Aquifer parameter optimization method and system based on data inversion

Abstract

The invention discloses an aquifer parameter optimization method and system based on data inversion, and the method comprises the steps: configuring an observation well in a research region, and carrying out the water pumping test through the observation well; according to the test data of the water pumping test, performing inversion analysis on the original parameters of the aquifer at the position of the observation well to obtain optimized aquifer parameters; and according to the optimized aquifer parameters at the observation well position, performing synchronous optimization of the aquifer parameters in the research area. The method can automatically converge to a high-precision parameter solution in multi-round disturbance optimization, and effectively overcomes the problem that a conventional analysis method is difficult to solve under nonlinear and unsteady working conditions. The whole scheme can be embedded into an existing digital twinborn platform, has good engineering landing performance, can remarkably improve the aquifer modeling efficiency and the regional hydrological simulation reliability, and has wide application prospects and popularization values.

Description

Technical Field

[0001] The present invention relates to the technical field of data inversion, and in particular to a method and system for optimizing aquifer parameters based on data inversion. Background Art

[0002] Aquifer parameters are key inputs for groundwater flow simulation, resource management, and hydrogeological assessment. Accurately understanding these parameters is crucial for groundwater extraction planning, pollution migration prediction, geological disaster prevention and control, and regional water resource regulation. However, obtaining these parameters in practical projects faces significant challenges, including severe spatial heterogeneity, limited field testing, and large data perturbation errors. This makes it difficult to directly obtain a complete and accurate parameter distribution through a single method or quantitative experiment.

[0003] Currently, commonly used methods for obtaining parameters include pumping tests, well tests, geophysical logging, and laboratory permeability tests. Pumping tests are the most representative field testing method because they provide the most direct response to groundwater systems. However, relying solely on pumping data, inferring the spatial distribution of multiple parameters from water level changes at a limited number of observation points remains a challenging technical challenge. This is especially true when the aquifer is significantly heterogeneous, the structure is complex, or the number of observation wells is limited. Traditional analytical methods struggle to meet the high-resolution and high-precision inversion requirements.

[0004] In addition, with the development of digital twins, data-driven modeling and artificial intelligence algorithms, how to integrate these advanced technologies into the aquifer parameter inversion process to further improve the stability, adaptability and regional generalization ability of the inversion results has also become a research hotspot in the current groundwater modeling field. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] To solve the above technical problems, the present invention provides the following technical solution: a method for optimizing aquifer parameters based on data inversion, comprising:

[0007] In the study area, an observation well is configured and a pumping test is performed using the observation well;

[0008] Based on the test data of the pumping test, the original parameters of the aquifer at the observation well are inverted and analyzed to obtain the optimized aquifer parameters;

[0009] Based on the optimized aquifer parameters at the observation well location, synchronously optimizing the aquifer parameters in the study area;

[0010] In the synchronous optimization process, if only one observation well is configured, the aquifer parameters of the entire study area are optimized in a geometric manner; if multiple observation wells are configured, the aquifer parameters of the entire study area are optimized in a gradient manner.

[0011] As a preferred solution of the aquifer parameter optimization method based on data inversion described in the present invention, wherein: the configuration strategy of the observation wells includes individual configuration and joint configuration;

[0012] The separate configuration includes configuring only one observation well in the study area; the joint configuration includes configuring m×n observation wells in the study area according to a matrix position relationship;

[0013] Wherein, n and m are both adjustable positive integer values.

[0014] As a preferred embodiment of the aquifer parameter optimization method based on data inversion described in the present invention, the pumping test includes obtaining initial static water level data of the observation well after standing still, which serves as a baseline; pumping water at a preset pumping rate; after completing one round of pumping test, standing still until the water level in the observation well reaches the baseline; after the water level in the observation well reaches the baseline, performing the next round of pumping test, and repeating the process C times;

[0015] Where C represents the maximum number of tests; during the pumping and standing process, a sampling frequency of 1 Hz is used to record the water level data in the observation well;

[0016] During each round of pumping test, the termination conditions include: condition 1: reaching the maximum pumping time; condition 2: the water level in the observation well is pumped to 0; when any termination condition is met, the current round of pumping test ends.

