatial reference. Coordinate transformation formulas can be used to transform feature data from different coordinate systems to the same coordinate system. Finally, the spatiotemporally aligned displacement constraint boundary condition feature set and force boundary condition feature set are integrated to generate the initial input feature sequence of the deformation transmission dynamics model module.

[0066] Step S320: Load the initial input features into the structural dynamics analysis model of the deformation transmission dynamics model module, and solve the structural dynamics analysis model in the time domain using the time integration method to obtain the structural vibration response features at different times.

[0067] The initial input features already include the displacement constraint boundary and force boundary conditions of the reservoir-dam group structure. Loading these features into the structural dynamics analysis model of the deformation transmission dynamics model module means applying these boundary conditions to the finite element model of the structural dynamics analysis model. The structural dynamics analysis model is used to describe the mechanical response and vibration characteristics of the reservoir-dam group structure under dynamic loads, considering factors such as the structure's mass, stiffness, and damping. The time integration method is a numerical method for solving dynamic equations. Solving the structural dynamics analysis model using the time integration method in the time domain means progressively solving for the structure's displacement, velocity, and acceleration responses in the time domain. The structural vibration response characteristics at different times refer to the vibration state of the reservoir-dam group structure at different time points, including the magnitude and changes of physical quantities such as displacement, velocity, and acceleration.

[0068] In one implementation, step S320 may include the following steps S321 to S325:

[0069] Step S321: Obtain the pre-constructed finite element dynamic model of the reservoir-dam group structure. The element types of the finite element dynamic model include solid elements, shell elements and contact elements. Solid elements are used to simulate the main structure of the reservoir-dam, shell elements are used to simulate the dam surface protective layer, and contact elements are used to simulate the contact behavior between the dam body and the bedrock.

[0070] The pre-constructed finite element dynamic model of the reservoir-dam group structure is a pre-established finite element model used to simulate the mechanical response and vibration characteristics of the reservoir-dam group under dynamic loads. Solid elements are used to simulate three-dimensional solid structures; in the reservoir-dam group, solid elements are used to simulate the main structure of the reservoir and dam, such as the dam body and dam foundation. Shell elements are used to simulate thin plate or thin shell structures; in the reservoir-dam group, shell elements are used to simulate the dam surface protective layer, such as concrete facing. Contact elements are used to simulate the contact behavior between two objects; in the reservoir-dam group, contact elements are used to simulate the contact between the dam body and the bedrock, considering issues such as force transmission and displacement coordination at the contact surface. Specifically, based on the design drawings, geological survey reports, and other data of the reservoir-dam group, finite element dynamic models of the reservoir-dam group structure are established using finite element software (such as ANSYS, ABAQUS, etc.). In the model, appropriate solid element, shell element, and contact element types are selected, and meshing is performed according to the geometry and structural characteristics of the reservoir-dam group.

[0071] Step S322: Load the initial input features onto the boundary nodes of the finite element dynamics model, apply the displacement constraint boundary condition feature set to the node degrees of freedom in the fixed boundary region, and apply the force boundary condition feature set to the node load vector in the free boundary region.

[0072] The initial input features already include displacement constraint boundary condition feature sets and force boundary condition feature sets. Loading these features onto the boundary nodes of the finite element dynamics model applies these boundary conditions to the corresponding nodes of the finite element model. The displacement constraint boundary condition feature set describes the displacement constraints in the fixed boundary region; applying it to the nodal degrees of freedom in the fixed boundary region restricts the displacement of these nodes in certain directions. The force boundary condition feature set describes the external forces acting in the free boundary region; applying it to the nodal load vectors in the free boundary region applies these external forces to the corresponding nodes. Specifically, firstly, based on the node numbers and boundary condition location information of the finite element dynamics model, the nodes in the fixed boundary region and the free boundary region are determined. For nodes in the fixed boundary region, the displacement values ​​from the displacement constraint boundary condition feature set are applied to the corresponding degrees of freedom of that node to restrict its displacement. For nodes in the free boundary region, the stress values ​​from the force boundary condition feature set are converted into nodal load vectors and applied to that node.

[0073] Step S323: Establish a structural dynamics control relationship model based on the finite element dynamics model. The structural dynamics control relationship model includes a mass matrix, a stiffness matrix, and a damping matrix. The mass matrix is ​​constructed based on the density characteristics of the reservoir dam material, the stiffness matrix is ​​constructed based on the elastic modulus characteristics, and the damping matrix is ​​constructed using the Rayleigh damping model.

[0074] The structural dynamics control relationship model is a mathematical model used to describe the mechanical response and vibration characteristics of a reservoir-dam group structure under dynamic loads. It includes important parameters such as the mass matrix, stiffness matrix, and damping matrix. The mass matrix describes the mass distribution of the reservoir-dam group structure and can be constructed based on the density characteristics of the reservoir-dam materials and the element information of the finite element model. The stiffness matrix describes the stiffness characteristics of the reservoir-dam group structure and can be constructed based on the elastic modulus characteristics of the reservoir-dam materials and the element information of the finite element model. The damping matrix describes the energy dissipation characteristics of the reservoir-dam group structure during vibration. It is constructed using the Rayleigh damping model, which represents the damping matrix as a linear combination of the mass matrix and stiffness matrix. Specifically, firstly, the mass matrix and stiffness matrix of each element are calculated based on the element type and material properties of the finite element dynamics model. Then, the mass matrices and stiffness matrices of all elements are assembled to obtain the mass matrix and stiffness matrix of the entire reservoir-dam group structure. For the damping matrix, the Rayleigh damping model is used, and the coefficients of the damping matrix are determined based on the natural frequency and damping ratio of the structure.

[0075] Step S324: Numerically solve the structural dynamics control relationship model using the time integration method, setting the integration time step to be coordinated with the time interval of the initial input feature sequence, and obtaining the nodal displacement vector, velocity vector and acceleration vector for each time step through iterative calculation.

[0076] The time integration method is used to numerically solve the structural dynamics control relationship model, that is, to solve for the displacement, velocity, and acceleration responses of the structure step by step in the time domain. The integration time step is an important parameter in the time integration method. Setting the integration time step to be consistent with the initial step size and the time interval of the initial input feature sequence can ensure the accuracy and stability of the calculation. Iterative calculation is the core process of the time integration method. By continuously updating the displacement, velocity, and acceleration vectors of the nodes, the actual response is gradually approximated.

[0077] Specifically, based on the selected time integration method (such as the Newmark-β method, Wilson-θ method, etc.), and combined with the mass matrix, stiffness matrix, and damping matrix of the structural dynamics control relationship model, an iterative calculation formula is established. Within each time step, the response at the next time step is calculated based on the nodal displacement vector, velocity vector, and acceleration vector at the current moment. The specific process is as follows: First, according to the formula of the time integration method, the displacement, velocity, and acceleration at the current moment are substituted to calculate the intermediate variables; then, according to the dynamic equations of the structural dynamics control relationship model, combined with the intermediate variables, the acceleration vector at the next time step is solved; next, according to the acceleration vector and the formula of the time integration method, the velocity vector and displacement vector at the next time step are calculated. By continuously repeating this process, the nodal displacement vector, velocity vector, and acceleration vector at each time step are obtained.

[0078] Step S325: Perform frequency domain transformation on the nodal displacement vector, velocity vector and acceleration vector, calculate the vibration amplitude and phase angle under different frequency components, and generate a structural vibration response feature set containing amplitude frequency characteristic curves and phase frequency characteristic curves.

[0079] Frequency domain transformation is the process of converting time-domain signals into frequency-domain signals. By performing frequency domain transformation on nodal displacement vectors, velocity vectors, and acceleration vectors, the vibration responses in the time domain can be converted into frequency components in the frequency domain. The vibration amplitude at different frequency components represents the vibration intensity of that frequency component, and the phase angle represents the vibration phase of that frequency component. The amplitude-frequency response curve describes the relationship between vibration amplitude and frequency, while the phase-frequency response curve describes the relationship between phase angle and frequency. The structural vibration response feature set is a feature set obtained by integrating the amplitude-frequency response curve and the phase-frequency response curve, which can comprehensively reflect the vibration frequency characteristics of the reservoir-dam group structure.

