A method and system for whole-process seepage field simulation and dynamic updating

Abstract

The application discloses a kind of whole process seepage field simulation and dynamic updating method and system.The application divides whole process into several stages by engineering event table, updates the geometric state, boundary condition and source and sink item of seepage numerical calculation model according to stage, and executes non-steady seepage forward calculation;The seepage field state at the end of last stage is initial condition above current stage, and the initial value of inversion parameter is inherited from the optimal parameter calibrated after last stage;Objective function is constructed based on head and leakage observation data, and adaptive trust region quasi-newton optimization method is used for parameter dynamic calibration, and the contribution proportion of each recharge source to leakage is calculated by progressive source activation / overlay decomposition method.The application realizes whole process seepage field continuous simulation, parameter dynamic calibration and leakage source quantitative diagnosis, and improves simulation accuracy and interpretability.

Description

Technical Field

[0001] This invention belongs to the field of seepage analysis and numerical calculation technology in water conservancy and hydropower engineering, and particularly relates to a seepage field simulation method and dynamic update system for the entire process of natural-exploration-construction-operation. Background Technology

[0002] During the natural state exploration, construction excavation, formation of seepage prevention and drainage systems, and subsequent reservoir impoundment and operation of water conservancy and hydropower projects, the seepage field in the site area is in a state of continuous evolution. Engineering disturbances will cause time-varying adjustments to model geometry, boundary conditions, and source-sink terms, thereby inducing cavern / tunnel leakage, abnormal seepage pressure (pore pressure) or uplift pressure, and local stability risks. Existing evaluations are mostly based on single static conditions or steady-state assumptions, which often fail to reflect the dynamic changes throughout the entire process; if seepage parameters and boundary conditions are not effectively calibrated, the evaluation error may increase significantly.

[0003] The following shortcomings still exist in existing engineering applications:

[0004] (1) Static working conditions or single observation inversion leads to non-uniqueness or insufficient representativeness: It is difficult to guarantee the uniqueness of parameter inversion by using only water head / seepage pressure monitoring data; if parameter inversion is carried out only during relatively stable periods, it is difficult to characterize the dynamic changes caused by engineering activities such as excavation, grouting, formation of seepage prevention and drainage system, water storage and operation.

[0005] (2) The continuous changes in the boundary and structure throughout the process make it difficult to update the model: the advancement of the excavation face, the connection of the tunnel, the implementation of the drainage system, the formation of the anti-seepage curtain, and the time-varying process of the reservoir water level will cause the seepage boundary to change continuously; if a staged and updatable whole-process model cannot be established, the simulation results will be difficult to match the actual measurements.

[0006] (3) Boundary conditions change frequently with the progress of the project and lack unified records, which can easily lead to inconsistencies between the calculation conditions and the calibration comparison conditions: the excavation advance / through, lining water stop, anti-seepage curtain and drainage facility operation, and reservoir water level changes will cause multiple adjustments to the tunnel wall boundary, drainage boundary and reservoir water level boundary; if there is a lack of boundary update records and management corresponding to the project events, the boundary configuration adopted by the model and the monitoring comparison conditions are prone to deviation, thus affecting the reliability of the dynamic calibration results. Summary of the Invention

[0007] To address the challenges of simultaneously managing geometric / boundary updates, parameter stage inheritance and dynamic calibration, and leakage source identification and quantitative diagnosis under the continuous evolution of hydropower projects during natural exploration, construction excavation, gradual formation of seepage prevention and drainage systems, water storage, and operation, this invention provides a method and system for full-process seepage field simulation and dynamic updating. Through staged model updates driven by engineering event tables, parameter stage inheritance, adaptive trust region quasi-Newton optimization, and source-by-source activation / superposition decomposition numerical decomposition, the invention achieves continuous simulation of the full-process seepage field, dynamic parameter calibration, and quantitative diagnosis of leakage sources.

[0008] According to one aspect of the present invention, a method for simulating and dynamically updating a seepage field throughout the entire process is provided, comprising: step S1, establishing a numerical calculation model of seepage throughout the entire process of a hydropower project, covering the natural exploration stage, construction and excavation stage, seepage prevention and drainage system construction stage, reservoir impoundment stage, and operation stage, and setting the geometric state, boundary conditions, and source and sink terms corresponding to each stage; step S2, determining and initializing the seepage parameters to be inverted, selecting observation points and constructing an objective function, wherein the observation points include head measurement points and leakage measurement points; step S3, dividing the entire process time domain into several stages based on an engineering event table, updating the geometric state, boundary conditions, and / or source and sink terms of the numerical calculation model of seepage throughout the entire process according to the stages, and performing unsteady seepage forward modeling to obtain the seepage at each stage. The flow field simulation results are as follows: The current stage of unsteady seepage forward modeling uses the seepage field state at the end of the previous stage as the initial condition, and the initial values ​​of the seepage parameters to be inverted in the current stage are inherited from the updated seepage parameters to be inverted in the previous stage. Step S4: Based on the seepage field simulation results and actual seepage field observation information, a residual vector is constructed and the objective function value is calculated. The sensitivity of the residual vector to the seepage parameters to be inverted is calculated to obtain the Jacobian matrix. Step S5: The seepage parameters to be inverted are updated using the adaptive trust region quasi-Newton optimization method, and convergence is determined. If the convergence condition is not met, steps S3 to S5 are repeated. Step S6: When the convergence condition is met, the optimal estimate of the seepage parameters to be inverted and the full-process seepage field calculation results are output.

[0009] As a further technical solution, step S6 also includes: using the source-by-source activation / overlay decomposition method to calculate the contribution ratio of the boundary conditions or source-sink terms corresponding to each preset supply source to the leakage of the target cavern or tunnel.

[0010] As a further technical solution, the phased update in step S3 is driven by an engineering event table, which records at least the event type, scope of application, and timestamp.

[0011] As a further technical solution, step S5 includes: in the j-th iteration, solving the parameter increment under the constraint of the trust region radius to satisfy the optimization of the local quadratic model, and calculating the return rate after obtaining candidate parameters; accepting candidate parameters and increasing or maintaining the trust region radius when the return rate is greater than a preset threshold, rejecting candidate parameters and shrinking the trust region radius when the return rate is less than or equal to the preset threshold, and resolving the parameter increment under the updated trust region constraint until the acceptance condition is met or the preset iteration limit is reached.

[0012] As a further technical solution, the source-by-source activation / overlay decomposition includes: when only a single source is activated and all candidate sources other than the current single source are deactivated, performing unsteady flow forward modeling to obtain the leakage amount corresponding to the source, and normalizing the leakage amount of the single source relative to the total leakage amount obtained when all sources are activated at the same time, so as to obtain the contribution ratio of each source to the leakage amount of the target chamber or tunnel.

[0013] As a further technical solution, in step S3, the changes in geometric state, boundary conditions and / or source and sink terms corresponding to each staged update are stored in association with the corresponding engineering events to form a versioned record of model configuration.

[0014] As a further technical solution, step S5 includes: selecting samples in the parameter space using orthogonal design and performing forward modeling, extracting the head and leakage response features at the observation point to form training data, training a neural network to establish a mapping relationship between parameters and observed responses to form a surrogate model, and using a genetic algorithm to pre-optimize or accelerate the optimization of the objective function on the surrogate model to obtain initial parameters or candidate parameter combinations for optimization iteration in step S5.