[0017] As a preferred embodiment of the aquifer parameter optimization method based on data inversion of the present invention, the original parameters of the aquifer include but are not limited to permeability coefficient, specific storage coefficient, and delayed water production coefficient;

[0018] If there are multiple observation wells, the original parameters of the aquifer also include anisotropy ratio.

[0019] As a preferred embodiment of the aquifer parameter optimization method based on data inversion described in the present invention, the inversion analysis includes simulating the water flow process of the aquifer in accordance with the geological conditions of the study area and the original parameters of the aquifer in the digital twin platform;

[0020] For each round of pumping test and the rest period after pumping, the trained temporal convolutional network is used to convert the data recorded in the pumping test into a low-dimensional intermediate vector. The input is the pumping rate and the water level data sequence recorded during the pumping and rest period. The output is the intermediate vector.

[0021] The original parameters of the aquifer are input into a mapping model, which outputs a vectorized representation of the simulated field as a comparison vector. The mapping model is a combination of a feedforward neural network and a self-attention encoding module. It uses an MLP to encode low-dimensional structural features to obtain a preliminary structural feature vector. An attention mechanism is then added to weight the influence of different parameters, thereby achieving vector conversion.

[0022] Calculating the distance L between the intermediate vector and the comparison vector as fitness;

[0023] In the parameter space of the aquifer, the original parameters are used as initial values, and each parameter is randomly perturbed. The fitness after the perturbation is evaluated. If the fitness decreases, the aquifer parameters are updated to the values after the perturbation; if the fitness increases, the aquifer parameters are not updated; if the fitness remains unchanged, the aquifer parameters before and after the perturbation are retained, and the next round of perturbations is performed respectively. In the next round of perturbations, the aquifer parameters with the minimum fitness are updated.

[0024] The perturbation and update process is repeated until the minimum fitness or convergence condition is met, and the iterative process ends. The parameters of the C group of aquifers are obtained by iterative optimization of C pumping tests. For each aquifer parameter, the maximum and minimum values of the parameters in the C group are taken as the interval after each parameter is optimized [s min , s max ];

[0025] The random perturbation includes, in {+Δ s , -Δ s ,0}, randomly select an adjustment direction and perturb the parameters; where, +Δ s Indicates adding an adjustment step to the sth parameter; -Δ s Indicates that the sth parameter is reduced by one adjustment step; 0 means no adjustment; s min Indicates the minimum value of the sth parameter in group C; s max Indicates the maximum value of the sth parameter in group C.

[0026] As a preferred embodiment of the aquifer parameter optimization method based on data inversion of the present invention, the geometric optimization includes optimizing the parameters at the same ratio at all locations in the region according to the ratio of each parameter before and after optimization in the observation well;

[0027] At the observation well location, the ratio of the sth parameter before and after optimization is: and

[0028] In the region, the result after optimizing the sth parameter is: at position (x, y), the minimum value s after optimizing the sth parameter is min,(x,y) ; At position (x,y), the maximum value s after the sth parameter is optimized max,(x,y) ;

[0029] in, s0 represents the value of the sth parameter before optimization at the observation well; s 0,(x,y) Represents the value of the sth parameter at position (x, y) before optimization.

[0030] As a preferred embodiment of the aquifer parameter optimization method based on data inversion of the present invention, the gradient optimization includes obtaining the spatial distribution of each optimized parameter in m×n space, and fitting a variation curve between the position and each optimized parameter according to the distribution;

[0031] In the study area, the location coordinates are input, and the optimization results of each parameter of the aquifer at the location are output using the change curve.

