An information-based simulation method and system for groundwater flow field

Abstract

The application provides a groundwater flow field information simulation method and system, including obtaining simulation basic data, obtaining hydrological event data through the simulation basic data, obtaining spatial correlation characteristics based on the hydrological event data and using clustering operation, and generating a geological hypothesis set; constructing an initial network topology framework and initializing through the simulation basic data and the geological hypothesis set, extracting an initial collaborative sub-network through the initial network topology framework and defining an execution rule; obtaining a correction parameter adjustment scheme set and a chaos parameter adjustment scheme set through the initial network topology framework respectively and generating a final scheme set, simulating based on the final scheme set to generate a deduction simulation result; performing cyclic deduction and obtaining a final simulation result based on a convergence condition, and providing a digital twin simulation method which can be based on and evolve with hydrogeological physical laws.

Description

Technical Field

[0001] This invention relates to the field of dynamic modeling technology, and more specifically, to an information-based simulation method and system for groundwater flow fields. Background Technology

[0002] In recent years, optimization algorithms, such as swarm intelligence algorithms, have been introduced into groundwater flow field simulation to perform global optimization during the simulation process. However, problems still exist when dealing with complex groundwater systems with strong heterogeneity and multi-peak characteristics. First, the optimization process is essentially a pure mathematical search, which is decoupled from physical laws such as aquifer structure and recharge / drainage mechanisms. This may lead to mathematically optimal but geologically unreasonable results. Second, in complex search spaces, the algorithm is still prone to premature convergence due to the early loss of population diversity, thus falling into suboptimal solutions. Therefore, there is an urgent need for a digital twin simulation method that can evolve based on hydrogeological and physical laws and evolve together with them. Summary of the Invention

[0003] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an information-based simulation method for groundwater flow fields, the method comprising:

[0004] Acquire basic simulation data, and obtain hydrological event data through the basic simulation data. Based on the hydrological event data, use clustering operations to obtain spatial correlation features and generate a set of geological hypotheses.

[0005] An initial network topology framework is constructed and initialized using simulation base data and a set of geological assumptions. Initial cooperative sub-networks are extracted from the initial network topology framework and execution rules are defined.

[0006] The set of corrected parameter adjustment schemes and the set of chaotic parameter adjustment schemes are obtained through the initial network topology framework, and the final scheme set is generated. Simulation is performed based on the final scheme set to generate the simulation results.

[0007] Perform iterative deductions and obtain the final simulation results based on the convergence conditions.

[0008] Furthermore, simulation baseline data is acquired, and hydrological event data is obtained from the simulation baseline data. Based on the hydrological event data, spatial correlation features are obtained through clustering operations, and a set of geological hypotheses is generated, including:

[0009] Acquire data of categories one, two, three, four, and five, and process them based on a preprocessing strategy;

[0010] The first, second, third, and fifth categories of data are respectively associated with the fourth category of data to obtain the basic simulation data;

[0011] Short-term separation is performed based on simulated basic data to obtain water level residual data, and event boundary points of the water level residual data are extracted to generate hydrological event data.

[0012] Based on the simulated basic data and hydrological event data, first-order clustering and second-order clustering are performed respectively to obtain a continuous spatial partition map, and at the same time, hydrological features and spatial correlation features are extracted.

[0013] Condition matching is performed based on hydrological features and spatial correlation features to obtain a set of geological hypotheses.

[0014] Furthermore, based on the simulated baseline data and hydrological event data, a first clustering is performed to obtain a discrete cluster distribution map. Based on this discrete cluster distribution map, a second clustering is performed to obtain a continuous spatial partition map, including:

[0015] Water level change curves are obtained by combining hydrological event data and simulation baseline data. Based on the water similarity matrix, a clustering is performed and the clusters are labeled to obtain a discrete distribution map of the clusters.

[0016] A monitoring well-event matrix is ​​constructed using hydrological event data, simulated basic data, and cluster labels. Secondary clustering is then performed based on the monitoring well-event matrix to obtain a continuous spatial partition map.

[0017] Hydrological characteristics and spatial correlation characteristics were obtained by using continuous spatial zoning maps and simulated basic data, respectively.

[0018] Furthermore, an initial network topology framework is constructed and initialized using simulation-based data and a set of geological assumptions. An initial cooperative sub-network is extracted from this initial network topology framework, and execution rules are defined, including:

[0019] Information element objects are instantiated based on simulated basic data, and an initial network topology framework is constructed using information element objects and based on connection relationships.

[0020] Initial estimates are obtained based on simulation data, and static attribute areas, dynamic attribute areas, and behavioral logic areas of information element objects are defined respectively. The information element objects are initialized based on the initial estimates and the set of geological assumptions.

[0021] The associated regional range is obtained based on the set of geological assumptions, and the initial collaborative sub-network is obtained from the initial network topology framework based on the associated regional range. At the same time, the initial collaborative sub-network is assigned a trust level and consensus label.

[0022] Define the rules for scheme generation, feedback, consensus reaching, and chaotic perturbation injection, respectively.

[0023] Furthermore, using the initial network topology framework, a set of corrected parameter adjustment schemes and a set of chaotic parameter adjustment schemes are obtained respectively, and a final scheme set is generated. Simulation is performed based on the final scheme set to generate inference simulation results, including:

[0024] The set of correction parameter adjustment schemes is obtained by using information element objects in the initial network topology framework and based on a consensus mechanism.

[0025] The chaotic parameter adjustment scheme set is obtained by executing the chaotic perturbation injection rule in the rule, and the final scheme set is obtained based on the chaotic parameter adjustment scheme set and the correction parameter adjustment scheme set.

[0026] Identify major bottleneck events based on the final solution set, obtain global optimization instructions through these major bottleneck events, and obtain update parameters and network consensus based on the global optimization instructions.

[0027] Based on the updated parameters, simulations are performed using a groundwater flow numerical simulation engine to generate inference simulation results.

[0028] Furthermore, a set of adjustment schemes for corrective parameters is obtained through information element objects in the initial network topology framework and based on a consensus mechanism, including:

[0029] The corresponding local observation error is obtained through the information element object;

[0030] By generating rule suggestions based on local observation errors and the schemes in the execution rules, and obtaining initial suggestion values ​​and an initial parameter adjustment scheme set based on the rule suggestion values;

[0031] The initial parameter adjustment scheme set is evaluated and feedback is provided based on the feedback rules in the execution rules, and a weighted consensus value is generated based on the consensus reaching rules in the execution rules.