[0080] Specifically, Fourier transform (such as Fast Fourier Transform, FFT) is used to perform frequency domain transformation on the nodal displacement vector, velocity vector, and acceleration vector. The time-domain vibration response data is used as input, and after Fourier transform, a complex spectrum in the frequency domain is obtained. For each frequency component in the complex spectrum, its amplitude and phase angle are calculated. The amplitude can be obtained by calculating the modulus of the complex number, and the phase angle can be obtained by calculating the argument of the complex number. Then, the vibration amplitudes and phase angles of different frequency components are arranged in frequency order, and amplitude-frequency response curves and phase-frequency response curves are plotted respectively. Finally, the amplitude-frequency response curves and phase-frequency response curves are integrated to generate a structural vibration response feature set.

[0081] Step S330: Based on the structural vibration response characteristics, construct a deformation transfer function in the deformation transmission dynamics model module. The deformation transfer function is used to describe the amplitude attenuation law and phase change characteristics of deformation propagating from the excitation source location to the surrounding area.

[0082] The structural vibration response characteristics reflect the vibration properties of the reservoir-dam group structure at different frequencies. Based on these characteristics, a deformation transfer function is constructed in the deformation propagation dynamics model module. The deformation transfer function is a function describing the deformation propagation law, considering the amplitude attenuation and phase change of deformation during propagation. The excitation source location refers to the origin of the deformation in the reservoir-dam group, which may be due to factors such as earthquakes or water level changes. When deformation propagates from the excitation source location to the surrounding area, its amplitude attenuates with increasing propagation distance, and its phase also changes. Specifically, the propagation law of deformation at different frequencies is first analyzed based on the amplitude-frequency characteristic curves and phase-frequency characteristic curves in the structural vibration response characteristics. The relationship between amplitude attenuation and phase change and factors such as propagation distance and frequency can be established by fitting experimental data or theoretical models. Then, these relationships are integrated into a function to construct the deformation transfer function. For example, through the analysis of a large amount of experimental data, it was found that the amplitude attenuation of deformation is inversely proportional to the square of the propagation distance, and the phase change is directly proportional to the propagation distance and frequency. Based on these relationships, a deformation transfer function that includes propagation distance and frequency can be constructed. During the construction process, the influence of factors such as the material properties and structural characteristics of the reservoir-dam complex on deformation propagation can also be considered. For example, the damping characteristics of different materials will affect the amplitude attenuation of deformation, and the geometry of the structure will affect the deformation propagation path.

[0083] Step S340: Perform topological modeling of the reservoir-dam group structure, abstract the key parts of the reservoir-dam group structure as graph nodes, and use the deformation transfer intensity between adjacent parts as edge weights to generate the topological graph of the reservoir-dam group structure.

[0084] Topological modeling is the process of abstracting a reservoir-dam group structure into a graph structure. It provides a more intuitive representation of the connections and deformation transfer between different parts of the reservoir-dam group structure. Key parts of the reservoir-dam group structure refer to those that significantly impact the safety and stability of the group, such as dam abutments, dam foundations, and the dam body. These key parts are abstracted as graph nodes, with each node representing a key part. The deformation transfer strength between adjacent parts refers to the ability of deformation to transfer from one key part to an adjacent key part, reflecting the degree of interaction and influence between the two parts. The deformation transfer strength between adjacent parts is used as the edge weight; a larger edge weight indicates a stronger deformation transfer capability between the two parts.

[0085] Specifically, firstly, based on the structural design drawings and monitoring data of the reservoir-dam group, the key components of the structure are identified. Then, based on the deformation transfer function and structural vibration response characteristics, the deformation transfer intensity between adjacent key components is calculated. The magnitude of the deformation transfer intensity can be determined by analyzing the changes in physical quantities such as displacement and stress between nodes. Finally, the key components are abstracted as graph nodes, and the deformation transfer intensity between adjacent components is used as edge weights to construct the topology graph of the reservoir-dam group structure.

[0086] Step S350: Calculate the deformation transfer efficiency of each edge in the topology graph of the reservoir-dam group structure using the deformation transfer function. Paths with deformation transfer efficiency exceeding a preset threshold are identified as critical deformation transmission paths. Critical deformation transmission paths consist of continuous graph nodes and edges and satisfy the principle of cumulative and maximum transmission efficiency.

[0087] The deformation transfer function describes the propagation law of deformation in the reservoir-dam group structure. This function can be used to calculate the deformation transfer efficiency of each edge in the topological graph of the reservoir-dam group structure. Deformation transfer efficiency refers to the ability of deformation to be transferred between two nodes connected by an edge, reflecting the importance of the edge in the deformation transfer process. A preset threshold is a pre-defined standard used to determine which edges have high deformation transfer efficiency. Paths with deformation transfer efficiency exceeding the preset threshold are identified as critical deformation transmission paths. These paths are the main channels for deformation transfer in the reservoir-dam group structure and have a significant impact on the overall deformation of the reservoir-dam group. Critical deformation transmission paths consist of continuous graph nodes and edges and satisfy the principle of maximizing the cumulative sum of transfer efficiency. This means that among all possible paths, the path with the largest cumulative sum of transfer efficiency is selected as the critical transmission path.

[0088] Specifically, for each edge in the topological graph of the reservoir-dam group structure, the deformation transfer efficiency of the edge is calculated based on the deformation transfer function and relevant information (such as distance and frequency) of the two nodes connected to the edge. The edge length, frequency, and other factors are substituted into the deformation transfer function to obtain the deformation transfer efficiency value. Then, the calculated deformation transfer efficiency is compared with a preset threshold, and edges with deformation transfer efficiencies exceeding the preset threshold are selected. Next, paths consisting of continuous graph nodes and edges are found from these edges, and the cumulative sum of transfer efficiencies for each path is calculated. Finally, the path with the largest cumulative sum of transfer efficiencies is selected as the critical deformation transmission path.

[0089] Step S400: Based on the key deformation transmission path, perform stress-deformation forward calculation and reverse source analysis in the digital twin model of the reservoir-dam group. The stress distribution status of each monitoring point on the key deformation transmission path is obtained through forward calculation. The core risk source that causes deformation transmission is located through reverse source analysis, and the influence weight of the core risk source on the deformation of the reservoir-dam group is calculated.

[0090] The critical deformation transmission path has been identified, revealing the main channels for deformation propagation within the reservoir-dam group structure. Based on this path, stress-deformation forward calculus and reverse source analysis were performed in the digital twin model of the reservoir-dam group. The stress-deformation forward calculus starts with known deformation information and uses physical models and computational methods to solve for the stress distribution at each monitoring point along the critical deformation transmission path. The reverse source analysis starts with the known stress distribution and uses mathematical models and optimization algorithms to locate the core risk sources that trigger deformation transmission and calculate their impact weights on the deformation of the reservoir-dam group.

[0091] In one implementation, step S400 may include the following steps S410-S450:

[0092] Step S410: Deploy a virtual monitoring point array along the critical deformation transmission path. The virtual monitoring point array is evenly distributed along the critical deformation transmission path, and each virtual monitoring point contains three-dimensional spatial coordinates and structural depth features.

[0093] A virtual monitoring point array is a series of virtual monitoring points set up to obtain more detailed information on the critical deformation transmission path. Evenly spaced distribution along the critical deformation transmission path ensures uniform data collection along the path. Each virtual monitoring point contains three-dimensional spatial coordinates and structural depth features. The three-dimensional spatial coordinates determine the point's location in space, while the structural depth features indicate the point's depth within the reservoir-dam complex structure.

[0094] Specifically, the spacing between virtual monitoring points is first determined based on the shape and length of the critical deformation transmission path. Then, starting from the beginning of the critical deformation transmission path, virtual monitoring points are sequentially deployed at the determined intervals. For each virtual monitoring point, its three-dimensional spatial coordinates are calculated based on its position on the path. Simultaneously, the structural depth characteristics of that point are determined based on the structural features and geological conditions of the reservoir-dam group. Through measurement and calculation, the three-dimensional spatial coordinates and structural depth characteristics of each virtual monitoring point are determined. The information from these virtual monitoring points is recorded to form a virtual monitoring point array. This array provides a more comprehensive understanding of the situation along the critical deformation transmission path, offering more detailed data for subsequent stress-deformation forward calculations and reverse source analysis.