[0015] According to one aspect of the present invention, a system for simulating and dynamically updating the seepage field throughout the entire process is provided, comprising: The full-process model management unit is used to build and maintain a seepage numerical calculation model that can be updated as the engineering stage changes. The model can express the geometric state, material partitioning, boundary conditions and source-sink configuration corresponding to different stages. The phase scheduling and model update unit is used to trigger phased updates of the model's geometric state, boundary conditions, and / or source and sink terms based on the project progress and project events, so as to form a phased model configuration consistent with the actual project progress. The parameter configuration and target construction unit is used to determine the initial values ​​of the seepage parameters to be inverted and establish value constraints, and to construct the objective function based on the head measurement point and the leakage measurement point; The unsteady seepage calculation unit is used to perform unsteady seepage forward modeling under given stage configuration and parameter values, and outputs seepage field response results and leakage process lines of the target cavern or tunnel. The error assessment and sensitivity analysis unit is used to construct the residual vector and calculate the objective function value, and obtain sensitivity information through parameter perturbation to form the Jacobian matrix; The parameter adaptive update unit is used to update the seepage parameters to be inverted using an adaptive trust region quasi-Newton optimization strategy, and to perform acceptance / rejection judgment and adaptive adjustment of the trust region radius until the convergence condition is met. The leakage source analysis unit is used to identify and decompose the boundary conditions or source-sink terms corresponding to different candidate supply sources, and output the proportion of leakage sources and their evolution patterns with different stages. The results summary and evolution output unit is used to output the optimal estimate of the seepage parameters to be inverted and their stage evolution results, as well as the full-process seepage field response results and the proportion of leakage sources.

[0016] According to one aspect of the present invention, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method for simulating and dynamically updating a full-process seepage field.

[0017] According to one aspect of the present invention, a non-transitory computer-readable storage medium is provided, on which computer instructions are stored, which, when executed by a processor, cause a computer to perform the method for simulating and dynamically updating a full-process seepage field.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention takes the evolution of the seepage field in the entire process of hydropower construction and operation as the main line, and integrates natural exploration, construction excavation, formation of seepage prevention and drainage system, water storage process and operation stage into the same unstable calculation framework. Based on the progress of the project and the project events, the model geometry, boundary conditions and source and sink terms are updated in stages, so that the simulation can continuously reflect the spatiotemporal response of the seepage field caused by project disturbance and reservoir water level changes.

[0019] (2) The present invention introduces monitoring data to carry out dynamic calibration in each stage, and inherits the optimal parameters obtained by calibration in this stage as the initial parameters of the next stage. At the same time, the seepage field state at the end of the previous stage is used as the initial condition for the unstable calculation of the next stage, thereby forming a parameter evolution sequence of "intra-stage calibration - inter-stage inheritance", reducing the representativeness and multiple solution risk caused by a single static inversion, and improving the stability and consistency of the whole process prediction.

[0020] (3) This invention simultaneously utilizes the time series of two types of observations, namely water head (calculated from groundwater level or seepage pressure) and leakage, to construct an objective function, and combines it with sensitivity information to update parameters, thereby enhancing the identifiability of key parameters such as permeability coefficient and water storage parameters, and thus improving the accuracy and reliability of the whole process seepage simulation and leakage prediction.

[0021] (4) This invention establishes a phased configuration and traceable management mechanism driven by engineering events, so that the updates of key boundary conditions and source and sink items such as reservoir water level boundary, cavern / tunnel boundary and drainage operation conditions have clear basis and can be traced back for verification, thereby ensuring the consistency of forward calculation conditions, calibration calculation conditions and monitoring comparison conditions; on this basis, the leakage process line and the proportion of leakage sources and their changes with the stage evolution are further output, enhancing the interpretability of leakage diagnosis, treatment scheme comparison and effect evaluation. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart illustrating a method for simulating and dynamically updating a seepage field throughout the entire process, provided in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the relationship between the entire process being divided into stages according to the project progress and the model configuration being updated as the stages switch, as provided in this embodiment of the invention. Figure 3 A schematic diagram of the three-dimensional unsteady seepage finite element model and its main structure for full-process calculation provided in this embodiment of the invention; Figure 4 This is a schematic diagram of the staged update of the model geometric state and boundary condition set corresponding to different stages provided in the embodiments of the present invention; Figure 5 This diagram illustrates the relationship between the iterative process of dynamic calibration within a stage, parameter updates, and convergence criteria provided in this embodiment of the invention. Figure 6 This is a schematic diagram of the output results of the full-process seepage field response, leakage process line, and leakage source ratio provided in the embodiments of the present invention. Detailed Implementation

[0024] For ease of understanding, the terms "comprising," "including," and "having" used in this specification, claims, and drawings should be understood as open-ended definitions, meaning that the process, method, system, or apparatus may include other steps, units, or elements that may be optionally provided by those skilled in the art to achieve the same technical effect, in addition to the explicitly listed steps, units, or elements.

[0025] The functional block diagram shown in the attached figures is used to illustrate the functional division relationship. A one-to-one correspondence between the functional units and specific physical structures is not required. Related functions can be implemented by software, hardware circuits, programmable logic, or a combination thereof, and can be implemented in a distributed manner on the same processor device or different processor devices. The flowchart shown in the attached figures is for illustrative purposes only, and the order of the steps does not constitute a limitation on the execution order. Without departing from the inventive concept, related steps can be combined, split, replaced, or implemented in an equivalent manner.

[0026] To address the challenges of simultaneously managing geometric / boundary updates, parameter stage inheritance and dynamic calibration, and leakage source identification and quantitative diagnosis under continuous evolution conditions during natural exploration, construction excavation, gradual formation of seepage prevention and drainage systems, water storage, and operation, this invention provides a seepage field simulation and dynamic calibration method and system for the entire process of hydropower engineering construction and operation. This method enables staged model updates and traceable management under continuously evolving geometry and boundaries, and combines monitoring information to perform intra-stage dynamic calibration and inter-stage parameter inheritance. Simultaneously, it outputs diagnostic results such as leakage sources and their proportions to support leakage control and seepage prevention and drainage optimization.

[0027] Specifically, this invention provides a method for simulating seepage fields throughout the entire process of natural exploration, construction, and operation. The method divides the entire time domain into stages based on project progress and events, updating the model geometry, boundary conditions, and source / sink terms within each stage, and conducting unsteady forward modeling calculations. An objective function is constructed based on multi-source monitoring time series such as head / pressure and leakage rate, and sensitivity information is obtained. Adaptive trust region quasi-Newton optimization is used to achieve dynamic calibration within each stage, and the optimal parameters of each stage are inherited as the initial parameters for the next stage. After meeting the convergence conditions, the method outputs the parameter evolution chain, the entire process results of the seepage field and leakage rate, and the proportion of leakage sources. This invention also provides a numerical analysis system for implementing the above method. This system focuses on full-process model management and stage scheduling, realizing versioned management of model configuration, stage-based calculation scheduling, dynamic calibration iteration, and result summary output.

[0028] In one embodiment, the entire process includes at least a natural exploration stage, a construction excavation stage, a seepage prevention and drainage system formation stage, a water storage stage, and an operation stage. During the construction excavation stage, the geometry of the tunnel / cavity is updated based on engineering events such as the advancement of the excavation face and the breakthrough of the tunnel / cavity, and the boundary conditions of the tunnel wall / excavation face are adjusted synchronously. During the water storage stage, time-varying constant head boundaries corresponding to the reservoir water level process line are applied. During the operation stage, unsteady forward modeling calculations can continue to be performed according to the scheduling conditions, thereby forming an unsteady full-process calculation chain covering "exploration-construction-water storage-operation".