[0032] A set of adjustment schemes for correction parameters is obtained based on a weighted consensus value and a judgment mechanism.

[0033] Furthermore, a set of chaotic parameter adjustment schemes is obtained by executing the chaotic perturbation injection rules in the execution rules. Based on the set of chaotic parameter adjustment schemes and the set of corrected parameter adjustment schemes, a final scheme set is obtained, including:

[0034] Whether to trigger chaotic perturbation injection is selected based on chaotic triggering conditions;

[0035] If triggered, the perturbation probability is obtained, and an information element object is randomly selected based on the perturbation probability. A chaotic sequence is generated using Logistic chaotic mapping. The perturbation term is obtained through the chaotic sequence. A set of chaotic parameter adjustment schemes is obtained based on the perturbation term and the set of correction parameter adjustment schemes. The perturbation term includes the direction sign and the perturbation amplitude.

[0036] The final set of solutions is obtained by combining the set of chaotic parameter adjustment schemes for the selected information element objects with the set of correction parameter adjustment schemes for the unselected information element objects.

[0037] Furthermore, the method also includes:

[0038] The convergence conditions include condition one, condition two, and condition three;

[0039] Condition one is that the current global objective function value has reached the set threshold or the improvement rate has stagnated;

[0040] Among them, condition two is that the consensus of the entire network is higher than a set threshold and the variance of the consensus sequence of the entire network is less than a set variance threshold.

[0041] Condition 3 is reaching the maximum number of iterations.

[0042] Furthermore, embodiments of the present invention also provide an information simulation system for groundwater flow fields, comprising:

[0043] A basic unit, which is used to acquire simulation basic data and generate a set of geological hypotheses based on the simulation basic data;

[0044] The building unit is used to build and initialize the initial network topology framework, and at the same time, the building unit is used to extract the initial cooperative sub-network and define the execution rules;

[0045] Execution unit, which is used to perform simulation through an initial network topology framework and generate simulation results;

[0046] A cyclic unit is used to perform cyclic deduction and obtain the final simulation result based on the convergence condition.

[0047] This embodiment fundamentally integrates physical mechanisms with data, and improves the global optimization and optimization efficiency in the deduction process through distributed negotiation and chaotic perturbation. It provides a digital twin simulation method that can evolve based on and co-evolve with hydrogeological and physical laws, thereby supporting the construction of a digital twin system that can be used for groundwater flow field simulation. Attached Figure Description

[0048] Figure 1 This is a flowchart of an information-based simulation method for groundwater flow field according to the present invention.

[0049] Figure 2 This is a flowchart of step C1 in the information simulation method for groundwater flow field of the present invention.

[0050] Figure 3 This is a flowchart of step C11 in the groundwater flow field information simulation method of the present invention.

[0051] Figure 4 This is a flowchart of step C12 in the information simulation method for groundwater flow field of the present invention.

[0052] Figure 5 This is a schematic diagram of an information simulation system for groundwater flow field according to the present invention. Detailed Implementation

[0053] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0054] The following is in conjunction with the appendix Figure 1 To be continued Figure 5 This invention provides a detailed description of an information-based simulation method and system for underground water flow fields.

[0055] Specifically, an information-based simulation method for groundwater flow fields includes the following steps:

[0056] Step A1: Obtain simulation base data, and obtain hydrological event data through simulation base data. Based on the hydrological event data, use clustering operations to obtain spatial correlation features and generate a set of geological hypotheses.

[0057] Furthermore, step A1 is performed through steps A11 to A15:

[0058] Step A11: Obtain the first, second, third, fourth and fifth types of data respectively, and process them based on the preprocessing strategy.

[0059] Step A12: Associate the first, second, third, and fifth types of data with the fourth type of data to obtain the basic simulation data.

[0060] Step A13: Perform short-term separation based on the simulated basic data to obtain water level residual data, extract event boundary points from the water level residual data, and generate hydrological event data through the event boundary points.

[0061] Step A14: Perform primary and secondary clustering based on the simulated basic data and hydrological event data respectively to obtain a continuous spatial partition map, and extract hydrological features and spatial correlation features.

[0062] Step A15: Perform condition matching based on hydrological features and spatial correlation features to obtain a set of geological hypotheses.

[0063] Specifically, step A14 is executed based on steps A141 to A144.

[0064] Step A141: Obtain water level change curves through hydrological event data and simulation baseline data, and perform similarity calculations based on the water level change curves to obtain a similarity matrix.

[0065] Step A142: Perform a clustering based on the similarity matrix and assign cluster labels to obtain a discrete distribution map of the clusters.

[0066] Step A143 involves constructing a monitoring well-event matrix using hydrological event data, simulated basic data, and cluster labels, and then performing secondary clustering based on the monitoring well-event matrix to obtain a continuous spatial partition map.

[0067] Step A144: Obtain hydrological characteristics and spatial correlation characteristics through continuous spatial zoning maps and simulated basic data, respectively.

[0068] For step A11, basic data including but not limited to Class I data, Class II data, Class III data, Class IV data and Class V data are obtained respectively. For ease of explanation, Class I data, Class II data, Class III data, Class IV data and Class V data will be collectively referred to as basic data in the following text.

[0069] One type of data represents time-series monitoring data, including water level time-series data from monitoring wells, pumping / recharge flow time-series data from centralized extraction wells or artificial recharge wells, and meteorological and hydrological related time-series data; the second type of data represents exploration data, including geological structure maps, borehole columnar sections, lithological zoning maps, stratigraphic profiles, and physical exploration data (which can be understood as the interpretation results of geophysical methods, usually represented as profile maps or three-dimensional physical property models); the third type of data represents remote sensing observation data, including time-series data of surface subsidence or uplift. The data includes: spatiotemporal series data (usually raster data) of soil moisture content, surface temperature, vegetation index, etc.; four types of data represent the basic data of the numerical model, including the predefined groundwater flow numerical model grid partition file of the simulation area (which should include model grid spatial coordinates, grid cell connection relationships, cell identifiers, etc.) and the possible value ranges of various hydrogeological parameters, such as permeability coefficient, specific yield, and storage rate; and five types of data represent the basic geographic information data, including basic geographic information data such as digital elevation model of the simulation area and water system distribution.