[0095] Step S420: Perform stress-deformation forward calculation, using the deformation value in the global deformation field as input condition, inputting it into the elastoplastic constitutive model in the digital twin model of the reservoir-dam group, to obtain the principal stress value, shear stress value and stress tensor invariant of each virtual monitoring point, and generate a stress distribution state dataset.

[0096] The stress-deformation forward calculus is the process of solving for the stress distribution state starting from known deformation information. The deformation values ​​in the global deformation field already describe the deformation of the surface and interior of the reservoir-dam group, and are used as input conditions into the elastoplastic constitutive model in the digital twin model of the reservoir-dam group. The elastoplastic constitutive model is a model used to describe the stress-strain relationship of a material under stress, considering the elastic and plastic deformation characteristics of the material. By solving the equations of the elastoplastic constitutive model, the principal stress values, shear stress values, and stress tensor invariants of each virtual monitoring point can be obtained. The stress distribution state dataset is a dataset obtained by integrating the principal stress values, shear stress values, and stress tensor invariants of each virtual monitoring point, which can comprehensively reflect the stress distribution along the key deformation transmission paths.

[0097] In one implementation, step S420 may include the following steps S421 to S425:

[0098] Step S421: Spatial discretize the global deformation field and represent it as a deformation value matrix with the virtual monitoring point array as the sampling points. The row dimension of the deformation value matrix corresponds to the virtual monitoring point number, and the column dimension corresponds to the deformation component.

[0099] Spatial discretization is the process of converting a continuous global deformation field into discrete data. Through spatial discretization, the global deformation field can be represented as a deformation value matrix with an array of virtual monitoring points as sampling points. The row dimension of the deformation value matrix corresponds to the virtual monitoring point number, and the column dimension corresponds to the deformation components. This clearly represents the values ​​of different deformation components for each virtual monitoring point.

[0100] Specifically, an interpolation method is used to spatially discretize the global deformation field. A suitable interpolation algorithm (such as bilinear interpolation, cubic spline interpolation, etc.) can be selected, and the deformation values ​​in the global deformation field are interpolated to the locations of the virtual monitoring points based on the distribution of the global deformation field and the positions of the virtual monitoring points. For each virtual monitoring point, the values ​​of its different deformation components (such as the x, y, and z components of displacement, and various components of strain) are collected, and these values ​​are arranged in the order of the virtual monitoring point numbers to form one row of the deformation value matrix. The deformation values ​​of all virtual monitoring points are combined to obtain the deformation value matrix.

[0101] Step S422: Input the deformation value matrix into the elastic-plastic constitutive model module of the reservoir-dam group digital twin model, and select the Drucker-Prager yield criterion as the material yield judgment condition. The elastic-plastic constitutive model module includes an elastic stage calculation submodule and a plastic stage calculation submodule.

[0102] The elastoplastic constitutive model module is used in the digital twin model of the reservoir-dam group to calculate the stress-strain relationship of materials. The deformation matrix is ​​input into this module, which transmits the deformation information of virtual monitoring points to the elastoplastic constitutive model for calculation. The Drucker-Prager yield criterion is a commonly used material yield criterion that considers the influence of hydrostatic pressure on yielding and is applicable to materials such as soil and rock. The elastoplastic constitutive model module includes an elastic stage calculation submodule and a plastic stage calculation submodule. The elastic stage calculation submodule calculates the stress-strain relationship of the material within the elastic range, while the plastic stage calculation submodule calculates the stress-strain relationship after the material enters the plastic stage.

[0103] Specifically, the deformation matrix is ​​input into the elastoplastic constitutive model module. The module first determines whether the material has entered the plastic stage based on the Drucker-Prager yield criterion. For each virtual monitoring point, the Drucker-Prager yield function is calculated based on its deformation value and the material parameters. If the yield function value is less than zero, the material is in the elastic stage, and the elastic stage calculation submodule is called for calculation; if the yield function value is greater than or equal to zero, the material has entered the plastic stage, and the plastic stage calculation submodule is called for calculation.

[0104] Step S423: In the elastic stage calculation submodule, when the deformation value does not exceed the material yield limit, the stress tensor is calculated by the stress-strain relationship established in advance by the generalized Hooke's law. The stress tensor is obtained by multiplying the elastic stiffness matrix and the deformation value matrix.

[0105] The elastic stage calculation submodule is used to calculate the stress-strain relationship of a material within its elastic range. When the deformation value does not exceed the material's yield strength, the material is in the elastic stage. The generalized Hooke's law describes the stress-strain relationship of elastic materials, and this law can be used to pre-establish the stress-strain relationship. The stress tensor is a tensor describing the internal stress state of a material, and it can be obtained by multiplying the elastic stiffness matrix by the deformation value matrix. The elastic stiffness matrix is ​​a matrix describing the elastic properties of a material and is related to parameters such as the material's elastic modulus and Poisson's ratio.

[0106] Specifically, the first step is to determine whether the deformation value at each virtual monitoring point exceeds the material's yield strength. This can be done using the yield function value calculated based on the Drucker-Prager yield criterion. If the yield function value is less than zero, it indicates that the deformation value has not exceeded the material's yield strength. For virtual monitoring points in the elastic stage, the stress tensor at that point is obtained by multiplying the elastic stiffness matrix by the deformation value matrix, according to the generalized Hooke's law. The elastic stiffness matrix can be calculated based on parameters such as the material's elastic modulus and Poisson's ratio.

[0107] Step S424: In the plastic stage calculation submodule, when the deformation value exceeds the material yield limit, the plastic strain increment is calculated by the plastic flow law, the stress tensor is updated and the strengthening effect is considered. The strengthening effect is described by the isotropic strengthening model.

[0108] The plastic stage calculation submodule is used to calculate the stress-strain relationship of a material after it enters the plastic stage. The material enters the plastic stage when the deformation exceeds its yield strength. The plastic flow law describes the relationship between strain increment and stress state during plastic deformation, and this law can be used to calculate the plastic strain increment. The stress tensor needs to be updated based on the plastic strain increment to reflect the stress changes after the material enters the plastic stage. The strengthening effect refers to the phenomenon that the strength of a material gradually increases during plastic deformation. The isotropic strengthening model is a commonly used model to describe the strengthening effect, assuming that the yield surface of the material expands uniformly during plastic deformation.

[0109] Specifically, when the deformation value at a virtual monitoring point exceeds the material's yield strength, the calculation submodule for the plastic stage is initiated. First, based on plastic flow laws (such as correlated plastic flow laws), and combined with the current stress state and deformation value, the plastic strain increment is calculated. Then, the stress tensor is updated based on the calculated plastic strain increment. During the update process, the strengthening effect described by the isotropic strengthening model is considered. The isotropic strengthening model describes the degree of material strengthening through a strengthening parameter, which increases with increasing plastic deformation. Based on the strengthening parameter and the plastic strain increment, the size of the yield surface is adjusted, thereby updating the stress tensor.

[0110] Step S425: Solve for the principal stresses of the calculated stress tensor. Calculate the three principal stress values ​​and their corresponding principal direction angles through eigenvalue decomposition. Simultaneously calculate the first, second, and third invariants of the stress tensor. Arrange the principal stress values, shear stress values, and stress tensor invariants in the order of virtual monitoring point numbers to generate a stress distribution state dataset.

[0111] Principal stress determination is the process of finding the three principal stress values ​​and their corresponding principal direction angles from the stress tensor. The principal stress values ​​represent the maximum and minimum stresses in different directions of the material, and the principal direction angles represent the directions of the principal stresses. The first, second, and third invariants of the stress tensor are important characteristic parameters of the stress tensor, related to the mechanical properties and deformation state of the material. Arranging the principal stress values, shear stress values, and stress tensor invariants in order of virtual monitoring point numbering generates a stress distribution dataset. This dataset comprehensively reflects the stress distribution at each virtual monitoring point along the critical deformation transmission path.

[0112] Specifically, for the stress tensor calculated at each virtual monitoring point, the principal stress values ​​and principal direction angles are solved using eigenvalue decomposition. Eigenvalue decomposition is a mathematical method that obtains the principal stress values ​​and principal direction angles by solving for the eigenvalues ​​and eigenvectors of the stress tensor. Simultaneously, based on the definition and calculation formula of the stress tensor, the first, second, and third invariants of the stress tensor are calculated. Shear stress values ​​can be calculated using the principal stress values ​​and principal direction angles. Finally, the principal stress values, shear stress values, and stress tensor invariants for each virtual monitoring point are arranged in numerical order according to the virtual monitoring point number, forming a stress distribution state dataset.