[0029] Furthermore, the objective function simultaneously includes time-series error terms for two types of observations: hydraulic head and leakage rate. The hydraulic head is obtained by adjusting the groundwater level or seepage pressure. By iteratively updating the permeability coefficient and other inversion parameters within each stage, the simulation results are made as consistent as possible with the observation information in the time domain. This improves the non-uniqueness that easily occurs in the inversion of a single observation and reflects the dynamic evolution characteristics caused by engineering activities.

[0030] In response to the frequent boundary changes caused by excavation and breakthrough, lining and water-stopping, seepage prevention curtain and drainage facilities formation, and water storage operation, an engineering event table-driven boundary and source-sink item configuration update mechanism is established. The reservoir water level process line, cavern / tunnel boundary, drainage operation conditions and surface recharge are included in the phased configuration and a traceable configuration record is formed to ensure that the forward simulation conditions, calibration conditions and monitoring comparison caliber are consistent.

[0031] Leakage source identification can adopt a combination of "candidate channel identification + numerical decomposition quantification": candidate recharge sources (or recharge channels) are screened based on the hydrochemical characteristics of the leaking water body, and the boundary conditions / source-sink terms corresponding to each candidate source are identified and decomposed in the numerical model, and the source proportion is normalized. The evolution trend of the proportion of each source with the water storage process or operating conditions can be output.

[0032] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined to form new technical solutions. Such combinations are not bound by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0033] This invention discloses a method for simulating and dynamically updating the seepage field throughout the entire process of hydropower project construction and operation. This method takes the evolution of the seepage field throughout the entire process of hydropower project construction and operation as the main thread, and integrates the initial hydrogeological conditions during the natural exploration stage, construction excavation disturbance, the gradual formation of the seepage prevention and drainage system, reservoir impoundment, and operation into a non-steady-state seepage analysis framework that can be updated as the project progresses. Furthermore, it combines monitoring data to dynamically calibrate the model parameters, thereby improving the accuracy and interpretability of the entire process seepage simulation, leakage diagnosis, and treatment scheme comparison.

[0034] For ease of understanding, the following explanation uses a hydropower station's underground powerhouse cavern complex as an example from engineering case studies. This project mainly consists of a roller-compacted concrete gravity dam and an underground powerhouse system: the dam is approximately 123 m high, with a crest elevation of approximately 1995 m and a normal water level of approximately 1987 m; construction began in July 2021, and is scheduled to complete water impoundment and reach the normal water level by March 2029; the underground powerhouse system includes the generating plant, main transformer room, and multiple water diversion tunnels, access / construction tunnels, etc. The cavern area contains multiple strata and structural zones, and faults (such as F...) exist within the cavern area. 152 ) and regional water-rich faults (such as F) 34 It can serve as an important lateral supply channel and needs to be specifically expressed and parameterized in the numerical model so as to quantitatively constrain the "channel-leakage" relationship in subsequent calibration and source identification.

[0035] Monitoring during construction revealed seepage in the bypass tunnel, the plant access tunnel, and the tailrace tunnel after excavation. Seepage was relatively concentrated in the bypass tunnel between K0+193 and K0+495, with some sections exhibiting a slight sulfide odor and higher temperatures (approximately 31.1–32.9℃), suggesting that the seepage may be caused by multiple recharge channels (or multiple sources of recharge). Monitoring data from January 2023 onwards showed that the total seepage from the bypass tunnel stabilized after February 2024, at approximately 47–50 L / s. The distribution of these tunnel sections, temperature characteristics, and seepage stability together constitute important observational constraints for model calibration and provide verifiable comparative evidence for the identification and decomposition calculation of seepage sources.

[0036] See attached document for details. Figure 1The method of this invention adopts the overall approach of "phased advancement—dynamic calibration—source decomposition": First, a full-process seepage numerical calculation model covering all engineering stages is established, and the entire process time domain is divided into several stages according to the project progress; in each stage, unsteady seepage forward modeling is performed and compared with monitoring data, the objective function and its sensitivity to the parameters to be inverted are calculated, and the parameters to be inverted are updated using an adaptive trust region quasi-Newton optimization method; after the convergence condition is met, the optimal parameters of the stage and their evolution results are output, and under the condition that the optimal parameters of the stage are fixed, the contribution ratio of each source to the leakage of the target chamber or tunnel is calculated using the "source-by-source activation / superposition decomposition" method. For ease of description, steps S1-S6 are used for explanation below.

[0037] Step S1: Establish a numerical calculation model for the entire seepage process.

[0038] In step S1, a three-dimensional unsteady seepage finite element numerical calculation model (hereinafter referred to as the "full-process model") is established, covering the entire process of hydropower project, including natural exploration, construction excavation, gradual formation of the seepage prevention and drainage system, and reservoir impoundment and operation. This model is used to simulate the seepage field response as engineering conditions continuously evolve with each stage. The full-process model comprehensively expresses the topography, lithology, geological structure, rock mass permeability zones, major hydraulic structures, and underground cavern / tunnel system of the project area within the same grid and partition framework. It can also enable, replace, or deactivate the model's geometry, boundary conditions, and / or source / sink terms in stages as the project progresses, thus providing a unified computational basis for continuous simulation of the entire process and subsequent dynamic calibration. Step S1 will be further explained below in conjunction with S11 to S14.

[0039] Step S11: Establish a full-process three-dimensional finite element model and configure the natural stage working conditions. Construct a full-process model covering the hub area and the power plant / dam area. The model includes topography, lithology, geological structure, rock mass permeability zones, and major engineering structures within a unified mesh and partitioning framework. The engineering structures include at least: dam and foundation structure, underground powerhouse and water diversion and power generation system, underground cavern / tunnel system, seepage prevention curtain, drainage gallery and drainage canopy, and diversion tunnel, etc.

[0040] Under the natural stage conditions, the engineering structures corresponding to subsequent construction and operation are configured in the model as "not built / not excavated / not in operation," and the application of seepage control boundaries and source / sink terms remains under natural conditions. When entering the corresponding engineering stage, the corresponding geometric and boundary conditions are activated according to the stage switching rules. Combined with borehole groundwater level (or equivalent head) observation data, the initial head field of the natural stage can be checked or inverted to verify the rationality of the values ​​of key parameters such as rock mass permeability coefficient. The model explicitly represents water-conducting structures such as major faults and tectonic fracture zones. Taking this embodiment as an example, as... Figure 3As shown, the model has a length of approximately 4102 m along the river, a length of approximately 3139 m across the river, and a height of approximately 1753 m. It is divided into approximately 6.377 million subdivisions and approximately 1.545 million nodes.

[0041] Step S12: Establish a staged representation and switching interface in the full-process model. In step S12, instead of establishing a new numerical model, within the same full-process model described in step S11, corresponding modeling representation methods and stage switching interfaces are pre-established for typical working condition changes in stages such as construction excavation, gradual formation of the seepage prevention and drainage system, and water storage operation. This allows the model to reflect the evolution of engineering conditions by switching geometric states, boundary conditions, and / or source and sink terms when entering different stages.