[0070] After acquisition, the spatial and temporal references of all basic data are unified. For spatial reference unification, regardless of the original coordinate system of the basic data, it is transformed to a planar coordinate system or a custom coordinate system consistent with the groundwater flow numerical model grid. For vector data in the basic data, coordinate transformation and reprojection are performed. For raster data in the basic data, resampling and registration are performed to ensure that its pixels are spatially aligned with the model grid. For temporal reference unification, a time axis is defined, and the timestamps of all time-series data in the basic data are unified to the same time zone and interpolated or aggregated to a unified time step. For non-equal interval data in the basic data, linear interpolation or cubic spline interpolation is used to generate a continuous sequence with a unified time step. Outlier detection and processing are performed on one type of data in the basic data. For example, statistical methods can be used to identify and remove obviously deviated values, and reasonable interpolation can be performed based on spatial proximity.

[0071] For step A12, after preprocessing, the coordinates of the monitoring well and other wells are obtained. The first-class data related to the monitoring well and other wells is assigned to the numerical model grid cells in the corresponding four-class data. Since the number of monitoring wells and other wells is limited and cannot cover all model grid cells, spatial interpolation is used to assign the first-class data to the remaining model grid cells, ensuring that model grid cells without wells have reference first-class data. The remaining first-class data is then assigned to the corresponding model grid cells. Vector surface data from the second-class data, such as geological maps, are analyzed using spatial overlay to determine the lithology of each model grid cell and assigned corresponding attribute labels. Simultaneously, the... For Class II data, such as two-dimensional profiles or three-dimensional physical property models of physical exploration data, spatial interpolation is used to map their attribute values ​​to each model grid cell, thereby forming a three-dimensional attribute field. For Class III data, such as planar raster data like remote sensing data, statistical correlation methods can be used to associate pixel-scale information with model grid cells. For example, the area-weighted average of each specific grid cell can be calculated based on the spatial relationship between the soil moisture raster and the numerical model grid cells. In addition, it is necessary to associate Class V data with specific grid cells. For example, the surface elevation can be assigned to the corresponding specific grid cell and given the corresponding attribute label, thereby generating the simulation base data.

[0072] Understandably, the simulation base data is a dataset that includes all the attributes of the model grid cells. Through the association of Class I, Class II, Class III, Class V and Class IV data, each model grid cell includes dynamic data, static data and driving data at each time step. The dynamic data should include dynamic data such as water level, soil moisture content, and landform variables. The static data should include static data such as lithology, grid cell spatial coordinates, and surface elevation. The driving data should include the source and sink terms that the model grid cell is subjected to, such as net rainfall, pumping volume and other external driving data that can have an impact. Different types of data in the driving data can be regarded as driving factors.

[0073] For step A13, the measured water level time series data for each monitoring well in the simulated baseline data is extracted. A moving average is used to separate data representing long-term trends and seasonal cycles, thereby obtaining water level residual data containing short-term event signals and random noise. This eliminates interference from slow processes such as climate change and long-term mining on the identification of short-term event signals. Next, for the random noise in the water level residual data, the data can be input into the VMD algorithm for decomposition, thereby obtaining higher-frequency and lower-frequency IMF components. Typically, higher-frequency IMF components contain noise and rapid micro-fluctuations, while lower-frequency IMF components contain smooth, continuous signals caused by short-term events. At this point, the IMF component matching the spectral characteristics of the short-term event is selected for reconstruction, thereby obtaining enhanced water level residual data. Then, the water level residual data is used as the observation... The measured values ​​are input into the BOCPD algorithm, which outputs the probability of a sudden change at each time point. When the probability exceeds a preset threshold, that time point is determined to be an event boundary point, i.e., the start or end of the corresponding event. The event boundary points are corrected by combining the driving vectors in the simulation base data. For example, the start point of a water level rise event should match a significant rainfall event or river flooding in time. Multiple event boundary points that are temporally adjacent and triggered by the same driving factor are merged into a complete event. All complete events are integrated into hydrological event data and assigned attributes. The attributes should at least include the event identifier and type, start / end time, main driving source and intensity index. The intensity index can be selected as the intensity of the corresponding driving factor, such as total rainfall. The main driving source is represented by the associated rainfall process, pumping well group, etc.

[0074] For step A14, for each event in the hydrological event data, the water level change curve corresponding to the start / end event is extracted from the measured water level time series data of each monitoring well in the simulation base data. Each curve is standardized by Z-score standardization to eliminate the influence of elevation differences. Since there are spatial differences in hydraulic conductivity in groundwater simulation, the influence of different locations on the same event may have significant event lag. If Euclidean distance is used to measure similarity, when two curves are similar in shape but have time lag, Euclidean distance will misjudge them as dissimilar. Therefore, the DTW algorithm is used to calculate the similarity distance between any two water level change curves, thereby generating an N×N similarity distance matrix. The smaller the similarity distance between matrix elements, the more similar the monitoring wells are to the event. After obtaining the similarity distance matrix, clustering is performed based on the similarity distance matrix and the DPC clustering algorithm. Each monitoring well is then assigned a cluster label and mapped back to the model grid cell to generate a cluster discrete distribution map. The above steps are repeated to generate a cluster discrete distribution map for each event.

[0075] Understandably, for each event in hydrological event data, analyzing the water level dynamics of all monitoring wells in the corresponding event process and clustering monitoring wells with high similarity into the same cluster can reveal the implicit spatial correlation pattern, thereby transforming time series data into spatial partitions.

[0076] Next, a monitoring well-event matrix is ​​constructed, where rows represent all monitoring wells and columns represent hydrological event data. Matrix elements store the cluster label of the monitoring well in the corresponding event. If a monitoring well has no data in an event, a special value is assigned to the corresponding matrix position. Each row of the matrix is ​​regarded as a high-dimensional feature vector, and dimensionality reduction is performed based on principal component analysis to extract the main variation patterns. Clustering is then performed in the low-dimensional principal component space to obtain clustering results, which are then assigned as labels to the corresponding model grid cells. Finally, based on the model grid cells with clustering result labels, a region growing algorithm is used to expand the model grid to the entire simulation area, generating continuous spatial partition surfaces. At the same time, the partition boundaries are reasonably repaired by combining the static data in the simulation base data, such as repairing ridge lines and river boundaries, in order to obtain a continuous spatial partition map covering the simulation area.