[0113] Step S430: Construct a reverse source analysis model, using the stress distribution state dataset as a known quantity, and establish an objective function based on the least squares method. The objective function is used to describe the sum of squared errors between the calculated stress value and the actual monitored stress value.

[0114] The reverse source tracing analysis model is used to locate the core risk source that triggers deformation transmission. It uses a stress distribution dataset as a known quantity, which already contains the stress distribution at each virtual monitoring point along the critical deformation transmission path. An objective function is established based on the least squares method, which finds the optimal solution by minimizing the sum of squared errors. The objective function describes the sum of squared errors between the calculated stress value and the actual monitored stress value. By optimizing the objective function, the calculated result that best matches the actual monitored stress value can be found, thus locating the core risk source.

[0115] In one implementation, step S430 may include the following steps S431 to S435:

[0116] Step S431: Obtain stress monitoring data of the physical monitoring points actually deployed on the critical deformation transmission path. The physical monitoring points and some nodes in the virtual monitoring point array overlap. The stress monitoring data includes the actual measured principal stress values ​​and shear stress values.

[0117] Physical monitoring points are points actually deployed along the critical deformation transmission path to monitor stress. These points overlap with some nodes in the virtual monitoring point array, ensuring that calculated and actual monitored stress values ​​are obtained at the same locations for easy comparison. Stress monitoring data, including principal stress and shear stress values, are obtained through actual measurements at the physical monitoring points. This data is accurate and reliable, reflecting the stress state of the reservoir-dam complex structure under actual conditions.

[0118] Specifically, physical monitoring points are strategically placed along the deformation transmission path based on its location and the area requiring monitoring. Stress sensors (such as strain gauges and pressure sensors) can be used to measure stress values. For each physical monitoring point, principal stress and shear stress values ​​are periodically collected and recorded to form stress monitoring data.

[0119] Step S432: Extract the virtual monitoring point data that coincides with the physical monitoring points in the stress distribution state dataset to obtain a set of calculated stress values. The dimension of the set of calculated stress values ​​is the same as the dimension of the stress monitoring data.

[0120] The stress distribution state dataset already contains the stress distribution at each virtual monitoring point along the critical deformation transmission path. Extracting the data from virtual monitoring points that overlap with the physical monitoring points in the stress distribution state dataset involves identifying the stress data at these virtual monitoring points located in the same positions as the physical monitoring points. The calculated stress value set is obtained by integrating the extracted stress data from the virtual monitoring points. Its dimensions are the same as those of the stress monitoring data, ensuring comparisons are performed on the same dimension.

[0121] Specifically, firstly, based on the location information of the physical monitoring points, virtual monitoring points that overlap with the physical monitoring points are found in the stress distribution state dataset. For each overlapping virtual monitoring point, its principal stress value and shear stress value are extracted, and these values ​​are arranged in the order of the physical monitoring point numbers to form a set of calculated stress values.

[0122] Step S433: Define the objective function as the sum of squared errors between the set of stress values ​​and the stress monitoring data. The independent variables of the objective function are the location coordinate vector and the intensity feature vector of the potential risk source. The location coordinate vector includes coordinate components in the x, y, and z directions, and the intensity feature vector includes the load amplitude and the range of action.

[0123] The objective function is used to measure the difference between the calculated stress value and the actual monitored stress value. It is defined as the sum of squared errors between the set of calculated stress values ​​and the stress monitoring data. By minimizing this objective function, the calculated result that is closest to the actual monitored stress value can be found, thereby locating the core risk source. The independent variables of the objective function are the location coordinate vector and intensity feature vector of the potential risk source. The location coordinate vector determines the spatial position of the potential risk source, and the intensity feature vector describes the intensity and range of the potential risk source.

[0124] Specifically, the calculated stress value set and the stress monitoring data are subtracted element-by-element to obtain an error vector. Then, each element of the error vector is squared, and the results are summed to obtain the sum of squared errors. This sum of squared errors is defined as the objective function. The independent variables of the objective function are the location coordinate vector and intensity characteristic vector of the potential risk sources. By adjusting the values ​​of these independent variables, the calculated stress value set can be changed, thereby minimizing the value of the objective function.

[0125] Step S434: Introduce a regularization term into the objective function. The regularization term, in the form of L2 norm, is used to constrain the magnitude of the intensity feature vector of potential risk sources and prevent overfitting.

[0126] Regularization is a correction term introduced to prevent overfitting, which refers to a model performing well on training data but poorly on test data. In reverse source analysis, the independent variables of the objective function are the location coordinate vector and intensity feature vector of potential risk sources. Without constraints, the magnitude of the intensity feature vector may become too large, making the model overly complex and leading to overfitting. L2 norm regularization is a commonly used regularization method that constrains the magnitude of the intensity feature vector, making the model more stable.

[0127] Specifically, a regularization term is added to the objective function. This regularization term is formed by multiplying the transpose of the intensity eigenvector by the regularization coefficient matrix, and then multiplying that product by the intensity eigenvector. The regularization coefficient matrix is ​​a diagonal matrix, with its diagonal elements being the regularization coefficients, used to control the strength of the regularization. By adjusting the regularization coefficients, the weights of the sum of squared errors and the regularization term can be balanced. Adjusting the regularization coefficients allows the model to fit both calculated and actual monitored stress values ​​while constraining the magnitude of the intensity eigenvectors of potential risk sources, thus preventing overfitting.

[0128] Step S435: Represent the objective function in matrix form, where the sum of squared errors is represented by the product of the transpose of the vector difference and the vector difference itself, and the regularization term is represented by the product of the transpose of the intensity eigenvector and the regularization coefficient matrix, and then multiplied by the intensity eigenvector, thus generating the complete objective function expression.

[0129] Representing the objective function in matrix form facilitates mathematical calculations and optimization. The sum of squared errors can be represented by the product of the transpose and the vector differences, concisely expressing the difference between the calculated stress values ​​and the stress monitoring data. The regularization term is represented by the product of the transpose of the intensity eigenvector and the regularization coefficient matrix, then multiplied by the intensity eigenvector, clearly demonstrating the calculation method of the regularization term. After generating the complete objective function expression, various optimization algorithms can be used to solve it.

[0130] Specifically, firstly, the calculated stress value set and stress monitoring data are represented as vectors, and their vector difference is calculated. Then, the vector difference is transposed and multiplied by itself to obtain the matrix representation of the sum of squared errors. For the regularization term, the intensity eigenvector is represented as a vector, multiplied by the regularization coefficient matrix, and then the result is multiplied by the transpose of the intensity eigenvector to obtain the matrix representation of the regularization term. Finally, the sum of squared errors and the regularization term are added to obtain the complete objective function expression.

[0131] Step S440: Optimize the objective function using the gradient descent algorithm, iteratively adjust the location and intensity characteristics of potential risk sources until the sum of squared errors is less than the preset convergence threshold. At this point, the potential risk source is the core risk source that triggers deformation propagation.

[0132] The gradient descent algorithm iteratively adjusts the values ​​of independent variables, gradually decreasing the value of the objective function until it reaches its minimum. In reverse source analysis, the objective function is defined as the sum of squared errors between the calculated stress value set and the actual monitored stress value, plus a regularization term. The independent variables are the location and intensity characteristics of the potential risk sources. Optimizing the objective function using the gradient descent algorithm involves continuously adjusting the location and intensity of the potential risk sources to gradually reduce the sum of squared errors.

[0133] Specifically, the location and intensity features of potential risk sources are first initialized as initial values ​​for the algorithm. Then, the gradient of the objective function at the current independent variable value is calculated; the gradient represents the rate of change and direction of the objective function at that point. Based on the direction of the gradient, the location and intensity features of the potential risk sources are adjusted to reduce the value of the objective function. In each iteration, the sum of squared errors is calculated and compared with a preset convergence threshold. If the sum of squared errors is less than the preset convergence threshold, the algorithm is considered to have converged, and the potential risk source at this point is the core risk source that triggers deformation propagation.

