Geological map scene implementation method and system based on measure potential tapping

Abstract

The invention discloses a geological map scene implementation method and system based on measure potential tapping, relates to the technical field of oil and gas field development, and is used for solving the problems that conflicts among a caliber layer, a method layer and a data layer are accurately positioned based on the same reference system, and a minimum recalculation set is difficult to stably solve according to the conflicts. Establishing a strong primary key index of a well ID, a layer ID, a grid unit and a time step, and registering and generating a caliber; a bidirectional Hausdorff distance is minimized through a multi-resolution pyramid and elastic registration, isolines are extracted and vectorized into a boundary set, scale factors and registration evidences are recorded, and co-located grids are generated; calculating four components including space overlapping, boundary consistency, trend correlation and abnormal co-location of CI in co-location grids, and performing structured output; and forming a conflict graph, locally replaying and generating a factor importance sequence by using a single factor, and solving a minimum recalculation set.

Description

Technical Field

[0001] This invention relates to the field of oil and gas field development technology, and more specifically, to a method and system for realizing geological map scenarios based on measures to tap potential. Background Technology

[0002] During oil and gas field development, reservoir engineers need to continuously monitor and evaluate the dynamic changes in the reservoir to optimize injection and production strategies and improve recovery rates. This typically involves the comprehensive analysis of various dynamic maps, such as water cut maps, pressure maps, bubble point maps, and profile maps. These maps originate from different model simulations and field monitoring data, and may be generated by different software and algorithms, resulting in differences in data caliber, spatial resolution, time step, and interpolation methods.

[0003] The existing technology has the following shortcomings:

[0004] Existing multi-source maps lack subpixel-level co-location registration and boundary assimilation. Boundary set alignment errors are difficult to record as registration evidence such as scale factors and displacement fields in the co-location raster and participate in the construction of CI four components. This makes it difficult to accurately locate the conflict between the aperture layer, method layer and data layer based on the same reference frame, and it is difficult to stably solve the minimum recalculation set based on this.

[0005] To address the above problems, this invention proposes a solution. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a method and system for implementing geological map scenarios based on measures to tap potential, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method and system for implementing geological map scenarios based on measures to tap potential includes the following steps;

[0009] Acquire multi-source maps and numerical sequences from both the model and field sides, establish strong primary key indexes for well ID, layer ID, grid cell, and time step under a unified reference system, and complete the specification of unit system and PVT caliber and generate method metadata field set record items.

[0010] The calculation projects the map onto a multi-resolution pyramid and performs co-location registration, extracts contour lines and vectorizes them into a boundary set, minimizes the symmetric Hausdorff distance with elastic transformation, records the scale factor, displacement field and registration evidence and establishes a co-location raster.

[0011] The four components of spatial overlap, boundary consistency, trend correlation, and anomalous co-location in the local neighborhood of the candidate well layer and the co-located grid are calculated and synthesized into a consistency index CI according to the grid sampling and concatenation rules. The component values ​​and identification field sets are recorded and associated with a unified number and time step label.

[0012] Determine the relationship between the consistency index CI and the threshold. If it is lower than the threshold, generate a conflict map of the caliber layer, method layer and data layer, construct a candidate recalculation operation set, and form a factor importance sequence by single-factor local replay and difference and solve the minimum recalculation set.

[0013] The incremental recalculation and block redrawing results are generated under the constraints of the minimum recalculation set. Differential tiles are replaced by tileization and version number management. The standardized map tiles, CI decomposition and potential a posteriori are synchronously sent to the external working panel through the parsing and write-back channel.

[0014] In a preferred embodiment, the data admission and map registration establishes a strong primary key index consisting of well ID, layer ID, grid cell, and time step, placing multi-source maps and numerical sequences from the model side and the field side into a unified grid and a unified reference system, registering the unit system and PVT caliber specifications, generating an index mapping table and a reverse index table, executing key conflict detection and conflict replacement strategies, saving index verification summaries and serialization snapshots, and setting the index version number.

[0015] In a preferred embodiment, a caliber standard dictionary, a field alias table, and a measurement conversion table are established. The multi-source maps and numerical sequences are corrected item by item using the caliber standard dictionary. The corrected caliber, conversion path, and batch number are recorded in the method metadata field set, and the storage format and version number identifier are set. A source data registration table is established to record the data source identifier, time step range, file path, and verification summary, and a reference relationship is established with the strong primary key index.