[0042] The excavation and connection of underground caverns / tunnels are achieved through birth and death units (unit activation / passivation) or equivalent excavation methods, and are consistent with the activation of boundary conditions of the tunnel walls / excavation faces to ensure that the drainage conditions after excavation are correctly expressed. The seepage prevention curtain and lining water-stopping structure are represented in the model as solid units or equivalent low-permeability zones, and their material parameters or permeability parameters are assigned according to the rules during the formation stage, so that they take effect according to the rules when switching stages. The drainage gallery and drainage hole curtain are set with drainage boundaries and operation parameters in the model. The drainage hole curtain is expressed using substructure modeling technology to achieve accurate application of drainage hole boundary conditions for each hole and avoid excessive densification of the overall mesh due to the hole size being too small. The time-varying head boundary corresponding to the reservoir water level process line is defined in the model as a boundary input that can be updated over time, so that it can be applied according to the process line in the future.

[0043] Step S13 involves calling the model configuration in stages according to the project progress and achieving continuous advancement between stages. In step S13, the entire time domain is divided into several stages based on the project construction and operation progress. When switching stages, the switching interface preset in step S12 is called according to the project events and schedule to update the geometric state, boundary conditions, and / or source and sink terms of the same whole-process model in stages, so that the numerical calculation conditions are consistent with the actual project progress. When the excavation of the underground cavern / tunnel progresses or is completed, the corresponding unit activation / passivation (or equivalent excavation) state switch is triggered, and the cavern wall / excavation face boundary conditions are activated simultaneously. When the drainage hole (or drainage curtain, drainage gallery) changes from not in operation to in operation, the corresponding drainage boundary or equivalent source and sink terms are activated and assigned values ​​according to the operating conditions. The reservoir water level is updated according to the process line changes to update the time-varying head boundary of the reservoir area. In this embodiment, the normal water level is 1987.0 m. During the operation stage, the reservoir water level boundary can continue to be updated according to the scheduling conditions and unsteady calculations can be carried out.

[0044] To ensure the continuous advancement of the entire computational chain, the unsteady seepage calculation in the current stage uses the seepage field state at the end of the previous stage as the initial condition, and the optimal parameters to be inverted obtained from the dynamic calibration of the previous stage are used as the initial parameters for the next stage, thereby achieving seamless transition between stages and continuous evolution of parameters as the project progresses. To improve process traceability, the update content, scope of application, and timestamp corresponding to each stage switch can be recorded to facilitate retrospective verification of stage configurations and calculation results.

[0045] Step S14: Determine the governing equations, initial conditions, and boundary conditions, and configure source and sink terms. In step S14, based on the same full-process model and its staged configuration formed in steps S11 to S13, determine the governing equations for unsteady seepage calculations, and uniformly provide the types of initial and boundary conditions and the configuration methods of source and sink terms, so that the model states at each stage are mathematically closed and solvable. The full-process model adopts a three-dimensional unsteady seepage finite element model, and the governing equations can adopt the basic equations of transient saturated seepage: , In the formula: h is the water head; t is the time; For vector differential operators; K is the permeability coefficient (which can be a tensor or an equivalent value); S s is the water storage rate; W is the source-sink term, used to characterize the effects of replenishment, infiltration, and drainage.

[0046] Furthermore, the model needs to satisfy both initial and boundary conditions. The initial condition can be taken as the initial head field under the natural stage configuration. Where x is a spatial location vector, t0 is the initial time, and h0(x) is the initial head distribution function; the boundary conditions include at least: constant head boundary (used to apply time-varying head boundaries such as reservoir water level process lines or river water levels), constant flux boundary (used to represent surface infiltration / recharge or drainage flux), impermeable boundary, and boundary terms related to the cavern / tunnel (e.g., excavation face, cavern wall seepage surface / potential overflow surface, etc.). The specific numerical processing methods for boundary types such as seepage surface and potential overflow surface can be implemented using methods known in the art, and this invention does not limit their specific implementation forms.

[0047] The calculation scope of the model is consistent with the engineering area determined in step S11, and at least covers the underground powerhouse cavern complex and related tunnels, reservoir area and river boundary, seepage prevention curtain and drainage system, surface recharge boundary, and key fault / valley recharge channels, etc. The fault zone / fracture zone can be represented by partitioned equivalents or strip-shaped units to ensure the identifiability of seepage channels and the adjustability of parameters, and to support subsequent dynamic calibration and seepage source decomposition calculations. It should be noted that the specific configuration of the model scope, grid size, structural representation method, and boundary / source / sink terms can be adjusted according to the engineering area scope, geological complexity, availability of monitoring data, and computing resources; this invention does not impose any limitations on this.

[0048] Step S2: Determine and initialize the seepage parameters to be inverted, select observation points, and construct the objective function.

[0049] In this embodiment, step S2 is used to determine the set of seepage parameters to be inverted, their feasible domain and initial values ​​for subsequent staged unsteady forward modeling and dynamic calibration iterations, and to organize the head and leakage observation sequence under a unified time scale, constructing an objective function that reflects the "simulation-observation" deviation, providing input for error assessment and sensitivity calculation in step S4 and parameter update in step S5.

[0050] Step S21: Determine the seepage parameters to be inverted for the entire process dynamic calibration.

[0051] Based on the stratigraphic zoning of the plant dam area, key water-conducting structures, and existing experimental / statistical results, the permeability parameters of the media zoning and structural zones that significantly contribute to the head response of the underground cavern group and the leakage of the target cavern section, while also exhibiting high uncertainty, are selected as the objects to be inverted. In this embodiment, the selected stratigraphic zoning... to The equivalent permeability coefficient and fault F 152 The equivalent permeability coefficients constitute the seepage parameter vector P to be inverted, that is: , in, This represents the equivalent permeability coefficient of the corresponding stratum. Fault F 152 The equivalent permeability coefficient is given by T, where T is the matrix transpose. Apart from the parameters to be inverted mentioned above, the other parameters required for seepage calculation (such as the permeability coefficients of other zones, water storage rate / specific yield, and other unsteady parameters or surface recharge-related parameters) can be regarded as known parameters and can be determined comprehensively based on geological survey recommendations, pressure test / field test results, and trial calculation verification.

[0052] Step S22: Set the feasible domain of the parameters and give initial values ​​to complete the initialization.

[0053] After determining the parameter vector P to be inverted, to ensure the physical feasibility of the parameters and the stability of the optimization process, a range of values ​​and initial values ​​can be set for each parameter component. Specifically, for each permeability coefficient K... i Set upper and lower limit constraints And give initial values ​​within the feasible region. Initialization complete. The upper and lower limits of the parameter value space can be determined based on geological recommendations, and the initial values ​​can be determined using statistical recommendations or combined experimental values; no specific limitation is made here. In this embodiment, the value ranges of each parameter P to be inverted are as follows: Fault F 152 The value constraint is given within a reasonable range for the project, referring to its recommended penetration value. The initial value can be taken as... m / s. The above initial values ​​can be determined using statistically suggested values ​​or experimental composite values, and used as the starting parameters for step S3.

[0054] Step S23: Select the observation point and obtain the observation time series values.