[0134] Step S450: Calculate the impact weights of core risk sources using the analytic hierarchy process (AHP) and construct a risk source impact assessment matrix. The element values ​​of the risk source impact assessment matrix represent the relative importance of different core risk sources. Solve for the eigenvector corresponding to the largest eigenvalue of the risk source impact assessment matrix using the eigenvalue decomposition method. After normalizing the eigenvector, obtain the impact weights of each core risk source.

[0135] The Analytic Hierarchy Process (AHP) is a method used to determine the relative importance of multiple factors, allowing for the calculation of the influence weights of core risk sources. The risk source impact assessment matrix is ​​used to evaluate the relative importance among different core risk sources; the element values ​​represent the degree of relative importance between different core risk sources. The eigenvalue decomposition method, by solving for the eigenvector corresponding to the largest eigenvalue of the risk source impact assessment matrix, can obtain the relative importance ranking of each core risk source. After normalizing the eigenvectors, the influence weights of each core risk source are obtained, accurately reflecting the degree of influence of each core risk source on the deformation of the reservoir-dam group.

[0136] Specifically, firstly, based on the characteristics of the core risk sources and their impact mechanisms on the deformation of the reservoir-dam group, a risk source impact assessment matrix is ​​constructed. For each element in the matrix, the relative importance of different core risk sources is determined based on expert experience, historical data, or numerical simulation results. For example, if core risk source A has a greater impact on the deformation of the reservoir-dam group than core risk source B, the value of the corresponding element in the matrix is ​​set to greater than 1; if the impact is the same, it is set to 1; if the impact is small, it is set to less than 1. Then, the eigenvector corresponding to the largest eigenvalue of the risk source impact assessment matrix is ​​solved using the eigenvalue decomposition method. Numerical calculation software (such as MATLAB, Python's NumPy library, etc.) can be used for eigenvalue decomposition. Finally, the obtained eigenvectors are normalized so that the sum of all elements in the vector is 1, thus obtaining the impact weight of each core risk source.

[0137] Step S500: Based on the impact weight of the core risk sources, construct simulation scenarios of various control strategies in the digital twin model of the reservoir-dam group, simulate and deduce the deformation trend of the reservoir-dam group under each simulation scenario, and obtain deformation early warning information.

[0138] The impact weights of core risk sources reflect the degree of influence of each core risk source on the deformation of the reservoir-dam group. Based on these weights, simulation scenarios with various control strategies are constructed in the digital twin model of the reservoir-dam group. Control strategies are a series of measures taken against core risk sources, such as adjusting water levels and reinforcing structures. Simulation scenarios simulate the operational state of the reservoir-dam group under different control strategies. By simulating these scenarios in the digital twin model, the deformation trend of the reservoir-dam group under different conditions can be predicted. Simulating and extrapolating the deformation trend of the reservoir-dam group under each simulation scenario involves using the digital twin model to calculate the changes in physical quantities such as stress, strain, and displacement of the reservoir-dam group under different control strategies. Deformation early warning information is based on the simulation results and includes risk level classification and emergency response priorities, used to guide the safety management and decision-making of the reservoir-dam group.

[0139] In one implementation, step S500 may include the following steps S510-S550:

[0140] Step S510: Based on the impact weight of the core risk source and the key deformation transmission path, execute a multi-path intervention simulation process in the digital twin model of the reservoir-dam group, apply virtual intervention variables of different intensities to the core risk source, and calculate the rate of change of transmission efficiency of the key deformation transmission path under each combination of variables.

[0141] The multi-path intervention simulation process simulates the intervention of core risk sources within a digital twin model of a reservoir-dam group. Based on the influence weights of core risk sources and critical deformation transmission paths, targeted interventions can be implemented. Virtual intervention variables are variables set to simulate intervention measures, such as changing water levels or increasing loads. Applying virtual intervention variables of varying intensities to core risk sources simulates different levels of intervention. The rate of change of transmission efficiency of the critical deformation transmission path refers to the proportionate change in transmission efficiency of the critical deformation transmission path relative to the initial state after applying virtual intervention variables. By calculating this rate of change, the impact of different intervention measures on deformation transmission can be evaluated.

[0142] In one implementation, step S510 may include the following steps S511 to S515:

[0143] Step S511: Load the geometric topology data of the key deformation transmission path and the spatial location data of the core risk source into the digital twin model of the reservoir-dam group, and establish the spatial mapping relationship between the point of action of the intervention variable and the node of the transmission path.

[0144] The geometric topological data of the critical deformation transmission paths describe the shape, length, and connectivity of these paths, while the spatial location data of the core risk sources determine their specific positions within the reservoir-dam group structure. Loading this data into the digital twin model of the reservoir-dam group involves inputting the information about the paths and risk sources into the model, enabling it to accurately simulate their interactions. Establishing a spatial mapping between the points of action of intervention variables and the nodes of the transmission paths is crucial for determining the effective locations of intervention measures within the model, ensuring that interventions are accurately applied to the core risk sources and critical deformation transmission paths.

[0145] Specifically, firstly, the geometric topological data of the critical deformation transmission path and the spatial location data of the core risk sources are organized into suitable formats, such as text files or database tables. Then, this data is loaded into the digital twin model of the reservoir-dam cluster. This loading process can be completed through the model's interface functions or data import tools. Next, based on the mode of action of the intervention variables and the location of the core risk sources, the points of action of the intervention variables are determined. For each point of action of an intervention variable, the nearest node on the critical deformation transmission path is found, and a spatial mapping relationship is established between them. For example, in a digital twin model of a reservoir-dam cluster, the geometric topological data of the critical deformation transmission path and the spatial location data of the core risk sources are organized into text files. This data is loaded into the model using the model's data import tool. Based on the mode of action of the intervention variables, several points of action of the intervention variables are determined. The nearest nodes on the critical deformation transmission path to these points are found, and a spatial mapping relationship is established. This mapping relationship ensures that intervention measures can accurately act on the core risk sources and the critical deformation transmission path, providing a foundation for subsequent simulations.

[0146] Step S512: Set the parameter space of the virtual intervention variables. The parameter space includes the dimension of effect intensity, the dimension of effect range, and the dimension of effect mode. Each dimension contains multiple discrete value levels. Generate a set of virtual intervention variable combinations covering the parameter space through the full factorial experimental design method.

[0147] The parameter space of a virtual intervention variable refers to the range of possible values ​​for the intervention variable, including the dimensions of intensity, scope, and mode of action. The intensity dimension represents the magnitude of the intervention, such as increasing the load or changing the water level; the scope dimension represents the area affected, such as the size of the region; and the mode of action dimension represents the implementation method, such as applying the intervention all at once or gradually. Each dimension contains multiple discrete value levels. Generating a set of virtual intervention variable combinations covering the parameter space using a full factorial experimental design method involves combining different values ​​from each dimension to obtain all possible combinations of intervention variables.

[0148] Step S513: Using the variable application module of the digital twin model of the reservoir-dam group, each group of variables in the virtual intervention variable combination set is applied sequentially to the spatial location corresponding to the core risk source at a preset time interval, and the deformation response time history curves of each monitoring node on the key deformation transmission path are recorded simultaneously.

[0149] The variable application module of the reservoir-dam cluster digital twin model is used to apply virtual intervention variables into the model, accurately transmitting variable parameters to the spatial locations corresponding to the core risk sources. Applying each group of variables from the virtual intervention variable set sequentially at preset time intervals means applying different combinations of intervention variables to the model according to a specific time sequence. Simultaneously recording the deformation response time history curves of each monitoring node along the key deformation transmission path involves recording the changes in physical quantities such as displacement and strain of each monitoring node over time during the application of intervention variables. These curves reflect the deformation response process of the reservoir-dam cluster under different intervention measures.

[0150] Specifically, the variable application module of the reservoir-dam cluster digital twin model is first activated, loading a set of virtual intervention variables into the module. Then, at preset time intervals, a set of variables from the virtual intervention variable set is selected sequentially and applied to the spatial locations corresponding to the core risk sources. Simultaneously with variable application, the model's monitoring function records the changes in physical quantities such as displacement and strain at each monitoring node along the critical deformation transmission path over time, generating deformation response time history curves.

[0151] Step S514: Calculate the peak decay rate and the time reduction to reach a steady state for each deformation response time history curve. Define the weighted sum of the peak decay rate and the time reduction as the rate of change of conduction efficiency. The weight allocation of the weighted sum is determined according to the influence weight of the core risk source.