[0016] In a preferred embodiment, the multi-resolution pyramid extracts contour lines layer by layer and vectorizes them into a boundary set. Morphological thinning and curvature regularization are performed, and the scale factor and displacement field are obtained by minimizing the symmetric Hausdorff distance using elastic transformation. The co-location raster and registration evidence are output, and the level number, node sequence number, sampling interval, and boundary length threshold are recorded. A registration diagnosis log and parameter snapshot are generated and associated with the method metadata field set.

[0017] In a preferred embodiment, a correspondence between boundary sets and grid cells is established on the co-located grid, node coordinates, orientation and topology labels are stored, and operation logs, timestamps and operator identifiers for topology connection, cutting and merging are recorded. A verification summary is generated and archived, and the topology change batch and audit trail number, as well as the associated boundary set version number and co-located grid version number, are registered in the method metadata field set.

[0018] In a preferred embodiment, the local neighborhood of the candidate well layer is limited by the well ID and layer ID, and the neighborhood radius and maximum neighborhood depth are set. The grid sampling step size and sampling start point are set on the co-location grid. The calculation order, window size and cache structure of the four components of spatial overlap, boundary consistency, trend correlation and abnormal co-location are established. The parameter source, default value and coverage relationship are recorded in the method metadata field set, and the persistent key name and calculation task number are set.

[0019] In a preferred embodiment, the co-position grid is traversed at a fixed step size to obtain the values ​​of the four components: spatial overlap, boundary consistency, trend correlation, and abnormal co-position. The values ​​are then combined into a consistency index (CI) cell by cell according to the conjunction rule. The sampling position, the values ​​of the four components, and the CI number are recorded. The values ​​and numbers are written into an identifier field set and associated with the well ID, layer ID, and time step label. A snapshot and data verification summary are also saved.

[0020] In a preferred embodiment, the threshold is configured in the method metadata field set and has a threshold number. When determining the relationship between the consistency index CI and the threshold, a conflict map of the caliber layer, method layer and data layer is generated, and a conflict list table and a conflict grouping table are established. A sorting key and a priority field are set, and a conflict index is generated for reference by the candidate recalculation operation set. The conflict cell number and the well layer object to which it belongs are recorded.

[0021] In a preferred embodiment, a candidate recalculation operation set is established based on the conflict map. The single factor is replayed locally in the local neighborhood of the candidate well layer and the difference is taken to form a factor importance sequence. The factor importance sequence is associated with the well ID and layer ID and stored. A persistent identifier and timestamp are generated. A foreign key association and retrieval key are established between the factor importance sequence and the well layer object, and the execution order and difference range are registered.

[0022] A geological map scene implementation system based on potential tapping measures includes:

[0023] The memory and processor, wherein the memory stores instructions, and the processor, when executing the instructions, implements the steps of the method of claim 1; the data structure of the instructions includes at least a strong primary key index, a co-location grid, a boundary set, a conflict map, a minimum recalculation set, and a tiled and version number management field, and associates the well ID, layer ID, grid cell, and time step with the strong primary key index and maintains consistency with the unified reference system.

[0024] The technical effects and advantages of the geological map scene implementation method and system based on the tapping of potential measures of this invention are as follows:

[0025] This invention generates a caliber by establishing and registering strong primary key indexes of well ID, layer ID, grid cell, and time step; minimizing the bidirectional Hausdorff distance using multi-resolution pyramids and elastic registration, extracting contour lines and vectorizing them into boundary sets, recording scale factors and registration evidence, and generating co-location rasters; calculating the spatial overlap, boundary consistency, trend correlation, and anomalous co-location four components of CI on the co-location rasters and outputting them in a structured manner; forming a conflict map based on this, generating a factor importance sequence using single-factor local replay, and solving for the minimum recalculation set; performing incremental recalculation and block redrawing using tile-based and version number management, and synchronizing standardized map tiles, CI decomposition, and potential posterior to EPBP through parsing and write-back channels. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the process for implementing a geological map scenario based on the tapping of potential measures according to the present invention;

[0027] Figure 2 This is a schematic diagram of the geological map scene implementation system based on the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0029] This invention provides a method and system for implementing geological map scenarios based on potential tapping measures, aiming to solve problems such as inconsistency in multi-source map data, inaccurate registration, lack of quantitative consistency assessment, low efficiency in conflict resolution, lack of confidence in potential assessment, and insufficient human-computer interaction in reservoir dynamic analysis. This method, through a series of structured steps, achieves unified management of multi-source dynamic maps, high-precision registration, quantitative consistency assessment, intelligent conflict location and incremental recalculation, confidence-based Bayesian fusion potential assessment, and a human-computer collaborative ranking and feedback mechanism. Finally, it outputs an auditable potential base map to the EPBP working panel.