[0032] A system for optimizing aquifer parameters based on data inversion using any of the methods described in the present invention, wherein:

[0033] The acquisition unit configures an observation well in the study area and uses the observation well to conduct a water pumping test;

[0034] The analysis unit performs an inversion analysis on the original parameters of the aquifer at the observation well location based on the test data of the pumping test to obtain the optimized aquifer parameters;

[0035] The prediction unit performs synchronous optimization of the aquifer parameters in the study area according to the optimized aquifer parameters at the observation well location.

[0036] A computer device comprises: a memory and a processor; the memory stores a computer program, wherein: when the processor executes the computer program, the steps of any one of the methods of the present invention are implemented.

[0037] A computer-readable storage medium stores a computer program, wherein: when the computer program is executed by a processor, the steps of any one of the methods of the present invention are implemented.

[0038] Beneficial effects of the present invention: The aquifer parameter optimization method based on data inversion provided by the present invention realizes the intelligent inversion and optimization of aquifer parameters by combining pumping test data, digital twin platform simulation and neural network vector mapping mechanism. During the inversion process, the similarity measurement between the intermediate vector and the comparison vector is used to establish a fitness evaluation system based on the collaboration of data-driven and physical modeling. It can automatically converge to a high-precision parameter solution in multiple rounds of perturbation optimization, effectively overcoming the problem that traditional analytical methods are difficult to solve under nonlinear and non-steady-state conditions. At the same time, the present invention flexibly selects regional optimization methods according to the layout strategy of observation wells. When the number of observation wells is insufficient, the global synchronous adjustment of parameters is achieved through geometric optimization; when the number of observation wells is large, the spatial distribution reconstruction is achieved through gradient fitting, which has both universality and accuracy. The overall solution can be embedded in the existing digital twin platform, has good engineering implementation, can significantly improve the efficiency of aquifer modeling and the reliability of regional hydrological simulation, and has broad application prospects and promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 An overall flow chart of an aquifer parameter optimization method based on data inversion provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0041] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0042] Example 1, with reference to Figure 1 , as an embodiment of the present invention, provides a method for optimizing aquifer parameters based on data inversion, comprising:

[0043] S1: In the study area, an observation well is configured and a water pumping test is performed using the observation well.

[0044] Observation well configuration strategies include individual and combined configuration. Individual configuration involves configuring only one observation well within the study area. Combined configuration involves configuring m × n observation wells within the study area according to a matrix-like positional relationship (distributing observation wells at regular grid nodes in rows and columns, with equal or variable spacing, with each observation well corresponding to a spatial sampling point). Both n and m are adjustable positive integers.

[0045] Furthermore, the pumping test includes obtaining the initial static water level data of the observation well after standing still, which serves as the baseline; pumping water at a preset pumping rate; after completing one round of pumping test, standing still until the water level in the observation well reaches the baseline; after the water level in the observation well reaches the baseline, conducting the next round of pumping test, and repeating it C times. Where C represents the maximum number of tests; during the pumping and standing process, the water level data in the observation well is recorded at a sampling frequency of 1 Hz.

[0046] During each round of pumping test, the termination conditions include: condition 1: reaching the maximum pumping time; condition 2: the water level in the observation well is pumped to 0; when any termination condition is met, the current round of pumping test ends.

[0047] It is important to note that by rationally configuring observation wells and combining them with high-frequency, multi-round pumping tests, water level-time data that fully reflects the hydraulic response characteristics of the aquifer can be obtained, providing a reliable input basis for subsequent aquifer parameter inversion analysis. Individually configuring observation wells is suitable for resource-constrained or structurally homogeneous areas. Combined configuration can establish a spatial sampling grid within complex regions, improving the regional representativeness of the data and the accuracy of the inversion. Using a multi-round pumping-and-resting alternating method, the dynamic response characteristics of the aquifer under different disturbance conditions, including drawdown rate, recovery speed, and hysteresis behavior, can be captured, enhancing the stability and identifiability of the parameter inversion. Setting pumping termination conditions prevents well drying or ineffective pumping, improving test safety and efficiency. Recording water level changes at a frequency of 1 Hz ensures sufficient temporal resolution in the acquired data, enabling a fine-grained characterization of the hydrological response process and laying the foundation for the subsequent construction of a high-precision, hydrological-data coupled inversion model.