[0152] Peak decay rate refers to the proportion of the peak value of the deformation response time history curve that decreases relative to the initial state after the application of the intervention variable, reflecting the suppression effect of the intervention on the deformation peak. The reduction in time to reach steady state refers to the reduction in the time required for the deformation response time history curve to reach steady state relative to the initial state after the application of the intervention variable, reflecting the impact of the intervention on the deformation stabilization rate. The rate of change in transmission efficiency is an indicator that comprehensively considers both the peak decay rate and the reduction in time to reach steady state, enabling a comprehensive assessment of the impact of the intervention on the transmission efficiency of key deformation transmission paths. The weighting of the weighted sum is determined based on the influence weights of the core risk sources, allowing for a reasonable allocation of weights for the peak decay rate and the reduction in time to reach steady state according to the importance of different core risk sources.

[0153] Specifically, for each deformation response time history curve, the initial peak value and the time to reach a steady state are first determined. Then, the peak value and the time to reach a steady state after applying the intervention variable are calculated. The peak decay rate is calculated as (initial peak value - peak value after intervention) / initial peak value, and the reduction in the time to reach a steady state is calculated as the initial time to reach a steady state minus the time to reach a steady state after intervention. The weights of the peak decay rate and the reduction in the time to reach a steady state are determined based on the influence weights of the core risk sources. For example, if the influence weight of a core risk source is large, a higher weight may be assigned to the peak decay rate. The peak decay rate and the reduction in the time to reach a steady state are multiplied by their respective weights and then summed to obtain the rate of change in transmission efficiency.

[0154] Step S515: Establish a correlation mapping table between virtual intervention variable combinations and transmission efficiency change rate. The row dimension of the correlation mapping table corresponds to the variable combination number, and the column dimension contains the parameter values ​​of each variable and the corresponding transmission efficiency change rate calculation results.

[0155] The correlation mapping table is used to record the correspondence between combinations of virtual intervention variables and the rate of change in transmission efficiency. The rows correspond to the variable combination number, and the columns contain the parameter values ​​of each variable and the corresponding calculated rate of change in transmission efficiency. This table clearly shows the changes in transmission efficiency under different combinations of virtual intervention variables, providing a basis for subsequent selection of the optimal control strategy.

[0156] Specifically, first, an empty association mapping table is created, determining its row and column dimensions. The row dimension is determined based on the number of combinations of virtual intervention variables, while the column dimension includes parameter values ​​such as the effect strength, scope, and mode of action of the virtual intervention variables, as well as the calculated result of the transmission efficiency change rate. Then, for each group of variables in the set of virtual intervention variables, the corresponding row is located in the association mapping table based on its number. The parameter values ​​of that group of variables are filled into the corresponding column, and the calculated transmission efficiency change rate is also filled into the corresponding column.

[0157] Step S520: Based on the rate of change of conduction efficiency, select the combination of intervention variables with the best effect on inhibiting deformation conduction as the benchmark control strategy. The benchmark control strategy includes the coordinates of the intervention point, the intensity level of the effect, and the timing characteristics of the effect.

[0158] Transmission efficiency reflects the impact of different combinations of virtual intervention variables on the transmission efficiency of key deformation transmission pathways. Selecting the optimal combination of intervention variables for inhibiting deformation transmission based on the rate of change in transmission efficiency involves finding the variable combination with the largest rate of change in transmission efficiency from the association mapping table. The baseline control strategy is a specific control strategy extracted from the selected optimal combination of intervention variables, including the coordinates of the intervention point, the level of action intensity, and the timing characteristics of the action. These characteristics clarify the specific implementation method of the control strategy.

[0159] Specifically, firstly, the transmission efficiency change rates in the association mapping table are sorted to identify the variable combination number with the largest change rate. Then, the parameter value of the corresponding variable combination is found in the association mapping table based on this number. For the coordinates of the intervention point, the specific location of the intervention is determined based on the previously established spatial mapping relationship between the intervention variable point and the transmission path node; for the intensity level, the specific intensity level is directly obtained from the parameter value of the variable combination; for the timing characteristics, the timing of the intervention is determined based on the time sequence and time interval of the variables applied in the virtual intervention variable combination set.

[0160] Step S530: Based on the baseline control strategy, by changing the distribution density of the intervention points along the key deformation transmission path, adjusting the gradient change mode of the action intensity level, and optimizing the phase difference parameter of the action timing, multiple sets of derivative control strategies containing parameter perturbations are generated. Each set of derivative control strategies corresponds to an independent simulation scenario.

[0161] The baseline control strategy has identified a relatively effective control method. Based on this, parameter perturbations are performed—adjusting the distribution density of intervention points, the gradient change pattern of the intensity level, and the phase difference parameter of the intervention timing—to explore more possible control strategies. Multiple sets of derived control strategies, including parameter perturbations, are a series of control strategies obtained by changing the parameters of the baseline control strategy to varying degrees. Each set of derived control strategies corresponds to an independent simulation scenario. By simulating these scenarios in a digital twin model, the effectiveness of different control strategies can be further evaluated.

[0162] Specifically, regarding the distribution density of intervention points along the critical deformation transmission path, the number of intervention points can be increased or decreased, altering their distribution interval along the path. For the gradient change mode of the intervention intensity level, the variation of the intensity with location or time can be altered. For example, if the intensity is uniformly distributed in the baseline control strategy, it can be adjusted to a linearly increasing or decreasing gradient change mode. Regarding the phase difference parameter of the intervention timing, the time difference between different intervention points can be changed. For example, if all intervention points act simultaneously in the baseline control strategy, it can be adjusted to have a certain time delay between different intervention points. By combining and adjusting these parameters in different ways, multiple sets of derived control strategies are generated. Each set of derived control strategies is assigned an independent number, corresponding to an independent simulation scenario.

[0163] Step S540: Input the control strategy parameters of each simulation scenario into the digital twin model of the reservoir-dam group, run the deformation trend simulation process for a preset duration, and simultaneously collect the spatial distribution dynamic data and time series change data of the global deformation field during the simulation process.

[0164] The control strategy parameters for each simulation scenario already contain specific information about the derived control strategies. Inputting these parameters into the digital twin model of the reservoir-dam group means passing the parameters of different control strategies into the model, allowing the model to simulate the operation of the reservoir-dam group under these strategies. Running a deformation trend simulation process for a preset duration involves calculating the changes in physical quantities such as stress, strain, and displacement of the reservoir-dam group within the preset time in the digital twin model. The spatial distribution dynamic data of the global deformation field describes the deformation distribution of the reservoir-dam group's surface and interior at different times, while the time series data records the magnitude and trend of deformation at different time points. Simultaneous acquisition of these data allows for a comprehensive understanding of the deformation trends of the reservoir-dam group under different control strategies.

[0165] Step S550: Perform fusion analysis on spatial distribution dynamic data and time series change data, extract trend characteristic parameters such as deformation diffusion rate, deformation range expansion acceleration and deformation of key monitoring points, and generate deformation early warning information including risk level classification and emergency response priority when any trend characteristic parameter reaches the preset early warning threshold.

[0166] Spatial distribution dynamic data and time series variation data describe the deformation of the reservoir-dam group from spatial and temporal perspectives, respectively. Fusion analysis involves comprehensively processing these two types of data to extract key characteristic parameters reflecting the deformation trend of the reservoir-dam group. Deformation diffusion rate represents the speed at which deformation propagates in space; deformation range expansion acceleration represents the change in the speed at which the deformation range expands in space; and deformation at key monitoring points represents the magnitude of deformation in critical parts of the reservoir-dam group. Preset warning thresholds are pre-defined standards used to determine whether the deformation of the reservoir-dam group reaches a dangerous level. When any trend characteristic parameter reaches the preset warning threshold, it indicates that the deformation of the reservoir-dam group may pose a threat to its safety, requiring the generation of deformation warning information. Deformation warning information includes risk level classification and emergency response priorities, providing clear guidance for the safety management and decision-making of the reservoir-dam group.