[0030] Example 1: Refer to Figure 1 The method flow of the present invention includes the following steps:

[0031] Step S101, Data Admission and Map Registration. This step is fundamental to the entire method and aims to address the heterogeneity of multi-source data. The system first receives multi-source data and maps from both the model side and the field side. Model-side data includes RSM data output from the reservoir numerical simulator, while field-side data includes monitoring data such as downhole pressure, flow rate, and water cut. To ensure that all data is managed and analyzed within a unified reference frame, the system establishes a strong primary key index consisting of well ID, layer ID, grid cells, and time step. This strong primary key index guarantees a one-to-one correspondence between maps and indicators within the same reference frame and forms the basis for all subsequent analyses.

[0032] Specifically, various data output from RSM and other models are fed into a unified grid system, such as pressure, saturation, and water cut, and bound to the corresponding well entities. For profile-type results, such as attribute profiles along the wellbore, the system attaches their corresponding well trajectory sequences to accurately locate and trace back in three-dimensional space.

[0033] To eliminate discrepancies in data caliber, the system normalizes the unit system, PVT caliber, and time step of all data. In one optional example, the unit system might be imperial versus metric, the PVT caliber might be volume under standard conditions versus volume under geological conditions, and the time step might be day, month, or year. During normalization, the system meticulously records the caliber conversion hierarchy, allowing users to trace the source and conversion path of any field at any time, ensuring transparency and auditability in data processing. After normalization, a data dashboard with unified caliber is generated.

[0034] Furthermore, for various maps such as water content contour maps, Meccan index maps, pressure contour maps, bubble point maps, and profile maps, the system registers metadata about the generation algorithm, interpolation kernel, filter half-width, smoothing order, sampling scale, and profile sampling density. In one optional example, the generation algorithm might be Kriging interpolation or inverse distance weighting; the interpolation kernel might be a Gaussian kernel or an exponential kernel. This metadata is crucial for locating conflict sources and determining the recalculation range in subsequent steps. Finally, this step generates a standardized data cube and metadata, providing a unified and standardized input for subsequent steps.

[0035] Furthermore, the details of parameter acquisition and default range refinement in step S101 of this application are as follows:

[0036] The selection of the local neighborhood Ω is typically based on reservoir geological characteristics and grid resolution. Its default range can be set to 3-5 times the average grid cell size, or adjusted according to empirical values ​​such as well spacing and fault spacing. For example, for conventional reservoirs, it can be set to 200-500 meters. Users can adjust this range on the system interface using sliders or input boxes and preview its impact on CI calculations in real time.

[0037] The MAD threshold used for extracting outlier set E is typically set to 2.5-3.5 by default. This value indicates the degree to which data points deviate from the median; values ​​exceeding this threshold are considered outliers. The specific value can be determined offline based on historical data statistical distribution or set by expert experience.

[0038] The boundary error tolerance scale τ can be set to 1-2 times the average grid cell size by default, or dynamically adjusted according to the average curvature radius of the reservoir boundary. For example, for a 100-meter grid, τ can be set to 100-200 meters. This value reflects the spatial scale at which the boundaries are considered similar.

[0039] The CI threshold θ can be set to a default value of 0.6-0.75. Values ​​below this threshold are considered to indicate significant conflict. This value can be set based on statistical analysis of the CI distribution of conflict cases in historical data, or by expert experience, and user-defined values ​​are also supported.

[0040] When CI uses the product t-norm, the weight w q The initial default values ​​can be set to a uniform distribution, such as 0.25. These weights are dynamically updated in step S106 through online learning and human feedback. The initial calibration can be obtained through expert scoring or regression analysis of historical cases.

[0041] Step S102, Multi-scale Co-location Registration and Boundary Assimilation. This step aims to solve the spatial registration problem between different maps, especially sub-pixel level deviations and precise boundary alignment. The system first projects various maps onto a multi-resolution pyramid, which facilitates registration at different scales and achieves sub-pixel level co-location, thereby improving registration accuracy.

[0042] For profile maps, since they are essentially two-dimensional slices along the well trajectory, in order to make them comparable to three-dimensional isosurface maps, the system backmaps the profile maps to voxel slices based on the well trajectory and stratigraphic control points, thereby obtaining profile projections comparable to isosurfaces.

[0043] Sampling raster / interpolation kernel settings for profile backprojection: During profile backprojection, the system samples the well trajectory at equal or adaptive intervals according to the resolution of the target 3D mesh, generating a series of sampling points. The value of each sampling point is obtained from the original profile data using methods such as linear interpolation, cubic spline interpolation, or kriging interpolation. These sampling points are then mapped to the corresponding 3D voxel elements, forming voxel slices. The choice of interpolation kernel affects the smoothness and accuracy of the backprojection and is typically configured based on the characteristics of the original profile data and the target accuracy requirements.