[0055] The observation points include head measurement points and leakage measurement points. Head measurement points are groundwater level measurement points or seepage pressure measurement points. Groundwater level observations can be directly used as the head sequence, while seepage pressure observations are converted to head to form a head time series. Leakage measurement points are cavern / tunnel leakage monitoring points, which are located at the outlet control surface / catchment section, drainage system, or segmental measurement location of the target cavern or tunnel. In this implementation, the right bank bypass tunnel section K0+193~K0+495 is taken as the target tunnel section, and the leakage of this section is summarized by tunnel section to form a leakage observation sequence. To ensure the comparability of multi-source observations, time-scale standardization, missing measurement and anomaly handling are performed on each observation sequence, and weights or covariance configurations are set based on instrument accuracy and data quality. When necessary, the observation sequence can be divided into a sequence for parameter calibration and a sequence for effect verification to allow for independent verification of calibration results.

[0056] Step S24: Construct an objective function that includes the water head and the total leakage of the tunnel section.

[0057] The objective function includes both a head time series error term and a tunnel section total leakage time series error term. Let the head observation sequence be h. obs The simulated sequence extracted by the model at the water head measuring point is h(P); the total leakage observation sequence of the tunnel section is Q. obs If the model outputs a summary leakage sequence for the corresponding hole segment, denoted as Q(P), then the residual vector is defined as: , The objective function is taken in weighted least squares form, for example: , In the formula: This is the head residual vector; The flow residual vector; Here is the error covariance matrix; It serves as a balancing weight for the two types of observations, used to coordinate the impact of different units and differences in data accuracy on the calibration results.

[0058] Step S3: Update the model in stages and perform unstable forward modeling.

[0059] In step S3, the entire time domain is divided into stages based on key engineering events and their occurrence times. The unsteady seepage forward modeling is organized using a method of "relatively stable operating conditions within a stage, with stage transitions triggered by engineering events" to ensure consistency between the numerical calculation conditions and the actual project progress. Key engineering events include, but are not limited to, tunnel excavation, completion of support or lining, activation or deactivation of drainage facilities, cofferdam water retention, reservoir water level changes, and the formation of the anti-seepage system. The occurrence time of each engineering event can be determined from construction progress data, operation scheduling records, or monitoring data. Combined with... Figure 2 This can be understood as follows: within each stage, an unsteady solution is performed with a fixed geometric state, boundary conditions, and source / sink term configuration, outputting a staged seepage response; when the stage switching conditions are met, an engineering event triggers an update to the model configuration, proceeding to the next stage of calculation. Combined with... Figure 4 This can be understood as follows: Phase switching mainly manifests as the activation / change of geometric states and the addition, replacement or deactivation of boundary condition sets and source / sink item sets, and can form a traceable phased configuration record.

[0060] To facilitate unified scheduling, a stage model configuration C is defined for any stage k. k for: , Among them: G k This represents the geometric and media zoning state of stage k, used to reflect geometric or zoning updates caused by the advancement and breakthrough of cavern / tunnel excavation, the formation of lining / waterstop structures, etc.; B k S represents the set of boundary conditions for stage k, used to apply river level boundaries, surface recharge boundaries, impermeable boundaries, tunnel / excavation face boundaries, drainage boundaries, and reservoir level boundaries, etc. k This represents the set of source and sink terms in the k-th stage, used to characterize the source and sink effects such as rainfall infiltration, gully / fault recharge, and pumping. This is the project event table corresponding to stage k, used to record event type, timestamp, scope, and triggered update items, and used to drive G. k B k or S k Phased updates; These are the initial values ​​of the parameters to be inverted when entering the k-th stage.

[0061] In one embodiment, based on an event-driven phased progression logic, the entire process is divided into Phase A (natural exploration phase), Phase B (construction and excavation phase), Phase C (formation phase of the seepage prevention and drainage system), Phase D (reservoir impoundment phase), and Phase E (operation phase). The configuration sets C for each phase are described below. k The update rules and output organization can be explained according to the above stages A to E. The natural exploration period is mainly used to give the boundary / source-sink configuration of river water level boundary and surface recharge under natural hydrogeological conditions, so as to obtain the initial head field and serve as the starting basis for subsequent stage calculations; the construction excavation period is triggered as the excavation of the cavern / tunnel progresses and is completed. k Activation / passivation of the corresponding units in B (birth / death unit or equivalent excavation treatment), and simultaneous activation / passivation in B. k The boundary condition types matched to the drainage from the tunnel wall / excavation face are used to reflect changes in the drainage channels caused by excavation; the seepage prevention and drainage system formation period is characterized by the improvement of the lining water-stop structure, the formation of the seepage prevention curtain, and the gradual commissioning of drainage facilities, respectively in G k The corresponding low-permeability zoning or material parameter assignment scheme is activated in B. k or S k The system utilizes drainage boundaries / equivalent source / sink terms and their operating parameters to demonstrate the gradual control of seepage field by seepage prevention and drainage; the reservoir's water level process curve is used during the impoundment and operation periods. As the main time-varying boundary input, in B k The system updates the reservoir head boundary according to the process line and can make corresponding adjustments to the drainage boundary and other parameters as the operating conditions change.

[0062] To ensure the continuous advancement of the entire computational chain, a constraint of "parameter inheritance + state inheritance" is adopted between stages: when entering stage k, the initial values ​​of the parameters to be inverted are inherited from the calibration results of the previous stage, and the unsteady forward modeling in this stage uses the seepage field state at the end of the previous stage as the initial condition, that is: , in, The optimal parameters are obtained by the convergence of the dynamic calibration in stage k-1. This represents the values ​​of the set of state variables of the seepage field (e.g., the head field) at the end of stage k. Stage k is given by C. k With time step Under unsteady conditions, the solution is performed, and the output results include the head field, water table location, flux distribution, and leakage process curve of the cavern / tunnel. When the engineering event table meets the stage switching conditions, it triggers the process of G. k B k and / or S kThe system updates and proceeds to the next stage of calculation. To facilitate full-process backtracking and verification, stage inputs (stage configuration and initial parameter values), stage outputs (seepage response and key indicators), and calibration process records (iteration and convergence information) can be archived and managed.

[0063] Step S4: Calculate the objective function value and the Jacobian matrix.

[0064] In step S4, based on the stage seepage field simulation results obtained in step S3, a residual vector is constructed by combining the head and leakage monitoring information, and the objective function value is calculated accordingly; at the same time, the sensitivity information of the objective function with respect to the parameters to be inverted is calculated to form the Jacobian matrix, which is used for parameter update in step S5.

[0065] In one embodiment, let the observed data vector be y. obs The numerical model outputs the model state under given parameters P as follows: And through observation operators Mapping to the observation space yields analog quantities The residual vector is then defined as: , in, It should include at least the components corresponding to each monitoring point, such as the head of the observation well (or the head converted from seepage pressure) and the leakage of the cavern / tunnel. The objective function can be expressed in the form of weighted least squares: , In the formula: W is the weight matrix or covariance weighted matrix, used to comprehensively reflect the differences in dimensions, sampling frequency, and instrument accuracy of different observations; when the covariance matrix V is used to express the statistical characteristics of error, it can be taken as... ;

[0066] To maintain consistency with the construction of the objective function in step S2, when the observations consist of the "head time series" and the "summary leakage time series of the tunnel section", we can set: , And write W in block form: , This ensures consistency with the objective function in step S2.

[0067] Jacobian matrix Defined as the partial derivative of the residual vector with respect to the inversion parameters: .

[0068] In one implementation, the Jacobian matrix can be approximated by the finite difference method with parameter perturbation: for the parameter vector Apply a small perturbation to the i-th component Repeat step S3 and reconstruct the residual vector. Then we have: , In the formula: This represents the i-th column of the Jacobian matrix. The perturbation quantity. It can be set according to relative disturbance, for example In the formula: ε is the perturbation proportionality coefficient, which is used to control the perturbation amplitude of the parameters in order to balance numerical stability and sensitivity resolution.