[0077] For each region within the continuous spatial zoning map, its hydrological characteristics are extracted by combining the simulation base data, including lag characteristics, amplitude characteristics, shape characteristics, and sensitivity characteristics. The lag characteristic represents the average lag time between the time when the water level begins to change significantly and the start time of the event. The amplitude characteristic represents the average water level amplitude caused by intensity indicators in the monitoring wells within the region during various events, which can reflect the overall water storage and release capacity of the corresponding region. The shape characteristic represents the statistical characteristics of the original water level curve corresponding to the region, such as the ratio of the slope of the rise to the fall, which reflects the dynamic characteristics of its hydraulic conduction and release. The sensitivity characteristic represents the morphological differences of the region to the main driving source, which can indirectly reflect its connectivity with different driving sources. Thus, the continuous spatial zoning map can describe which regions within the simulation area consistently exhibit cooperative behavior under various hydraulic disturbances, which can indirectly reflect good hydraulic connectivity, and which regions behave independently or with lag, which can indirectly indicate the existence of isolated or weakly permeable areas.

[0078] By obtaining spatial correlation features between each region within the continuous spatial partition map and the simulation base data, and various geological features, including spatial overlap features and morphological consistency features, the spatial overlap features represent the percentage of area within the region that is marked as a specific geological feature, such as a fault zone or a high resistivity anomaly zone. The morphological consistency features represent the strike similarity or average distance between the boundary of the calculated region and the geological features.

[0079] For step A15, the spatial correlation features are combined with the hydrological features of the corresponding region and matched with a preset rule base. Each rule in the preset rule base adopts the form of condition-hypothesis-confidence. For the condition part, it is a logical combination of multiple facts, such as the condition (fault zone greater than 70%). (Fault line greater than 0.8). For the hypothesis part, to propose geological hypotheses, such as "the area constitutes a dominant runoff channel controlled by a fault", for the confidence part, the initial confidence of the hypothesis can be calculated by a preset function based on the degree of condition matching and the input value. For example, the confidence can be a weighted comprehensive score of various spatial correlation features and hydrological features, which needs to be normalized to the range of 0 to 1. During the matching process, the same area may trigger multiple rules and generate multiple hypotheses. At this time, the highest hypothesis needs to be set as the geological hypothesis based on the confidence. If the confidence is close and not geologically mutually exclusive, a composite geological hypothesis can be generated. Thus, all geological hypotheses are encapsulated into a geological hypothesis set, where each geological hypothesis can contain a unique identifier, hypothesis content description, hypothesis type, associated regional range, inference conditions, etc. The geological hypothesis set is transformed into initialization instructions and constraint rules that can be directly used in step B1. For example, the dominant runoff channel hypothesis is mapped to adding a rule with high priority for permeability coefficient collaborative adjustment to the behavioral logic area of ​​all information element objects in the area.

[0080] Step B1: Construct an initial network topology framework and initialize it using simulation base data and a set of geological assumptions. Extract initial collaborative sub-networks from the initial network topology framework and define execution rules.

[0081] Furthermore, step B1 is performed through steps B11 to B14:

[0082] Step B11: Instantiate information element objects based on simulated basic data, and construct an initial network topology framework based on the information element objects and connection relationships.

[0083] Step B12: Obtain initial estimates based on simulation data, and define the static attribute area, dynamic attribute area, and behavioral logic area of ​​the information element object respectively. Initialize the information element object based on the initial estimates and the geological hypothesis set.

[0084] Step B13: Obtain the associated regional range based on the set of geological assumptions, obtain the initial collaborative sub-network through information element objects and the associated regional range, and assign trust and consensus labels to the initial collaborative sub-network.

[0085] Step B14 defines the scheme generation rules, feedback rules, consensus-reaching rules, and chaos perturbation injection rules, respectively.

[0086] For step B11, all model grid cells in the simulation base data are read. For each model grid cell, a corresponding information element object with a unique identifier is instantiated within the network framework. The information element object is directly mapped to the index of the model grid cell, for example, using a triple (X, Y, Z), where X represents the Xth layer, Y represents the Yth row, and Z represents the Zth column of the model grid cell. Then, based on the connection relationship of the model grid cells, the direct neighborhood of each information element object is determined. The direct neighborhood includes horizontal and vertical neighborhoods. For horizontal neighborhoods, the information element objects adjacent to it in the row and column directions in the same layer are included. For vertical neighborhoods, the information element objects adjacent to it in the upper and lower layers are included. At the same time, a neighborhood relation table is generated to store the corresponding neighborhood relations. All information element objects and their neighborhood relations are organized into an initial network topology framework, where the nodes are information element objects and the edges are defined by the neighborhood relation table.

[0087] For step B12, initial estimates of hydrogeological parameters and their reasonable physical adjustment ranges are obtained from the simulation baseline data through regional experience or experiments. Based on the initial network topology framework, static attribute areas, dynamic attribute areas, and behavioral logic areas are defined and initialized for each information element object. The static attribute area stores the current hydrogeological data and the physical range within which parameters can be adjusted. The dynamic attribute area stores the water level and other results of the information element object after the latest round of simulation, the error and historical error of the information element object, its own proposed historical parameter adjustment scheme, and neighbor feedback. The behavioral logic area is a rule base, initially empty, and will be loaded later. It also stores the weight vector of the local optimization objective, which can be dynamically adjusted according to global optimization instructions. Initial estimates are assigned to each information element object, and the physical adjustment range is loaded into the static attribute area. For some special information element objects, such as bedrock outcrops, the upper limit of their permeability coefficient may be set to an extremely low value. At the same time, the set of geological assumptions is loaded into the behavioral logic area. The rule base uses a two-layer structure. The architecture includes general basic rules and specific rules. The basic rules are common rules loaded for all information element objects. For example, rule 1 calculates a basic parameter adjustment direction scheme based on the latest observation error. Rule 2 requires that when generating a scheme, the relative relationship with the parameter values ​​of the neighboring domain should be considered to avoid physically unreasonable drastic jumps. The specific rules are rules loaded only for specific information element objects based on the mapping of the geological hypothesis set. For example, for information element objects marked as covered by the dominant runoff channel hypothesis, rule 3 is loaded to adjust the permeability coefficient with an initiative weight of 0.9. In the negotiation, the scheme with increased parameter value is given higher support. Then, the dynamic attribute area is initialized to empty or default value, and an initial weight is assigned to the rule base according to the loaded rules. General basic rules can be assigned a preset basic weight, while specific rules can be assigned an initial weight that is proportional to the confidence level of their geological hypothesis or directly select the confidence level. For example, the rule corresponding to a geological hypothesis with a confidence level of 0.8 may have an initial weight set to 0.8. At this point, the definition and initialization of the information element object are completed.