[0044] Next, the system extracts contour lines from the maps and vectorizes them into boundary sets. To suppress pseudo-aliasing caused by interpolation or sampling, the system employs morphological thinning and curvature regularization techniques to process the boundaries. Subsequently, the system achieves flexible registration between boundary sets of different maps, aiming to minimize the bidirectional Hausdorff distance. During the registration process, the system records scale factors and boundary deviations as registration evidence, which is crucial for subsequent evaluation of registration quality and location conflicts.

[0045] Furthermore, the details of the flexible registration in this step are as follows:

[0046] Cost function: The goal of elastic registration is to minimize a comprehensive cost function E(T), where T is a spatial transformation function, such as an affine transformation or a thin-plate spline transformation. This cost function typically includes the following terms:

[0047] E(T)=α·dH(B k ,T(B j ))+β·R(T)+γ·A(T);

[0048] in:

[0049] dH(B k ,T(B j )) is the boundary set B of the two types of maps after registration. k and T(B) j The symmetric Hausdorff distance between the two boundaries is used as a data term to measure the degree of geometric mismatch at the boundary. R(T) is a smoothing regularization term for the transformation T, such as bending energy or gradient norm, used to penalize excessive distortion transformations and ensure the physical rationality of registration. A(T) is an area / length constraint term used to maintain the relative invariance of the area or boundary length of the region enclosed by the boundary before and after registration, preventing unreasonable geological deformation caused by registration. α, β, and γ are the weighting coefficients of each term, which are adjusted according to actual needs and experience.

[0050] An example: In reservoir map registration, the weight of α is relatively high because more emphasis is usually placed on accurate boundary matching, for example, α∈[0.5,0.8]. The smoothing regularization term β is usually given a moderate weight to avoid unnatural distortions in the registration results, for example, β∈[0.1,0.3]. The area / length constraint γ is used to maintain overall geological characteristics, for example, γ∈[0.1,0.2]. These weights can be calibrated according to the actual situation through cross-validation or expert experience.

[0051] Minimizing E(T) usually employs iterative optimization algorithms. The specific choice depends on the map size and accuracy requirements. For example, gradient descent is chosen for large-scale reservoir maps, while Newton's method is chosen for small-scale fine registration scenarios.

[0052] The initial transformation T0 can be obtained through a simple rigid body transformation. An optional example: extract key boundary features from two sets of maps, such as the top interface of the oil layer and fault lines, calculate their geometric centers and perform translation and alignment, determine the rotation angle, and remove isolated interference points on the boundary, such as mislabeled patches and annotation lines, during preprocessing to ensure the accuracy of angle calculation; finally, determine the scaling factor in conjunction with the map scale.

[0053] Furthermore, in each iteration, the gradient of the cost function is calculated, and the transformation parameters are updated along the gradient direction.

[0054] The iterative process continues until one of the following conditions is met:

[0055] 1. The change in the cost function E(T) is less than a preset threshold, for example, the relative change is less than 0.001.

[0056] 2. The change in the transformation parameters is less than the preset threshold.

[0057] 3. Reach the maximum number of iterations.

[0058] By combining a multi-resolution pyramid, registration is first performed at a coarse scale to obtain an approximate transformation. This transformation is then used as the initial value for fine-scale registration, gradually improving accuracy. An optional example is to construct a three-layer pyramid. First, registration is performed at the lowest resolution and coarsest scale to obtain a global affine transformation. Then, this transformation is upsampled and applied to the medium-resolution image, upon which more refined non-rigid registration is performed. Finally, the medium-resolution transformation is upsampled and applied to the highest resolution image for final sub-pixel-level registration.

[0059] To facilitate subsequent quantitative use, the system calculates the boundary similarity index B = exp(-d) at the end of the registration process. H / τ). Where d H The symmetric Hausdorff distance for the boundary sets of two aligned map classes quantifies the maximum mismatch between the two boundary sets. τ is the boundary error tolerance scale, whose value is set according to the grid resolution or the average bending radius of the boundary, and is used to normalize the Hausdorff distance to a similarity value between 0 and 1.

[0060] The dynamic adjustment of τ can be based on the local region's mesh resolution and the geometric complexity of the boundary. An optional example: the system can first calculate the average mesh cell size G of the local region. avg Then, by performing curvature analysis on the boundary curve, the average radius of curvature R of the local boundary is calculated. curv τ can be dynamically set as τ = max(k1·G) avg ,k2·R curv), where k1 and k2 are empirical coefficients, for example, k1∈[1.0,2.0], k2∈[0.5,1.0]. Thus, in regions with coarser meshes or smoother boundaries, τ will be larger, allowing for greater errors; in regions with finer meshes or complex boundaries, τ will be smaller, requiring higher matching accuracy.