[0069] Furthermore, to improve computational efficiency, the forward modeling tasks corresponding to each parameter perturbation can be executed in parallel, thereby generating the columns of the Jacobian matrix in parallel. Based on J and r, the gradient information of the parameters can be further constructed. For step S5: .

[0070] In addition, see Figure 5 Step S5, together with steps S3 and S4, constitutes an intra-stage closed loop: when the parameter update is accepted but the convergence condition has not yet been met, return to step S3 to re-perform the unstable forward modeling calculation to obtain a new model response and update the error and sensitivity information; until the convergence criterion is met, proceed to step S6 to output the results.

[0071] Step S5: Update the parameters to be inverted using the adaptive trust region quasi-Newton method and determine convergence.

[0072] In step S5, based on the objective function F(P), residual vector r(P), weight matrix W, and Jacobian matrix J(P) obtained in step S4, the parameter vector P to be inverted is iteratively updated to gradually reduce F(P) and make it consistent with the observation information. For the k-th stage, let the parameter in the j-th iteration within the stage be... The corresponding residuals and Jacobians are denoted as follows: and And at this iteration point, construct a trust region with a radius of... A local quadratic approximation model is used to solve for the parameter increments.

[0073] Furthermore, this step determines whether to accept the update through a process of "candidate update - forward modeling verification - return rate determination," and adaptively adjusts the trust region radius accordingly. If the update is accepted, the unstable forward modeling calculation in step S3 is re-executed with the updated parameters, and the process returns to step S4 for updating. and The process continues until the convergence condition is met, at which point the process proceeds to step S6 to output the result.

[0074] Step S51: Construct the gradient and quasi-Newton approximation matrix.

[0075] Using the objective function F(P) described in step S24, the gradient vector of the j-th iteration It can be represented as: , In the case of a weighted least squares objective function, second-order approximation information can be constructed from the Jacobian matrix, and the quasi-Newton approximation matrix can be used to construct the second-order approximation information. Take as Alternatively, it can be obtained by recursively applying a quasi-Newton update rule based on this approximation to improve iterative stability.

[0076] Step S52: Solve for the parameter update amount under the trust region constraint and obtain the candidate parameters.

[0077] In the trust region radius Solving the following trust region subproblem under constraints yields the parameter update. : , Among them, the local quadratic model (trust region sub-model) for Based on this, candidate parameters are constructed. If the parameters have physical constraints (e.g., upper and lower limits for permeability coefficient, range corresponding to logarithmic parameterization), then for Constraints are applied to ensure parameter feasibility.

[0078] Step S53: Calculate the yield and decide whether to accept or reject the update, while adaptively adjusting the trust region radius.

[0079] Candidate parameters Substituting the values ​​into steps S3-S4, an unstable forward model is performed to obtain the new objective function value. And calculate the rate of return ρ k,j : , In the formula: the numerator is the actual decrease in the objective function, and the denominator is the decrease predicted by the quadratic model. This represents the value of the quadratic model at the zero increment. When The time interval indicates that the update produces a valid decrease in the true objective function, meaning that the step size is effective within the current trust region. Therefore, the update is accepted and set... At the same time, according to ρ k,j Size pair Maintain or increase; when If the update does not decrease or increase the true objective function, it indicates that the quadratic model is mismatched near that step size. Therefore, the update should be rejected, the trust region reduced, and the process returned to step S52 to solve for the new objective function. Until the acceptance criteria are met.

[0080] Step S54, convergence determination and iterative loop.

[0081] The iteration within stage k is terminated and the optimal parameters for that stage are output when any of the following convergence conditions are met: , Convergence conditions may include, but are not limited to: 1. ≤ ε F (The objective function changes relatively little); 2. ≤ ε r (The residual is small enough); 3. ≤ ε p (The parameter increment is small enough); 4. ≤ ε g (The gradient is small enough); 5. The number of iterations j reaches the upper limit. max .

[0082] If the convergence condition is not met, then, provided the update is accepted, return to step S3 to perform a new unsteady seepage forward modeling calculation to update r. k,j+1 With J k,j+1 Then proceed to the next round of steps S4 to S5.

[0083] Step S6: Output the optimal parameters, the overall process results, and the percentage of leakage sources, and archive them.

[0084] Step S6: After step S5 meets the convergence condition, output the optimal parameters to be inverted obtained from the dynamic calibration of each stage, the seepage response results of the whole process and the leakage process line, and give the proportion of leakage sources based on the numerical decomposition idea of ​​"enabling / overlaying decomposition by source"; at the same time, archive the key data related to stage configuration, engineering events and monitoring comparison to support subsequent review and traceability.

[0085] Step S61: Output of the optimal parameters and their evolution chain for each stage. When there are k stages in the entire process, the output of the stage parameters is organized in the form of "stage index - end time of stage - optimal parameters of stage". It includes at least the initial parameters of each stage and the optimal parameters of the stage after convergence, and explicitly gives the stage inheritance relationship: the next stage enters the calibration of the current stage with the optimal parameters of the previous stage as the initial value, thus forming a parameter chain that evolves continuously as the project progresses.

[0086] To enhance interpretability, "relative change characteristics" and "sensitivity ranking" can be output for key zoning parameters (such as permeability coefficients and water storage parameters for each permeable zone, fault zone / fracture zone, etc.). It is recommended to use a logarithmic scale for permeability coefficient changes to facilitate the identification of media zones or structural channels that undergo significant adjustments at a certain stage, and to establish a correspondence between these adjustments and engineering event tables (such as excavation completion, commissioning of drainage facilities, rapid rise in reservoir water level, etc.).

[0087] Step S62: Output the entire process seepage field and leakage rate process line. The entire process seepage response results are output in stages and can be spliced ​​into a continuous process line, including at least the evolution of the head field / ground level over time, and the response comparison of key parts (powerhouse area, cavern junction area, structural zone crossing area, near-dam / near-curtain area, etc.).

[0088] The leakage rate process line is used to statistically analyze the leakage rate of the target chamber / tunnel or drainage system according to the control surface (or outlet boundary). This can be obtained by integrating the normal outflow flux: , In the formula: The water outlet boundary or catchment section of the chamber / tunnel; q n This represents the normal outflow (outflow is considered positive). The output organization method should be consistent with the engineering statistics, summarizing at the "section-system-total" level (for example, summarizing the right bank bypass dam tunnel by section and providing the total seepage process line of the tunnel), and simultaneously providing statistics such as peak value, stable value, and cumulative outflow to support scheme comparison and operation evaluation.

[0089] Simultaneously, key evaluation indicators can be output for risk identification and calibration rationality verification, such as: seepage pressure / pore pressure (calculated from head and elevation), hydraulic gradient (calculated from the spatial gradient of head), and drainage drawdown effect (obtained by comparing head / water level changes at key locations before and after drainage activation), etc. The extreme value locations and evolution trends near key sections and key structural zones are also provided as a basis for subsequent control recommendations. The results of the entire process can be combined with… Figure 6 (a) to (b) are presented in the form of “comparison of measured value and fitted value + stage division mark” to reflect the fitting effect and stage response characteristics after dynamic calibration.