[0048] S2: Based on the test data of the pumping test, the original parameters of the aquifer at the observation well location are inverted and analyzed to obtain the optimized aquifer parameters.

[0049] The aquifer's original parameters include, but are not limited to, permeability coefficient, specific storage coefficient, and delayed water production coefficient. Specifically, the permeability coefficient is derived from fitting the pumping rate and water level drawdown; it characterizes the flow capacity of water within the aquifer and is the primary target parameter for inversion. The specific storage coefficient is obtained by fitting the water level rise during the recovery phase; it characterizes the water storage and compressibility per unit volume. The delayed water production coefficient is determined by curve hysteresis analysis to determine whether the recovery process is delayed; it is applicable to free aquifers and is used to determine the water release rate. If multiple observation wells are configured, the aquifer's original parameters also include the anisotropy ratio (multiple well joint testing and directional water level differences can be fitted; multiple wells are required to obtain hydraulic anisotropy parameters). This ratio reflects the differences in horizontal and vertical water conductivity of the groundwater system and helps improve the accuracy of three-dimensional flow predictions in regional models. Therefore, this parameter system not only meets the basic inversion requirements for single-well configurations but also has the ability to be expanded to multi-well scenarios, demonstrating excellent applicability and scalability.

[0050] Furthermore, in the digital twin platform, the water flow process of the aquifer is simulated based on the geological conditions of the study area and the original parameters of the aquifer.

[0051] For each round of pumping test and the resting process after pumping, the trained temporal convolutional network is used to convert the data recorded in the pumping test into a low-dimensional intermediate vector; input: pumping rate, water level data sequence recorded during the pumping and resting process (this is a time sequence); output: intermediate vector.

[0052] The original parameters of the aquifer are input into a mapping model, and a vectorized expression of the simulated field is output as a comparison vector. The mapping model is a combination of a feedforward neural network and a self-attention encoding module. The low-dimensional structural features are encoded through the MLP to obtain a preliminary structural feature vector. The attention mechanism is then added to weight the influence of different parameters, thereby achieving vector conversion.

[0053] Temporal Convolutional Network (TCN) → Input to the intermediate vector extractor: pumping rate sequence Water level change sequence Concatenate the two into one input tensor:

[0054]

[0055] Model structure example: Conv1D (convolution kernel = 3, output channels = 64) + ReLU. Dropout + ResidualLayer (2 to 4 layers). Flatten → Dense (128) → output intermediate vector

[0056] Output: An intermediate vector that can compress and express the behavior of the hydrological response sequence.

[0057] MLP+self-attention network→simulated parameter vector expressor, the specific process is:

[0058] Input: Original parameter set: K is the permeability coefficient, R is the specific storage coefficient, W is the delayed water production coefficient, and Z is the anisotropy ratio.

[0059] First stage (MLP encoding): Dense(64)→ReLU→Dense(128)→get feature vector

[0060] The second stage (self-attention mechanism): Applying a multi-head self-attention mechanism, each head models the mutual influence between parameters and outputs a weighted fusion vector

[0061] Output: The vector representing the original parameters after nonlinear structure and weighting is used for comparison with the true response.

[0062] Calculate the distance L between the intermediate vector and the comparison vector as the fitness: L(θ) represents the distance between the intermediate vector and the comparison vector.