[0088] For step B13, for each set of geological hypotheses in the set of geological hypotheses, obtain its associated region range, that is, the region of all model grid cells covered by the hypothesis. Based on the associated region range, check whether each information element object appears in the associated region range of the above hypothesis. If it appears, assign a consensus label to the information element object, that is, the unique identifier of the hypothesis. An information element object may belong to multiple non-mutually exclusive geological hypotheses at the same time, so it can have multiple consensus labels. If there is an information element object that is not in the associated region range of any geological hypothesis, its consensus label can be set as the default label. After the consensus label is assigned, identify information element objects with the same consensus label and regard these information element objects as an initial cooperative sub-network.

[0089] Then, an M×M trust matrix is ​​initialized, where M is the total number of information element objects. Matrix element T12 represents the initial trust level of information element object 1 towards information element object 2, with the initial trust level ranging from 0 to 1. The principle for assigning trust levels is that the more similar the perceptions, the higher the trust level. The trust level assignment rules include internal trust level assignment, external trust level assignment, and special trust level assignment. Internal trust level assignment means that for two information element objects belonging to the same initial collaborative sub-network, their initial trust level is set to a higher value, such as 0.7-0.9, reflecting that they are bound by the same geological assumption. The bundle is expected to have strong synergy in the adjustment of behavior and parameters. The external trust degree assignment is a low base value, such as 0.3-0.5, for two information element objects that do not belong to the same initial cooperative sub-network. The special trust degree assignment means that the diagonal element is set to 1, while for information element objects that are direct neighbors but do not belong to the same initial cooperative sub-network, a small bonus can be given on the external trust degree to reflect the potential impact of physical connection. The trust degree is then assigned to the corresponding information element object. At this time, the initial cooperative sub-network has trust degree and consensus label.

[0090] For step B14, the following rules are defined: scheme generation rule, feedback rule, consensus reaching rule, and chaotic perturbation injection rule. The scheme generation rule calculates a basic adjustment amount for its own parameters based on its rule base and the historical simulation and observation errors of the information element object. The specific calculation formula can be combined with its rule base weights; for example, basic adjustment amount = Σ(weight of the k-th rule * specific adjustment suggestion value obtained according to the k-th rule). The feedback rule requires that after an information element object broadcasts a scheme to its neighborhood, the neighboring information element objects must evaluate and provide feedback based on their own state. This feedback includes a support coefficient, ranging from -1 to 1. The support coefficient quantifies the degree of acceptance of the scheme by the neighboring information element objects; positive values ​​indicate support, and negative values ​​indicate opposition. The specific evaluation logic is based on the rule base. The consensus reaching rule requires that after collecting feedback from all neighbors (neighboring information element objects), the information element object obtains a weighted consensus value. The consensus for calculating the weighted consensus value is:

[0091] ;

[0092] in, This is represented as the weighted consensus value of information element object i. This represents the trust level of information element object i towards its neighboring information element object j. This is represented as the support coefficient of information element object j to information element object i. This is represented by the information meta-object j belonging to the neighbor. The closer a value is to 1, the stronger the consensus support.

[0093] The chaotic perturbation injection rule is used to prevent the network from getting stuck in a local consensus deadlock. Its trigger condition is that when the average objective function value of the entire network improves by less than the improvement threshold for N consecutive rounds, it is triggered. After triggering, a portion of information element objects are randomly selected with the chaotic perturbation injection probability. A perturbation term generated by the Logistic chaotic mapping sequence is superimposed on the final solution formed by the normal rules. The assignment of the perturbation term should be limited to a reasonable range of its parameters, but its direction can be different from the gradient or consensus direction, thereby providing better exploration capabilities.

[0094] Step C1: Obtain the set of corrected parameter adjustment schemes and the set of chaotic parameter adjustment schemes respectively through the initial network topology framework, and generate the final scheme set. Perform simulation based on the final scheme set to generate the simulation results.

[0095] Furthermore, step C1 is performed through steps C11 to C14:

[0096] Step C11: Obtain a set of adjustment schemes for correction parameters by using the information element objects in the initial network topology framework and based on the consensus mechanism.

[0097] Step C12: Obtain a set of chaotic parameter adjustment schemes by executing the chaotic perturbation injection rules in the execution rules, and obtain the final scheme set based on the set of chaotic parameter adjustment schemes and the set of corrected parameter adjustment schemes.

[0098] Step C13: Identify the main bottleneck events based on the final solution set, obtain global optimization instructions through the main bottleneck events, and obtain update parameters and network consensus based on the global optimization instructions.

[0099] Step C14: Based on the updated parameters, simulate the groundwater flow using a numerical simulation engine to generate the simulation results.

[0100] Specifically, step C11 is executed based on steps C111 to C114.

[0101] Step C111: Obtain the corresponding local observation error through the information element object.

[0102] Step C112: Generate rule suggestion values ​​by using local observation errors and generating rules based on the schemes in the execution rules; obtain initial suggestion values ​​and an initial parameter adjustment scheme set based on the rule suggestion values.

[0103] Step C113: Evaluate and provide feedback on the initial parameter adjustment scheme set based on the feedback rules in the execution rules, and generate a weighted consensus value based on the consensus achievement rules in the execution rules.