[0090] Step S63: Leakage Source Identification and Proportion Quantification Output. To output the "leakage source proportion," a combination of "candidate source identification + numerical decomposition quantification" can be used. When water chemical indicators are available, the characteristics of the leaked water sample can be used for source screening or as prior constraints for proportion estimation. For example, given M types of candidate sources, the mixed weight (proportion coefficient) of the m-th candidate source on the observed leaked water sample characteristics is... This is used to characterize the relative contribution of this source to the chemical composition of the leachate, satisfying... and The typical hydrochemical eigenvector of source m is C. m The observed leakage feature vector is C. obs Then a hybrid model can be established: , Therefore, we obtain The initial range or priority sorting is used to guide the setting of source sets and weights for subsequent numerical decomposition calculations. This step is not limited to a specific solution method and can employ least squares or robust fitting, etc.

[0091] optimal parameters at each stage Below, distinguishable boundaries or source-sink identifiers are set for each source (e.g., different recharge boundaries, different fault zone recharge terms, etc.), and the contribution of each source to the target leakage is calculated by decomposition. One approach is "enable / decompose on a per-source basis": Parameters during the retention phase Under the condition that remains unchanged, only the boundary / source-sink term of the m-th source is enabled, and other sources are set to zero or disabled to obtain the corresponding target outflow: , The total amount is obtained by summing the contributions from each source: , Calculate the contribution curves of each source to the target leakage (either instantaneous or phased average): , or , In the formula: Let k be the time interval for stage k. The above decomposition calculation is performed under the assumptions of fixed optimal parameters for each stage and equivalent decomposability of the engineering process, and is used to obtain the relative contribution ratio of each supply source to the target leakage and the stage evolution law.

[0092] Furthermore, the percentage results can be output using a three-dimensional index of "stage - reservoir water level - operating conditions" to form an evolution sequence of the source percentage as the project progresses: , This is used to track key conclusions such as the dynamic changes in leakage sources caused by human engineering activities such as construction disturbances. In this example, for instance... Figure 6 (c) in the figure presents a comparison of the proportion of leakage sources at different stages or under different working conditions in the form of a pie chart (or other forms).

[0093] Step S64, Control Recommendations and Archived Output. Based on the output parameter evolution chain, seepage response, and source proportion, suggestive conclusions related to engineering control can be given and archived. The suggestive information can establish a logical chain around "changes in source proportion—abnormal indicators in key components—engineering events": for example, when the source proportion of a certain structural zone continues to increase and a high-value area of ​​hydraulic gradient in the corresponding region appears stably, it suggests focusing on this channel and considering reinforcement grouting, curtain continuity improvement, or water-stop lining strengthening in conjunction with engineering conditions; when the water level drop effect is insufficient after drainage measures are implemented and the seepage pressure exceeds the threshold, it suggests reviewing the drainage conditions or optimizing the drainage layout; when certain parameters are sensitive to the fitting target but monitoring coverage is insufficient, it suggests adding observation wells or segmented metering points to improve identifiability and calibration stability.

[0094] Archived content may include: records of geometric states and boundary / source / sink configurations at each stage, records of engineering events triggered by stage switching, initial values ​​and optimal parameters for each stage, monitoring comparison curves and statistical errors, summary tables of leakage by hole segment / system, source proportion results and their index relationships, thereby ensuring that the results of the entire process are traceable and verifiable.

[0095] Based on the same inventive concept as the aforementioned method, this invention also provides a numerical analysis system for simulating and dynamically updating the seepage field throughout the entire process of natural exploration, construction, and operation. The system includes:

[0096] The full-process model management unit is configured to establish and maintain a three-dimensional unsteady seepage finite element numerical calculation model, complete mesh generation and medium partitioning, uniformly configure reservoir water level boundaries, river boundaries, surface recharge boundaries, impermeable boundaries, drainage boundaries, and potential overflow boundaries, and pre-set switching interfaces and calling rules for geometric states and boundary conditions for engineering events such as tunnel / cavity excavation and connection, drainage facility commissioning, reservoir water level rise to normal storage level and its process line changes during the project's progress.

[0097] The stage scheduling and model update unit is configured to divide the entire time domain into several stages based on engineering events, and to trigger the activation, replacement or deactivation of geometric states, boundary conditions and / or source and sink terms within the same numerical model when switching stages, so that the model operating conditions are consistent with the actual engineering situation.

[0098] The parameter configuration and target construction unit is configured to acquire and manage engineering event tables, reservoir water level process lines, geological survey test data, and groundwater level / seepage pressure monitoring data and leakage monitoring data. It performs time-scale unification, missing and anomaly handling, and weight configuration on the monitoring time series; determines the seepage parameter vector to be inverted, sets the initial values ​​and value constraints of the parameters, selects the head measurement points and leakage measurement points, and constructs an objective function to characterize the deviation between the simulated response and the observed information.

[0099] The unsteady seepage calculation unit is configured to perform unsteady seepage forward modeling under given stage configuration and parameter values, and output response results such as head field, flux field and seepage process line of target cavern / tunnel.

[0100] The error assessment and sensitivity analysis unit is configured to calculate the objective function value based on simulation results and observation information, and obtain the sensitivity information of the objective function with respect to the parameters to be inverted, forming the Jacobian matrix or equivalent gradient information; at the same time, it can schedule multiple forward modeling tasks corresponding to parameter perturbations in parallel to shorten the inversion time.

[0101] The parameter adaptive update unit is configured to update the parameters to be inverted using an adaptive trust region quasi-Newton optimization strategy, and to adaptively adjust the update step acceptance / rejection and trust region radius according to the return criterion until the convergence condition is met.

[0102] The leakage source analysis unit is configured to identify and decompose the boundary conditions or source-sink terms corresponding to different preset supply sources, and obtain the contribution ratio of each source to the leakage of the target cavern or tunnel and its evolution law with stage by using the source-by-source activation / overlay decomposition method.

[0103] The results summary and evolution output unit is configured to output the optimal estimate of the parameters to be inverted and their stage evolution chain, the calculation results of the seepage field throughout the process and the leakage process line when the convergence condition is met, and output the proportion of leakage sources, evaluation indicators of key parts and archived information; when the convergence condition is not met, it drives the unsteady seepage calculation unit, error assessment and sensitivity analysis unit and parameter adaptive update unit to continue iterative updates until convergence.

[0104] Based on the same inventive concept as the aforementioned method, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program. When the processor executes the computer program, it implements the steps of the method for simulating and dynamically updating a full-process seepage field.

[0105] Specifically, the computer device can be a personal computer, server, workstation, or other electronic device with data computing and storage capabilities. The processor can be a central processing unit (CPU), graphics processing unit (GPU), or application-specific integrated circuit (ASIC), etc.

[0106] Based on the same inventive concept as the aforementioned method, the present invention also provides a non-transitory computer-readable storage medium storing computer instructions thereon. When executed by a processor, the computer instructions cause the computer to perform the following steps:

[0107] A numerical calculation model for seepage throughout the entire process, covering natural exploration, construction excavation, construction of the seepage prevention and drainage system, water storage, and operation, is established. The seepage parameters to be inverted are determined and initialized. Head measurement points and leakage measurement points are selected, and an objective function is constructed. Based on an engineering event table, the entire time domain is divided into several stages. The model's geometric state, boundary conditions, and / or source-sink terms are updated according to each stage, and unsteady seepage forward modeling is performed. The current stage uses the seepage field state at the end of the previous stage as the initial condition, and the initial values ​​of the parameters to be inverted are inherited from the optimal parameters calibrated in the previous stage. A residual vector is constructed based on simulation results and observation information, and the objective function value is calculated. The sensitivity of the residual vector to the parameters to be inverted is calculated to obtain the Jacobian matrix. The parameters to be inverted are updated using an adaptive trust region quasi-Newton optimization method, and convergence is determined. If the convergence condition is not met, the model update, forward modeling, and parameter update steps are repeated. When the convergence condition is met, the optimal estimate of the parameters to be inverted, the calculation results of the entire seepage field, and the contribution ratio of each preset supply source to the leakage are output.