[0063] In the aquifer parameter space, starting with the original parameters, each parameter is randomly perturbed and its fitness is evaluated after the perturbation. If the fitness decreases, the aquifer parameters are updated to the perturbed values; if the fitness increases, no update is performed; if the fitness remains unchanged, the aquifer parameters before and after the perturbation are retained and the next perturbation round is performed. In the next perturbation round, the aquifer parameters with the minimum fitness are updated. (This optimization is not limited to two rounds of testing; according to this theory, multiple rounds of judgment can be performed until the "aquifer parameters with the minimum fitness" are obtained.) By introducing a perturbation iteration mechanism based on fitness feedback in the aquifer parameter space, the discrepancy between the hydrological response behavior reflected by the pumping test and the simulation model is effectively reduced, ultimately obtaining the optimal parameter combination. Starting with the original parameters as the initial solution, small random perturbations are gradually applied to each parameter, allowing for optimization without relying on gradient information. This approach is applicable to the nonlinear, multi-peak, and discontinuous response function structures commonly found in aquifer systems. By using a fitness function as the judgment basis, the perturbed parameters are updated when they reduce the error, and otherwise the current parameters are retained, ensuring that the overall optimization direction gradually converges toward the minimum error. When the perturbation result remains unchanged, the parameters before and after the perturbation are retained and updated selectively in subsequent iterations, further avoiding falling into local optimal solutions. This strategy has the advantages of strong robustness, efficient computation, and stable convergence. It can adapt to diverse scenarios such as inconsistent parameter sensitivity and complex response space in aquifers, ensuring that the final inversion results have higher credibility and generalization capabilities.

[0064] The perturbation and update process is repeated until the minimum fitness or convergence condition is met, and the iterative process ends. The parameters of the C group of aquifers are obtained by iterative optimization of C pumping tests. For each aquifer parameter, the maximum and minimum values of the C group of parameters (here "parameter" is a single coefficient in θ) are taken as the interval [s min , s max ]. By performing interval selection on each parameter in θ (or the parameter of each class) in the C group of parameters, the optimization of each parameter is finally completed.

[0065] By performing parameter perturbations and fitness assessments on each round of pumping test data, and independently obtaining Group C optimization results, we can effectively offset single-shot inversion biases caused by initial perturbation sensitivity, boundary perturbation effects, and data contingency, thereby improving the reliability and robustness of the overall parameter inversion. Furthermore, based on the Group C results, we extract the maximum and minimum values of each parameter across all rounds and construct a parameter interval representation. This not only provides more comprehensive inversion results but also provides a fault tolerance range and sensitivity analysis basis for subsequent numerical simulations.

[0066] The random perturbation includes, in {+Δ s , -Δ s ,0}, randomly select an adjustment direction and perturb the parameters; where, +Δ s Indicates adding an adjustment step to the sth parameter; -Δ s Indicates that the sth parameter is reduced by one adjustment step; 0 means no adjustment; s min Indicates the minimum value of the sth parameter in group C; s max represents the maximum value of the sth parameter in group C. The perturbation strategy employs a parameter-by-parameter, random direction fine-tuning mechanism, and introduces a three-state control scheme of "+1 / -1 / 0" in the perturbation direction. This makes the parameter search path more diverse, and the optimization process more robust and exploratory. Ultimately, through the triple mechanism of multiple rounds of optimization, random perturbation, and interval extraction, this design achieves a highly reliable reconstruction of the aquifer parameter space, providing a highly adaptable solution path for parameter inversion in nonlinear and complex hydrological scenarios.

[0067] S3: Based on the optimized aquifer parameters at the observation well location, synchronously optimize the aquifer parameters of the study area; and obtain the optimized aquifer parameters in the study area.

[0068] During the synchronous optimization process, if only one observation well is configured, geometric optimization of the aquifer parameters is performed for the entire study area; if multiple observation wells are configured, gradient optimization of the aquifer parameters for the entire study area is performed. If the initial data is optimized data, the initial data is represented as an interval; in this case, for a particular parameter, the minimum and maximum values before and after optimization are respectively mapped to each other, and the ratio is calculated. The ratios calculated using the minimum and maximum values are then used to perform global calculations. Specifically, if the initial parameters are in interval form, the maximum and minimum values before and after optimization are directly extracted as representatives, and their proportional relationship is calculated as the control factor for regional synchronous adjustment. In the specific implementation, the parameter value at any location in the region is multiplied by the proportional coefficient corresponding to the minimum and maximum values at the observation well, thereby generating the optimized interval value of the parameter at that location. This design is not only applicable to single-point value optimization, but is also compatible with interval optimization, with stronger model generalization capabilities and data adaptability. Compared with the traditional fixed-amplitude adjustment method, this method can keep the original spatial distribution structure of the parameters unchanged, while promoting the overall evolution of the system towards the credible interval. It is particularly suitable for complex hydrogeological environments where field survey data is incomplete or some parameters are estimated values. It helps to improve the controllability, stability and prediction accuracy of the regional parameter model.