[0104] Step C114: Obtain a set of adjustment schemes for correction parameters based on the weighted consensus value and using a judgment mechanism.

[0105] Specifically, step C12 is executed based on the value of step C121 in step C123.

[0106] Step C121: Select whether to trigger chaotic perturbation injection based on the chaotic triggering conditions.

[0107] Step C121: If triggered, obtain the perturbation probability, randomly select an information element object based on the perturbation probability, generate a chaotic sequence using Logistic chaotic mapping, obtain the perturbation term through the chaotic sequence, and obtain a set of chaotic parameter adjustment schemes based on the perturbation term and the set of correction parameter adjustment schemes, wherein the perturbation term includes the direction sign and the perturbation amplitude.

[0108] Step C123: Obtain the final set of schemes based on the set of chaotic parameter adjustment schemes for the selected information element objects and the set of correction parameter adjustment schemes for the unselected information element objects.

[0109] For step C11, after defining and initializing the initial network topology, initial cooperative subnetwork, and execution rules in step B1, each information element object first calculates its local observation error. The calculation method depends on its data association method. If there is a direct observation well, the calculation is based on the formula: Local Observation Error = Measured Water Level - Water Level Value of the Corresponding Grid Cell Obtained in the Previous Simulation. If it is an interpolated reference value, the calculation is based on the formula: Local Observation Error = Reference Water Level - Water Level Value of the Corresponding Grid Cell Obtained in the Previous Simulation. Simultaneously, each information element object updates its dynamic attribute area with the local observation error and combines the current error with historical errors for analysis to obtain its trend and magnitude. Then, the information element object activates all relevant rules in its rule base. Each rule outputs a rule suggestion value and the strength of the suggestion based on the current historical error, the current permeability coefficient value, and the status of neighboring parameters. For example, a general basic rule might output a suggestion = (Sensitivity Coefficient × Local Observation Error), where the sensitivity coefficient... The number needs to be predefined, and the strength is the absolute value of the error. For example, a special rule might output a suggestion = β × (maximum physical range of permeability coefficient - current parameter value), where β is the weight, which can encourage the parameter to adjust towards the high value direction of the dominant runoff channel. The strength can be determined by the rule weight and confidence level. Each information element object performs a weighted fusion of all output rule suggestion values ​​to generate a basic adjustment scheme. The weighting coefficient combines the inherent weight of each rule and the strength obtained this time. The basic formula for calculation can be expressed as: Basic adjustment scheme = (Σ(weight of the kth rule × strength of the kth rule × suggestion value of the kth rule)) / (Σ(weight of the kth rule × strength of the kth rule)). The basic adjustment scheme is obtained in this way. After obtaining it, the basic adjustment scheme also needs to be physically trimmed to ensure that the current permeability coefficient value + the basic adjustment scheme does not exceed the physical range defined by the rule. The value after trimming is the initial parameter adjustment scheme of the information element object. This is used to obtain a set containing each initial parameter adjustment scheme.

[0110] Each information element then broadcasts its initial plan to all its direct neighbors listed in its neighborhood table via a message passing mechanism. Upon receiving a plan from neighbor i, any information element j evaluates it according to feedback rules. This evaluation process simulates a simplified local physical prediction, including virtual impact assessment and benefit consistency assessment. Virtual impact assessment involves information element j estimating the potential virtual impact on its own state if its parameters are adjusted according to the initial parameter adjustment plan, based on the current local flow field relationship. This primarily affects the simulated water level and can be approximated using Darcy's law. Benefit consistency assessment involves information element j comparing the estimated virtual impact with its own objective, i.e., reducing error. If the virtual impact helps reduce error, information element j tends to support the plan; otherwise, it opposes it. Then, information element j generates structured feedback, which includes a support coefficient as described above. Subsequently, it obtains the weighted consensus value mentioned above. After obtaining the value, it compares the weighted consensus value with a preset consensus value threshold. If the weighted consensus value is greater than the threshold, it indicates that the scheme has obtained strong consensus in the neighborhood, and information element object j can adopt the scheme. If the weighted consensus value is less than the threshold, it indicates that the scheme is controversial, and information element object j needs to modify the original scheme or adopt the neighbor's scheme, thereby adjusting the scheme direction to be more in line with the collective interest. In some possible embodiments, a simple multi-round mechanism can also be designed to improve the negotiation quality. For example, after the first round of modification, information element object j can broadcast the scheme again for further processing until the consensus value meets the threshold or the maximum number of negotiations is reached. Finally, the information element object can obtain the set of schemes to adjust the modification parameters.

[0111] For step C12, the chaotic perturbation injection rule is enabled and a global objective function is selected. The global objective function is used to calculate a single scalar value, i.e., the global objective function value, from the error set obtained in the previous round of deduction through a predefined aggregation function. This value is used to quantify the overall poor fit under the current parameters. The smaller the value, the better the fit. The aggregation function can use the commonly used weighted root mean square error. If the global objective function value is less than a preset threshold for N consecutive rounds, it indicates that it may be trapped in a local optimum. Or, if the variance of the overall solution is lower than a certain threshold, it indicates that the group's decision-making tends to be homogeneous. If any of the above triggering conditions are met, chaotic perturbation injection is enabled; otherwise, it is disabled. Once enabled, the perturbation probability is obtained. This probability is not fixed but adaptively adjusted based on the degree of stagnation. For example, perturbation probability = min(base perturbation probability × (1 + number of stalled rounds / number of rounds performed)). The longer the stagnation, the higher the perturbation probability. Information element objects are randomly selected based on the perturbation probability. For each selected information element object, a chaotic sequence is generated iteratively using a chaotic mapping, illustrated using a Logistic chaotic mapping. The equation is:

[0112] ;

[0113] in, The value of the chaotic variable in the nth iteration is represented by the interval between 0 and 1. Represented as control parameters, when When the value is 4, the system is in a completely chaotic state. Through this mapping, iteratively starting from a random initial value, a chaotic sequence distributed in the 0-1 interval is generated. Then, the direction and amplitude of the disturbance are calculated. By transformation, the chaotic sequence is mapped to the -1 to 1 interval, resulting in a random direction sign, where the direction sign = 2 × chaotic sequence - 1. For the disturbance amplitude, the base value can be set to a small proportion of the allowable adjustment range of this parameter, such as 0.5, thus obtaining the final disturbance term = direction sign + disturbance amplitude. The chaotic disturbance term is then synthesized with the correction parameter adjustment scheme of the selected information element object to generate a chaotic parameter adjustment scheme. At the same time, the physical rationality of the chaotic parameter adjustment scheme must be ensured. If it exceeds the limit, it is pruned. For those information element objects that are not selected, their correction parameter adjustment scheme is their final scheme. The final scheme set is obtained by combining the correction parameter adjustment scheme set and the chaotic parameter adjustment scheme set.