[0167] Specifically, the spatially distributed dynamic data and time-series change data are first preprocessed, such as through data cleaning and interpolation, to improve data quality and consistency. Then, based on the spatially distributed dynamic data, the boundaries and extent of deformation regions at different times are calculated. By comparing the deformation extent at adjacent times, the deformation diffusion rate and deformation extent expansion acceleration are calculated. For the deformation of key monitoring points, the deformation values ​​of these points are extracted from the time-series change data. The calculated deformation diffusion rate, deformation extent expansion acceleration, and key monitoring point deformation are compared with preset warning thresholds. If any trend characteristic parameter reaches the preset warning threshold, a comprehensive risk index is calculated based on the number of elements exceeding the threshold and the extent of the exceedance. Risk levels are then classified according to the range of the comprehensive risk index, such as low risk, medium risk, and high risk. Emergency response priorities are determined based on the influence weight of core risk sources; core risk sources with higher influence weights have higher emergency response priorities. Finally, deformation warning information containing risk level identifiers, priority rankings, and key monitoring point coordinates is generated.

[0168] The various algorithms involved in the above descriptions of the embodiments of this invention can all be obtained from relevant content in the prior art. To save space, they will not be elaborated on in the embodiments of this invention. In addition, those skilled in the art can supplement the details based on common knowledge in the art when implementing the solutions of this invention. For example, they can use standardization or normalization to eliminate dimensional conflicts before feature fusion (e.g., before feature fusion, splicing, and addition, features of different units are normalized first, and then fused), use interpolation to eliminate dimensional differences, reasonably set thresholds based on historical data, experience, or business scenario requirements, train the model based on a general model training method, set the number of layers in the model structure based on actual needs, select activation functions, etc. This invention will not provide redundant descriptions of overly detailed implementation processes here.

[0169] Please see Figure 2 , Figure 2 This is a schematic diagram of a monitoring system provided in an embodiment of the present invention. The monitoring system includes at least a processor 101, a communication interface 102, and a memory 103. The processor 101, communication interface 102, and memory 103 can be connected via a bus or other means. The processor 101 (or Central Processing Unit, CPU) is the computing and control core of the monitoring system, capable of parsing various instructions and processing various data within the monitoring system. The communication interface 102 may optionally include a standard wired interface or a wireless interface (such as Wi-Fi, mobile communication interface, etc.), and can be used to send and receive data under the control of the processor 101; the communication interface 102 can also be used for data transmission and interaction within the monitoring system. The memory 103 is a memory device in the monitoring system used to store programs and data. It is understood that the memory 103 here can include the built-in memory of the monitoring system, or it can include extended memory supported by the monitoring system. The memory 103 provides storage space, which stores the operating system of the monitoring system; this invention does not limit this.

[0170] In one embodiment, the processor 101 executes the BeiDou multi-source fusion-based reservoir and dam group deformation monitoring method provided above in the embodiments of the present invention by running a computer program in the memory 103.

Claims

1. A method for monitoring deformation of a dam group based on multi-source fusion of Beidou, characterized in that, The method comprises: receiving satellite observation signals of a reservoir dam group monitoring area and sensing data returned by a plurality of source sensing devices arranged in the reservoir dam group monitoring area, unifying the satellite observation signals and the sensing data in time and space reference to obtain multi-source input data; inputting the multi-source input data into a preset reservoir dam group digital twin model, simulating physical behavior of the reservoir dam group structure to generate a global deformation field representing surface and internal deformation state of the reservoir dam group; inputting the global deformation field as an initial boundary condition into a deformation conduction dynamics model module in the reservoir dam group digital twin model, simulating the transmission process of deformation in the reservoir dam group structure, and identifying a key deformation conduction path existing in the reservoir dam group; based on the key deformation conduction path, performing stress-deformation forward calculation and reverse tracing analysis in the reservoir dam group digital twin model, obtaining stress distribution state of each monitoring point on the key deformation conduction path through forward calculation, locating a core risk source causing deformation conduction through reverse tracing analysis, and calculating an influence weight of the core risk source on deformation of the reservoir dam group; according to the influence weight of the core risk source, constructing simulation scenarios of a plurality of regulation strategies in the reservoir dam group digital twin model, simulating and deducing deformation trend of the reservoir dam group under each simulation scenario to obtain deformation warning information.

2. The method of claim 1, wherein, The method comprises: performing feature layering on the multi-source input data, extracting spatial geometric features, time sequence features and physical field correlation features in the multi-source input data, and generating a structured feature set containing a plurality of feature channels; wherein feature layering is a process of classifying and extracting multi-source input data according to different feature types, so as to decompose complex multi-source data into features with different physical meanings; inputting the structured feature set into a multi-scale physical field coupling module of the reservoir dam group digital twin model, solving through coupling relationship between elasticity and fluid dynamics to obtain deformation response features of the reservoir dam group structure under the action of multiple physical fields; based on the deformation response features, performing adaptive grid densification on the reservoir dam group structure in a finite element grid division module of the reservoir dam group digital twin model to generate a layered grid model containing different accuracy levels; performing physical behavior simulation on the surface and internal structure of the reservoir dam group through the layered grid model, collecting displacement vectors and strain tensors of each grid node, and constructing a deformation state data set covering the whole reservoir dam group; performing time and space interpolation on the deformation state data set to generate a continuously distributed global deformation field, and the spatial resolution of the global deformation field is the same as the highest accuracy level of the layered grid model.

3. The method of claim 2, wherein, The method comprises: performing feature layering on the multi-source input data, extracting spatial geometric features, time sequence features and physical field correlation features in the multi-source input data, and generating a structured feature set containing a plurality of feature channels; wherein feature layering is a process of classifying and extracting multi-source input data according to different feature types, so as to decompose complex multi-source data into features with different physical meanings; Modal separation is performed on the multi-source input data to decompose the multi-source input data into satellite observation modal data and sensing device modal data, the satellite observation modal data including carrier phase observations and pseudo-range observations, and the sensing device modal data including strain sensing data, seepage pressure data and inclination sensing data; Spatial geometric features are extracted from the satellite observation modal data, three-dimensional coordinate variation of a monitoring point is calculated through carrier phase difference processing, and a spatial coordinate time series is constructed in combination with a Beidou satellite ephemeris feature, the spatial geometric features including amplitude and direction angle features of the coordinate variation; Time series features are extracted from the sensing device modal data, time series decomposition is performed on the strain sensing data, seepage pressure data and inclination sensing data to separate trend component, periodic component and random disturbance component, and a multi-dimensional time series feature vector is generated; A preset physical field correlation feature extraction rule is obtained, cross correlation between the spatial geometric features and the time series features is obtained according to material attribute features and geological condition features of the dam group structure, and a physical field correlation feature matrix including elastic modulus correlation features, Poisson's ratio influence features and permeability coefficient coupling features is generated; The spatial geometric features, the time series feature vector and the physical field correlation feature matrix are spliced according to channel dimensions to generate a structured feature set with three-dimensional feature dimensions of space, time and physical field, and the product of the channel number and the feature dimension of the structured feature set is equal to the sum of the channel numbers of each single feature.

4. The method of claim 3, wherein, The structured feature set is input into a multi-scale physical field coupling module of the dam group digital twin model, a deformation response feature of the dam group structure under the action of multiple physical fields is obtained through the coupling relationship of elastic mechanics and fluid dynamics, including: The structured feature set is input into a feature distribution layer of the multi-scale physical field coupling module, the structured feature set is distributed to an elastic mechanics calculation branch and a fluid dynamics calculation branch according to the feature dimensions, the elastic mechanics calculation branch receives solid mechanics features in the spatial geometric features and the physical field correlation feature matrix, and the fluid dynamics calculation branch receives seepage features in the time series feature vector and the physical field correlation feature matrix; In the elastic mechanics calculation branch, a three-dimensional elastic constitutive relationship constructed based on Hooke's law is used, three-dimensional coordinate variation in the spatial geometric features is taken as a displacement boundary condition, the three-dimensional elastic constitutive relationship is input for finite element solving, and a structural stress distribution feature tensor is generated; In the fluid dynamics calculation branch, a seepage field control relationship is constructed, seepage pressure data in the time series feature vector is taken as an initial condition, the seepage field control relationship is solved, and a pore water pressure distribution feature matrix is generated; A fluid-structure coupling interface processing layer is constructed, interface data exchange is performed between the structural stress distribution feature tensor and the pore water pressure distribution feature matrix, a two-way coupling relationship between structure deformation and seepage field distribution is updated through iterative calculation until a preset convergence condition is met. According to the converged structural stress distribution feature tensor and the pore water pressure distribution feature matrix, a comprehensive deformation response feature of the reservoir dam group structure under the action of multiple physical fields is obtained, and a dimension of the deformation response feature matches a number of feature channels of the structural feature set.