[0108] The storage medium can be a read-only memory (ROM), random access memory (RAM), USB flash drive, portable hard drive, magnetic disk, optical disk, or solid-state drive (SSD) or any other medium capable of storing program code.

[0109] In summary, this invention addresses the continuous evolution of seepage fields throughout the entire construction and operation process of hydropower projects. It unifies event-driven switching of geometric and boundary conditions, data-driven dynamic calibration within stages, inter-stage state and parameter inheritance, and quantitative output of leakage source proportions within a single computational framework. This improves the accuracy and interpretability of seepage simulation and leakage diagnosis under continuously changing geometric and boundary conditions, solving problems in existing technologies such as the difficulty in retrospectively updating model configurations as the project progresses, difficulties in connecting parameter calibration with stage evolution, and the difficulty in quantitatively decomposing leakage sources. Compared to existing technologies, this invention enables dynamic model updates and parameter calibration under continuously evolving site geometry and boundary conditions, improving the accuracy and interpretability of seepage simulation, leakage diagnosis, and treatment scheme comparison throughout the entire process.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for simulating and dynamically updating the seepage field throughout the entire process, characterized in that, include: Step S1: Establish a full-process seepage numerical calculation model covering the natural exploration stage, construction and excavation stage, seepage prevention and drainage system construction stage, reservoir impoundment stage and operation stage of hydropower project, and set the geometric state, boundary conditions and source and sink terms corresponding to each stage. Step S2: Determine and initialize the seepage parameters to be inverted, select observation points and construct the objective function. The observation points include head measurement points and leakage measurement points. Step S3: Based on the engineering event table, the entire process time domain is divided into several stages. The geometric state, boundary conditions, and source and sink terms of the whole process seepage numerical calculation model are updated according to the stages, and unsteady seepage forward modeling is performed to obtain the seepage field simulation results of each stage. Among them, the unsteady seepage forward modeling of the current stage uses the seepage field state at the end of the previous stage as the initial condition, and the initial values ​​of the seepage parameters to be inverted in the current stage are inherited from the seepage parameters to be inverted after the update of the previous stage. Step S4: Construct a residual vector and calculate the objective function value based on the simulation results of the seepage field and the actual seepage field observation information, and calculate the sensitivity of the residual vector with respect to the seepage parameters to be inverted to obtain the Jacobian matrix; Step S5: Update the seepage parameters to be inverted using the adaptive trust region quasi-Newton optimization method and determine convergence; if the convergence condition is not met, repeat steps S3 to S5. Step S6: When the convergence condition is met, output the optimal estimate of the seepage parameters to be inverted and the calculation results of the seepage field throughout the process.

2. The method for simulating and dynamically updating the seepage field throughout the entire process according to claim 1, characterized in that, Step S6 further includes: using the source-by-source activation and superposition decomposition method to calculate the contribution ratio of the boundary conditions or source-sink terms corresponding to each preset supply source to the leakage of the target cavern or tunnel.

3. The method for simulating and dynamically updating the seepage field throughout the entire process according to claim 1, characterized in that, The phased update in step S3 is driven by the project event table, which records at least the event type, scope, and timestamp.

4. The method for simulating and dynamically updating the seepage field throughout the entire process according to claim 1, characterized in that, Step S5 includes: in the first j In the next iteration, at the trust region radius Solving parameter increments under constraints To satisfy the optimization of the local quadratic model, the return rate is calculated after obtaining candidate parameters. When the return rate is greater than a preset threshold, the candidate parameter is accepted and the trust region radius is increased or maintained. When the return rate is less than or equal to the preset threshold, the candidate parameter is rejected and the trust region radius is reduced. The parameter increment is resolved under the updated trust region constraint until the acceptance condition is met or the preset iteration limit is reached.

5. The method for simulating and dynamically updating the seepage field throughout the entire process according to claim 2, characterized in that, The source-by-source activation and superposition decomposition includes: when only a single source is activated and all candidate sources other than the current single source are deactivated, unsteady flow forward modeling is performed to obtain the leakage amount corresponding to the source, and the leakage amount of the single source is normalized relative to the total leakage amount obtained when all sources are activated at the same time, so as to obtain the contribution ratio of each source to the leakage amount of the target chamber or tunnel.

6. The method for simulating and dynamically updating the seepage field throughout the entire process according to claim 1, characterized in that, In step S3, the changes in geometric state, boundary conditions and / or source and sink terms corresponding to each staged update are stored in association with the corresponding engineering events, forming a versioned record of the model configuration.

7. The method for simulating and dynamically updating the seepage field throughout the entire process according to claim 1, characterized in that, Step S5 includes: selecting samples in the parameter space using orthogonal design and performing forward modeling; extracting the head and leakage response features at the observation points to form training data; training a neural network to establish the mapping relationship between parameters and the observed response to form a surrogate model; and using a genetic algorithm to pre-optimize or accelerate the optimization of the objective function on the surrogate model to obtain the initial parameters or candidate parameter combinations for optimization iteration in step S5.

8. A system for simulating and dynamically updating the seepage field throughout the entire process, characterized in that, include: The full-process model management unit is used to build and maintain a seepage numerical calculation model that can be updated as the engineering stage changes. The model can express the geometric state, material partitioning, boundary conditions and source-sink configuration corresponding to different stages. The phase scheduling and model update unit is used to trigger phased updates of the model's geometric state, boundary conditions, and source and sink terms based on the project progress and project events, so as to form a phased model configuration consistent with the actual project progress. The parameter configuration and target construction unit is used to determine the initial values ​​of the seepage parameters to be inverted and establish value constraints, and to construct the objective function based on the head measurement point and the leakage measurement point; The unsteady seepage calculation unit is used to perform unsteady seepage forward modeling under given stage configuration and parameter values, and outputs seepage field response results and leakage process lines of the target cavern or tunnel. The error assessment and sensitivity analysis unit is used to construct the residual vector and calculate the objective function value, and obtain sensitivity information through parameter perturbation to form the Jacobian matrix; The parameter adaptive update unit is used to update the seepage parameters to be inverted using an adaptive trust region quasi-Newton optimization strategy, and to perform acceptance or rejection decision and adaptive adjustment of the trust region radius until the convergence condition is met. The leakage source analysis unit is used to identify and decompose the boundary conditions or source-sink terms corresponding to different candidate supply sources, and output the proportion of leakage sources and their evolution with the stage. The results summary and evolution output unit is used to output the optimal estimate of the seepage parameters to be inverted and their stage evolution results, as well as the seepage field response results and the proportion of leakage sources throughout the entire process.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method for simulating and dynamically updating the seepage field throughout the entire process as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions thereon, characterized in that, When the computer instructions are executed by the processor, the computer performs a method for simulating and dynamically updating a seepage field throughout the entire process, as described in any one of claims 1 to 7.

Images