[0114] For step C13, during the iterative forward inference process, the region with the largest error is continuously obtained through the final solution set. Based on the time period and spatial distribution characteristics of this error (e.g., concentrated in a certain partition), and using the region with the largest error, the decoder identifies the main bottleneck events hindering the improvement of the global objective function value. For example, if region A has an error of 40% in a rainfall event, the decoder generates a global optimization instruction. This instruction should include the instruction type, target object, action parameters, and instruction strength. The instruction type represents priorities such as priority increase or constraint reduction. The target object clarifies the scope of the information element objects involved and can be defined through consensus tags. The action parameters represent the specific adjustment amount, and the instruction strength represents the quantitative intensity of the main bottleneck event, which can be based on the ratio of errors. For example, the global optimization instruction is broadcast to all information element objects in the initial network topology framework. After receiving the global optimization instruction, each information element obtains the action to be executed based on the global optimization instruction and the rule base. First, it determines whether it belongs to the area. If it does, it modifies its temporary state according to the global optimization instruction, such as temporarily increasing the weight of the local optimization target. If it does not belong, it does not take any action. If it does not directly belong but is closely connected to the area hydraulically, it can be judged through adjacency relationship and historical negotiation record, and may make fine adjustments to indicate cooperation. In this way, a global execution plan is obtained. Then, for each information element object, its final weighted consensus degree for global execution is obtained, and the overall network consensus degree is also obtained. The consensus degree of the overall network is calculated as (1 / M×Σ weighted consensus degree), where M is the total number of information element objects. The overall network consensus degree can reflect the overall cooperation and consistency.

[0115] After confirming all final weighted consensus degrees, each information element object executes its parameter update operation based on global optimization instructions and checks the physical boundary range, thereby outputting the updated parameters after a complete round of deduction.

[0116] Step C14: Obtain the physical value range conditions, encapsulate the updated parameters and the physical value range conditions, and input them into a groundwater flow numerical simulation engine, such as FEFLOW. The groundwater flow numerical simulation engine solves the three-dimensional governing equations describing groundwater flow motion based on the provided input, such as the differential partial equations under confined / unconfined flow conditions. For unsteady flow, the equations can be expressed as:

[0117] ;

[0118] in, Represented as the divergence operator, Represented as the gradient operator, K(x,y,z) represents the permeability coefficient at spatial coordinates (x,y,z), obtained from the update parameters; h(x,y,z,t) represents the water level distribution to be calculated, which is a function of space and time; and S represents the storage rate. The rate of change of water level over time is represented by W(x,y,z,t), which represents the source and sink terms. The groundwater flow numerical simulation engine discretizes, assembles, and iteratively solves the equations using numerical methods such as finite difference, finite element, or finite volume methods until the convergence criteria are met. After the groundwater flow numerical simulation engine finishes running, the spatiotemporal water level field is extracted from the output. The spatiotemporal water level field represents the simulated water level values ​​at all observation wells and all calibration time steps. Key flux data are represented as data used for auxiliary analysis, such as flow rate through a specific cross section and exchange between aquifer and river. The simulated water level values ​​are spatiotemporally aligned and matched with the corresponding monitoring wells and the measured water levels at the corresponding time steps to generate water level pairs. For each valid water level pair, its error is calculated. The spatiotemporal water level field, key flux data, error, and water level pairs are stored as the simulation results.

[0119] Step D1: Perform iterative deduction and obtain the final simulation results based on the convergence conditions.

[0120] For step D1, during the iterative simulation, the improvement rate of the global objective function value in each round compared to the previous round is calculated. Simultaneously, the moving average improvement rate of the global objective function value in the most recent round A is calculated, and the network consensus is obtained. Then, a judgment is made based on convergence conditions, including condition one, condition two, and condition three. Condition one is that the current global objective function value has reached a set threshold or the improvement rate has stagnated, with stagnation indicating a small improvement rate. Condition two is that the network consensus is higher than a set threshold and the variance of the network consensus sequence is less than a set variance threshold. Condition three is reaching the maximum number of iterations. When judging based on convergence conditions, there are two judgment results: convergence and non-convergence. Convergence can be determined based on actual needs, satisfying any one of conditions one, two, or three, where any one of these conditions should be greater than 1. Non-convergence indicates that any condition is not met, causing the simulation to return to step C1 for re-simulation. After convergence, the final simulation result is output.

[0121] Steps A1 to D1 above can be encapsulated into a digital twin system for simulating groundwater flow fields, and subsystems such as graphical user interfaces, data management, and visualization can be added.

[0122] Specifically, an information simulation system for groundwater flow fields includes:

[0123] The basic unit 100 is used to acquire simulation basic data and generate a set of geological hypotheses through the simulation basic data.

[0124] The construction unit 200 is used to construct and initialize the initial network topology framework, and at the same time, the construction unit 200 is used to extract the initial cooperative sub-network and define the execution rules.

[0125] The execution unit 300 is used to perform simulations using an initial network topology framework and generate simulation results.

[0126] The loop unit 400 is used to perform loop deduction and obtain the final simulation result based on the convergence condition.

[0127] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0128] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An information-based simulation method for groundwater flow field, characterized in that, It includes the following steps: Acquire basic simulation data, and obtain hydrological event data through the basic simulation data. Based on the hydrological event data, use clustering operations to obtain spatial correlation features and generate a set of geological hypotheses. An initial network topology framework is constructed and initialized using simulation base data and a set of geological assumptions. Initial cooperative sub-networks are extracted from the initial network topology framework and execution rules are defined. The set of corrected parameter adjustment schemes and the set of chaotic parameter adjustment schemes are obtained through the initial network topology framework, and the final scheme set is generated. Simulation is performed based on the final scheme set to generate the simulation results. Perform iterative deductions and obtain the final simulation results based on the convergence conditions.