5. The method of claim 1, wherein, The global deformation field is input into a deformation conduction dynamics model module in the reservoir dam group digital twin model as initial boundary conditions, a deformation transmission process in the reservoir dam group structure is simulated, and a key deformation conduction path existing in the reservoir dam group is identified, including: Boundary conditions of the global deformation field are extracted, displacement constraint boundaries and force boundary conditions of the reservoir dam group structure are determined, displacement values of the displacement constraint boundaries and stress values of the force boundary conditions are taken as initial input features, the initial input features are loaded into a structural dynamics analysis model of the deformation conduction dynamics model module, time-domain solving of the structural dynamics analysis model is performed, and structural vibration response features at different times are obtained; Based on the structural vibration response features, a deformation transmission function is constructed in the deformation conduction dynamics model module, the deformation transmission function is used to describe amplitude attenuation rules and phase change characteristics of deformation propagation from an excitation source position to surrounding areas; Topology modeling of the reservoir dam group structure is performed, key parts of the reservoir dam group structure are abstracted as graph nodes, and deformation transmission strengths between adjacent parts are taken as edge weights, and a reservoir dam group structure topology graph is generated; Deformation transmission efficiencies of edges in the reservoir dam group structure topology graph are calculated through the deformation transmission function, and a path with a deformation transmission efficiency higher than a preset threshold is determined as a key deformation conduction path, the key deformation conduction path is composed of continuous graph nodes and edges and meets a cumulative maximum principle of transmission efficiency. The global deformation field is input into a deformation conduction dynamics model module in the reservoir dam group digital twin model as initial boundary conditions, a deformation transmission process in the reservoir dam group structure is simulated, and a key deformation conduction path existing in the reservoir dam group is identified, including:

6. The method of claim 5, wherein, Boundary conditions of the global deformation field are extracted, displacement constraint boundaries and force boundary conditions of the reservoir dam group structure are determined, displacement values of the displacement constraint boundaries and stress values of the force boundary conditions are taken as initial input features, the initial input features are loaded into a structural dynamics analysis model of the deformation conduction dynamics model module, time-domain solving of the structural dynamics analysis model is performed, and structural vibration response features at different times are obtained; Based on the structural vibration response features, a deformation transmission function is constructed in the deformation conduction dynamics model module, the deformation transmission function is used to describe amplitude attenuation rules and phase change characteristics of deformation propagation from an excitation source position to surrounding areas; Topology modeling of the reservoir dam group structure is performed, key parts of the reservoir dam group structure are abstracted as graph nodes, and deformation transmission strengths between adjacent parts are taken as edge weights, and a reservoir dam group structure topology graph is generated; Deformation transmission efficiencies of edges in the reservoir dam group structure topology graph are calculated through the deformation transmission function, and a path with a deformation transmission efficiency higher than a preset threshold is determined as a key deformation conduction path, the key deformation conduction path is composed of continuous graph nodes and edges and meets a cumulative maximum principle of transmission efficiency. ​ The displacement constraint boundary condition feature set and the force boundary condition feature set are spatio-temporally aligned so that all features have a unified time stamp and spatial coordinate system, generating an initial input feature sequence of the deformation conduction dynamics model module, and a time interval of the initial input feature sequence is the same as a time resolution of the global deformation field.

7. The method of claim 6, wherein, The initial input feature is loaded into a structural dynamics analysis model of the deformation conduction dynamics model module, and the structural dynamics analysis model is solved in the time domain to obtain structural vibration response features at different times, including: An existing finite element dynamics model of the reservoir dam group structure is obtained, and element types of the finite element dynamics model include solid elements, shell elements and contact elements, the solid elements are used to simulate a main structure of the reservoir dam, the shell elements are used to simulate a dam surface protection layer, and the contact elements are used to simulate a contact behavior between the dam body and the bedrock; The initial input feature is loaded onto boundary nodes of the finite element dynamics model, the displacement constraint boundary condition feature set is applied to node degrees of freedom of a fixed boundary region, and the force boundary condition feature set is applied to a node load vector of a free boundary region; A structural dynamics control relationship model is established based on the finite element dynamics model, and the structural dynamics control relationship model includes a mass matrix, a stiffness matrix and a damping matrix; The structural dynamics control relationship model is numerically solved by a time integration method, an integral time step is set to be coordinated with a time interval of the initial input feature sequence, and node displacement vectors, velocity vectors and acceleration vectors at each time step are obtained through iterative calculation; The node displacement vectors, the velocity vectors and the acceleration vectors are converted into a frequency domain to calculate vibration amplitudes and phase angles at different frequency components, and a structural vibration response feature set including amplitude-frequency characteristic curves and phase-frequency characteristic curves is generated.

8. The method of claim 1, wherein, Based on the deformation key conduction path, stress-deformation forward calculation and reverse tracing analysis are performed in the reservoir dam group digital twin model, stress distribution states of each monitoring point on the deformation key conduction path are obtained through forward calculation, a core risk source causing deformation conduction is located through reverse tracing analysis, and an influence weight of the core risk source on the deformation of the reservoir dam group is calculated, including: A virtual monitoring point array is arranged on the deformation key conduction path, the virtual monitoring point array is distributed at equal intervals along the deformation key conduction path, and each virtual monitoring point includes a three-dimensional spatial coordinate and a structural depth feature; Deformation values in the global deformation field are input into an elastoplastic constitutive model in the reservoir dam group digital twin model as input conditions, and principal stress values, shear stress values and stress tensor invariants of each virtual monitoring point are obtained, generating a stress distribution state data set; The stress distribution state data set is taken as a known quantity, a target function is established based on a least square method, and the target function is used to describe a sum of squares of errors between calculated stress values and actual monitoring stress values; The target function is optimized and solved, and position features and intensity features of potential risk sources are iteratively adjusted until the sum of squares of errors is less than a preset convergence threshold. An influence weight of the core risk source is calculated, a risk source influence evaluation matrix is constructed, an element value of the risk source influence evaluation matrix represents a relative importance between different core risk sources, an eigenvector corresponding to a maximum eigenvalue of the risk source influence evaluation matrix is solved by an eigenvalue decomposition method, and the influence weight of each core risk source is obtained after normalization of the eigenvector.

9. The method of claim 8, wherein, The deformation value in the global deformation field is input as an input condition into an elastoplastic constitutive model in the reservoir dam group digital twin model, to obtain a principal stress value, a shear stress value and a stress tensor invariant of each virtual monitoring point, and generate a stress distribution state data set, including: The global deformation field is spatially discretized, and the global deformation field is represented as a deformation value matrix with the virtual monitoring point array as sampling points, a row dimension of the deformation value matrix corresponds to a virtual monitoring point number, and a column dimension corresponds to a deformation component; The deformation value matrix is input into an elastoplastic constitutive model module of the reservoir dam group digital twin model, and the elastoplastic constitutive model module includes an elastic stage calculation submodule and a plastic stage calculation submodule; In the elastic stage calculation submodule, when the deformation value does not exceed a material yield limit, a stress tensor is calculated through a pre-established stress-strain relationship, and the stress tensor is obtained by multiplying an elastic stiffness matrix and a deformation value matrix; In the plastic stage calculation submodule, when the deformation value exceeds the material yield limit, a plastic strain increment is calculated, the stress tensor is updated and a strengthening effect is considered, and the strengthening effect is described by an isotropic hardening model; The principal stress is solved from the calculated stress tensor, three principal stress values and corresponding principal direction angles are calculated by eigenvalue decomposition, a first invariant, a second invariant and a third invariant of the stress tensor are calculated, the principal stress values, the shear stress values and the stress tensor invariants are arranged in order of the virtual monitoring point numbers, and a stress distribution state data set is generated.

10. A monitoring system, characterized by The memory has stored therein a computer program; The processor is configured to load the computer program to implement the reservoir dam group deformation monitoring method based on Beidou multi-source fusion according to any one of claims 1-9. ​

Images