[0069] The geometric optimization includes optimizing the parameters at all locations in the region at the same ratio according to the ratio of each parameter before and after optimization in the observation well.

[0070] At the observation well location, the ratio of the sth parameter before and after optimization is: and

[0071] In the region, the result after optimizing the sth parameter is: at position (x, y), the minimum value s after optimizing the sth parameter is min,(x,y) ; At position (x,y), the maximum value s after the sth parameter is optimized max,(x,y) .

[0072] in, s0 represents the value of the sth parameter before optimization at the observation well; s 0,(x,y) Represents the value of the sth parameter at position (x, y) before optimization.

[0073] The gradient optimization approach involves obtaining the spatial distribution of each optimized parameter in an m×n space and fitting a curve for each optimized parameter based on the distribution. For a study area, the coordinates of a location are input, and the curve is used to output the optimization results for each aquifer parameter at that location.

[0074] m×n observation wells have been laid out in the study area; inversion has been completed for each observation well, and the optimized parameter set has been obtained:

[0075] θ i,j ={K i,j ,R i,j ,W i,j ,Z i,j}

[0076] Where (i, j) represents the position of the well in the i-th row and j-th column in the spatial grid.

[0077] The fitting process is step 1: extract the spatial distribution data set. For each type of parameter (taking K as an example), extract the following structured data:

[0078] Dataset K ={(x i,j ,y i,j ,K i,j )|i=1..m,j=1..n}

[0079] Among them, (x i,j ,y i,j ) is the spatial coordinate of the i,j observation well; K i,j Optimize the permeability coefficient of the well. Construct similar data sets for parameters such as R, W, and Z.

[0080] Step 2: Fit the changing curve (or surface) For each parameter, use a spatial fitting algorithm to establish a functional relationship, for example:

[0081] K(x,y)=f K (x,y)

[0082] R(x,y)=f R (x,y)

[0083] W(x,y)=f W (x,y)

[0084] Z(x,y)=f Z (x,y)

[0085] In this embodiment, Gaussian process regression is used for fitting; in an optional embodiment, cubic spline interpolation, two-dimensional polynomial fitting, radial basis function (RBF) fitting and other strategies may also be used for fitting.

[0086] Step 3: Coordinate input → parameter output. * ,y * ), input to the fitting function:

[0087] θ * ={f K (x * ,y * ),f R (x * ,y * ),f W (x * ,y * ),f Z (x * ,y * )}

[0088] The optimized parameter value at that position can be obtained.

[0089] Step 4: Generate the global parameter field. Traverse the study area grid and call the fitting function in batches to generate the following parameter distribution maps: permeability coefficient field K(x,y), specific storage coefficient field R(x,y), delayed water production coefficient field W(x,y), and anisotropy ratio field Z(x,y). This will be used for subsequent work such as digital twin modeling and groundwater flow field simulation.

[0090] This gradient optimization strategy converts the discrete inversion results of multiple observation points into a continuously queryable and spatially perceptible regional parameter field by fitting the change curves (or change surfaces) of various parameters in space, thus realizing the expansion process from "point inversion" to "surface optimization". Compared with the global scaling method of proportional optimization, this scheme can more accurately reflect the gradient transition law of each parameter in the study area, and is particularly suitable for complex areas with obvious heterogeneity or drastic changes in geological conditions. This method has the advantages of result visualization, model interpolation, and structural scalability, providing high-resolution and continuous parameter input for applications such as groundwater simulation and pumping impact assessment.