2. The method for information-based simulation of groundwater flow field according to claim 1, characterized in that, Acquire basic simulation data, and then obtain hydrological event data from the simulation data. Based on the hydrological event data, perform clustering operations to obtain spatial correlation features, and generate a set of geological hypotheses, including: Acquire data of categories one, two, three, four, and five, and process them based on a preprocessing strategy; The first, second, third, and fifth categories of data are respectively associated with the fourth category of data to obtain the basic simulation data; Short-term separation is performed based on simulated basic data to obtain water level residual data, and event boundary points of the water level residual data are extracted to generate hydrological event data. Based on the simulated basic data and hydrological event data, first-order clustering and second-order clustering are performed respectively to obtain a continuous spatial partition map, and at the same time, hydrological features and spatial correlation features are extracted. Condition matching is performed based on hydrological features and spatial correlation features to obtain a set of geological hypotheses.

3. The method for information-based simulation of groundwater flow field according to claim 2, characterized in that, A primary clustering process is performed based on simulated baseline data and hydrological event data to obtain a discrete cluster distribution map. A secondary clustering process is then performed based on this discrete distribution map to obtain a continuous spatial partition map, including: Water level change curves are obtained by combining hydrological event data and simulation baseline data. Based on the water similarity matrix, a clustering is performed and the clusters are labeled to obtain a discrete distribution map of the clusters. A monitoring well-event matrix is ​​constructed using hydrological event data, simulated basic data, and cluster labels. Secondary clustering is then performed based on the monitoring well-event matrix to obtain a continuous spatial partition map. Hydrological characteristics and spatial correlation characteristics were obtained by using continuous spatial zoning maps and simulated basic data, respectively.

4. The method for information-based simulation of groundwater flow field according to claim 1, characterized in that, An initial network topology framework is constructed and initialized using simulation data and a set of geological assumptions. Initial cooperative subnetworks are extracted from this framework, and execution rules are defined, including: Information element objects are instantiated based on simulated basic data, and an initial network topology framework is constructed using information element objects and based on connection relationships. Initial estimates are obtained based on simulation data, and static attribute areas, dynamic attribute areas, and behavioral logic areas of information element objects are defined respectively. The information element objects are initialized based on the initial estimates and the set of geological assumptions. The associated regional range is obtained based on the set of geological assumptions, and the initial collaborative sub-network is obtained from the initial network topology framework based on the associated regional range. At the same time, the initial collaborative sub-network is assigned a trust level and consensus label. Define the rules for scheme generation, feedback, consensus reaching, and chaotic perturbation injection, respectively.

5. The method for information-based simulation of groundwater flow field according to claim 1, characterized in that, Using the initial network topology framework, sets of corrected parameter adjustment schemes and chaotic parameter adjustment schemes are obtained respectively, and a final scheme set is generated. Simulations are performed based on the final scheme set to generate simulation results, including: The set of correction parameter adjustment schemes is obtained by using information element objects in the initial network topology framework and based on a consensus mechanism. The chaotic parameter adjustment scheme set is obtained by executing the chaotic perturbation injection rule in the rule, and the final scheme set is obtained based on the chaotic parameter adjustment scheme set and the correction parameter adjustment scheme set. Identify major bottleneck events based on the final solution set, obtain global optimization instructions through these major bottleneck events, and obtain update parameters and network consensus based on the global optimization instructions. Based on the updated parameters, simulations are performed using a groundwater flow numerical simulation engine to generate inference simulation results.

6. The method for information-based simulation of groundwater flow field according to claim 5, characterized in that, The set of adjustment schemes for corrective parameters is obtained by using information meta-objects in the initial network topology framework and based on a consensus mechanism, including: The corresponding local observation error is obtained through the information element object; By generating rule suggestions based on local observation errors and the schemes in the execution rules, and obtaining initial suggestion values ​​and an initial parameter adjustment scheme set based on the rule suggestion values; The initial parameter adjustment scheme set is evaluated and feedback is provided based on the feedback rules in the execution rules, and a weighted consensus value is generated based on the consensus reaching rules in the execution rules. A set of adjustment schemes for correction parameters is obtained based on a weighted consensus value and a judgment mechanism.

7. The method for information-based simulation of groundwater flow field according to claim 5, characterized in that, A set of chaotic parameter adjustment schemes is obtained by executing the chaotic perturbation injection rules in the execution rules. Based on the set of chaotic parameter adjustment schemes and the set of corrected parameter adjustment schemes, a final scheme set is obtained, including: Whether to trigger chaotic perturbation injection is selected based on chaotic triggering conditions; If triggered, the perturbation probability is obtained, and an information element object is randomly selected based on the perturbation probability. A chaotic sequence is generated using Logistic chaotic mapping. The perturbation term is obtained through the chaotic sequence. A set of chaotic parameter adjustment schemes is obtained based on the perturbation term and the set of correction parameter adjustment schemes. The perturbation term includes the direction sign and the perturbation amplitude. The final set of solutions is obtained by combining the set of chaotic parameter adjustment schemes for the selected information element objects with the set of correction parameter adjustment schemes for the unselected information element objects.

8. The method for information-based simulation of groundwater flow field according to claim 1, characterized in that, The method further includes: The convergence conditions include condition one, condition two, and condition three; Condition one is that the current global objective function value has reached the set threshold or the improvement rate has stagnated; Among them, condition two is that the consensus of the entire network is higher than a set threshold and the variance of the consensus sequence of the entire network is less than a set variance threshold. Condition 3 is reaching the maximum number of iterations.

9. An information-based simulation system for groundwater flow field, applied to the method described in any one of claims 1 to 8, characterized in that, include: A basic unit, which is used to acquire simulation basic data and generate a set of geological hypotheses based on the simulation basic data; The building unit is used to build and initialize the initial network topology framework, and at the same time, the building unit is used to extract the initial cooperative sub-network and define the execution rules; Execution unit, which is used to perform simulation through an initial network topology framework and generate simulation results; A cyclic unit is used to perform cyclic deduction and obtain the final simulation result based on the convergence condition.

Images