[0091] Example 2: This example further provides an aquifer parameter optimization system based on data inversion, which includes:

[0092] The acquisition unit configures an observation well in the study area and uses the observation well to perform a water pumping test.

[0093] The analysis unit performs an inversion analysis on the original parameters of the aquifer at the observation well location based on the test data of the pumping test to obtain the optimized aquifer parameters.

[0094] The prediction unit performs synchronous optimization of the aquifer parameters in the study area according to the optimized aquifer parameters at the observation well location.

[0095] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0096] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0097] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0098] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0099] Example 3 is an embodiment of the present invention, which provides an aquifer parameter optimization method based on data inversion. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0100] The experimental area was a 4.5 km x 4.5 km groundwater development zone, primarily composed of interbedded sandy silt and fine sand. Nine observation wells, numbered P1 to P9, were evenly distributed in a 3 x 3 matrix, with a spacing of 1.5 km between wells.

[0101] Each well underwent three pumping-and-standing experiments at a sampling frequency of 1 Hz to generate a complete water level-time response sequence. All data were then fed into both the traditional Theis method and the proposed method for inversion, with the goal of accurately evaluating aquifer parameters for each well and the entire region.

[0102] Experimental statistics show that the average inversion error of the traditional method for nine wells was 23.7%, while the average error of the proposed method was only 1.3%, a significant reduction in error, demonstrating its significant advantage in point-level inversion accuracy. This error primarily stems from the traditional method's failure to account for dynamic variations in hydrological response behavior (such as hysteresis and perturbation inertia). However, the proposed method, through a vector space mapping mechanism, can deeply characterize the entire pumping response curve, achieving higher data fitting quality.

[0103] In regional prediction, this method further establishes a continuous spatial function through gradient fitting based on the inversion results of multiple well points, enabling high-precision predictions for non-observed points. The prediction errors for each point are kept within 2%, demonstrating the smoothness and spatial consistency of the constructed parameter distribution surface. This capability is crucial in practical groundwater modeling, enabling high-precision parameter mapping across a region and providing more reliable boundary input for subsequent flow field simulations, water supply scheduling, and other applications.

[0104] The invention not only improves the inversion accuracy of aquifer parameters (well points), but also realizes spatial generalization capability (regions), providing important support for the construction and calibration of groundwater models under complex geological conditions, and has outstanding practical value and promotion significance.

[0105] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for optimizing aquifer parameters based on data inversion, characterized in that: include: In the study area, an observation well is configured and a pumping test is performed using the observation well; Based on the test data of the pumping test, the original parameters of the aquifer at the observation well are inverted and analyzed to obtain the optimized aquifer parameters; Based on the optimized aquifer parameters at the observation well location, synchronously optimizing the aquifer parameters in the study area; In the synchronous optimization process, if only one observation well is configured, the aquifer parameters of the entire study area are optimized in a geometric manner; if multiple observation wells are configured, the aquifer parameters of the entire study area are optimized in a gradient manner.

2. The method for optimizing aquifer parameters based on data inversion according to claim 1, wherein: The configuration strategies of the observation wells include individual configuration and joint configuration; The separate configuration includes configuring only one observation well in the study area; the joint configuration includes configuring m×n observation wells in the study area according to a matrix position relationship; Wherein, n and m are both adjustable positive integer values.

3. The method for optimizing aquifer parameters based on data inversion according to claim 2, wherein: The pumping test includes obtaining initial static water level data of the observation well after standing still, which serves as a baseline; pumping water at a preset pumping rate; after completing one round of pumping test, standing still until the water level in the observation well reaches the baseline; after the water level in the observation well reaches the baseline, performing the next round of pumping test, and repeating the process C times; Where C represents the maximum number of tests; during the pumping and standing process, a sampling frequency of 1 Hz is used to record the water level data in the observation well; During each round of pumping test, the termination conditions include: condition 1: reaching the maximum pumping time; condition 2: the water level in the observation well is pumped to 0; when any termination condition is met, the current round of pumping test ends.

4. The method for optimizing aquifer parameters based on data inversion according to claim 3, wherein: The original parameters of the aquifer include, but are not limited to, permeability coefficient, specific storage coefficient, and delayed water production coefficient; If there are multiple observation wells, the original parameters of the aquifer also include anisotropy ratio.

5. The method for optimizing aquifer parameters based on data inversion according to claim 4, characterized in that: The inverse analysis includes simulating the water flow process of the aquifer based on the geological conditions of the study area and the original parameters of the aquifer in the digital twin platform; For each round of pumping test and the rest period after pumping, the trained temporal convolutional network is used to convert the data recorded in the pumping test into a low-dimensional intermediate vector. The input is: the pumping rate, and the water level data sequence recorded during the pumping and rest period. Output: intermediate vector; The original parameters of the aquifer are input into a mapping model, which outputs a vectorized representation of the simulated field as a comparison vector. The mapping model is a combination of a feedforward neural network and a self-attention encoding module. It uses an MLP to encode low-dimensional structural features to obtain a preliminary structural feature vector. An attention mechanism is then added to weight the influence of different parameters, thereby achieving vector conversion. Calculating the distance L between the intermediate vector and the comparison vector as fitness; In the parameter space of the aquifer, the original parameters are used as initial values, and each parameter is randomly perturbed. The fitness after the perturbation is evaluated. If the fitness decreases, the aquifer parameters are updated to the values after the perturbation; if the fitness increases, the aquifer parameters are not updated; if the fitness remains unchanged, the aquifer parameters before and after the perturbation are retained, and the next round of perturbations is performed respectively. In the next round of perturbations, the aquifer parameters with the minimum fitness are updated. The perturbation and update process is repeated until the minimum fitness or convergence condition is met, and the iterative process ends. The parameters of the C group of aquifers are obtained by iterative optimization of C pumping tests. For each aquifer parameter, the maximum and minimum values of the C group parameters are taken as the interval after each parameter is optimized [s min , s max ]; The random perturbation includes, in {+Δ s , -Δ s ,0}, randomly select an adjustment direction and perturb the parameters; where, +Δ s Indicates adding an adjustment step to the sth parameter; -Δ s Indicates that the sth parameter is reduced by one adjustment step; 0 means no adjustment; s min Indicates the minimum value of the sth parameter in group C; s max Indicates the maximum value of the sth parameter in group C.

6. The method for optimizing aquifer parameters based on data inversion according to claim 5, characterized in that: The geometric optimization includes optimizing the parameters at the same ratio at all locations in the area according to the ratio of each parameter before and after optimization in the observation well; At the observation well location, the ratio of the sth parameter before and after optimization is: and In the region, the result after optimizing the sth parameter is: at position (x, y), the minimum value s after optimizing the sth parameter is min,(x,y) ; At position (x,y), the maximum value s after the sth parameter is optimized max,(x,y) ; in, s0 represents the value of the sth parameter before optimization at the observation well; s 0,(x,y) Represents the value of the sth parameter at position (x, y) before optimization.

7. The method for optimizing aquifer parameters based on data inversion according to claim 6, characterized in that: The gradient optimization includes obtaining the distribution of each optimized parameter in an m×n space, and fitting a change curve between the position and each optimized parameter according to the distribution; In the study area, the location coordinates are input, and the optimization results of each parameter of the aquifer at the location are output using the change curve.

8. An aquifer parameter optimization system based on data inversion using the method according to any one of claims 1 to 7, characterized in that: The acquisition unit configures an observation well in the study area and uses the observation well to conduct a water pumping test; The analysis unit performs an inversion analysis on the original parameters of the aquifer at the observation well location based on the test data of the pumping test to obtain the optimized aquifer parameters; The prediction unit performs synchronous optimization of the aquifer parameters in the study area according to the optimized aquifer parameters at the observation well location.

9. A computer device comprising: A memory and a processor; the memory stores a computer program, wherein the processor implements the steps of the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images