[0061] The similarity B, along with the scale factor and displacement field, is output as a component of the co-location grid and alignment boundary, and is used in steps S103 and S104.

[0062] Step S103, Structured Assessment of the Local Consistency Index (CI). This step is used to quantitatively assess the consistency of different maps in local areas. The system calculates four complementary components—spatial overlap, boundary consistency, trend correlation, and anomaly isomorphism—within the local neighborhood Ω of candidate wells and layers, and forms a consistency score based on these components.

[0063] The system first identifies potential hotspot sets and anomaly sets in the local neighborhood Ω, which form the basis for assessing consistency.

[0064] Hotspot region extraction can be achieved through various methods. One possible approach is threshold-based segmentation. For instance, in a water-cut map, regions with water cut exceeding a certain threshold, such as 50%, are defined as high-water-cut hotspots; in a pressure map, regions with pressure below a certain threshold, such as bubble point pressure, are defined as low-pressure hotspots. The threshold can be set by expert experience based on reservoir type and production stage, or automatically determined through statistical analysis, such as the Otsu method based on histograms. Another method is based on cluster analysis, clustering grid cells with similar attribute characteristics into hotspot regions.

[0065] Then, the following four pieces of evidence were obtained:

[0066] First, spatial overlap component (O): characterized by penalized IoU (Intersection over Union), specifically... Among them, A k With A j Let s be the set of hotspot regions of the two types of maps within Ω. k With s j This is the scale factor obtained in step S102. This component not only considers the degree of overlap in hotspot regions, but also penalizes spurious overlaps caused by scale mismatch through an exponential term, ensuring the authenticity of spatial overlap.

[0067] Second, the boundary consistency component (B): directly adopt the boundary similarity B = exp(-d) output from step S102. H This directly utilizes the results of geometric registration, reflecting the degree of fit between map boundaries.

[0068] Third, the trend correlation component (T): Spearman's rank correlation coefficient is used to enhance robustness to dimensional differences and nonlinear responses, specifically T = Spearman(M k (Ω),M j (Ω)). Among them, M k (·) represents scalar field sampling on a uniform grid. Rank correlation can effectively assess whether the variation trends of two maps are consistent in a local region, regardless of their absolute values ​​or dimensions.

[0069] Fourth, anomalous isotopes (C): measured by isotope ratio, specifically... Here, E represents the set of anomalies extracted based on the MAD threshold. This component assesses the degree of overlap of anomalous regions in different maps and is crucial for identifying potential anomalies or data errors.

[0070] In a more refined implementation, the calculation of the anomalous isotope component C can be based on Ripley's K function to evaluate the spatial clustering of point patterns, i.e., the spatial clustering of the anomalous point set.

[0071] For a point pattern, the K-function K(t) is defined as the average number of points j contained within a distance t from any randomly selected point i, divided by the point pattern density λ. That is...

[0072] For two sets of outliers E k and E j Their cross K-functions can be calculated. kj (t), which measures E k Point in E j The spatial correlation of points at different distances t in the equation. An anomalous co-located component C can be defined as... Where L(t) is the K-function under completely random space, i.e., L(t) = πt 2 The integration range [t1,t2] is the scale range of interest.

[0073] The specific details of the K(t) value range, boundary correction, and threshold strategy are as follows:

[0074] The value of K(t) is usually related to πt. 2 Compare them. If K(t) > πt 2 This indicates that points are clustered within a distance t; if K(t) < πt 2 This indicates that the points are distributed in a dispersed manner.

[0075] In addition, boundary correction is required when calculating the K-function to avoid underestimating the point density due to boundary effects in the study area. Commonly used boundary correction methods include Ripley's isotropic correction, translation correction, or edge correction.

[0076] The final value of the anomalous isotope component C can be obtained by adjusting K. kj The statistical test for the difference between L(t) and L(t) is used to determine this. For example, it can be calculated... The function is defined, and a threshold is set. When L(t) exceeds the threshold, the outlier is considered to have significant isotopic inclination at scale t. The final C value can be a comprehensive measure of isotopic intensity at multiple scales, ensuring stability at different densities.

[0077] The choice of the integration range [t1, t2] for the Ripley-K function is also determined based on the specific circumstances. One possible example is that t1 can be set to the smallest grid cell size, such as 10 meters, to capture nearest-neighbor co-locations; t2 can be set to the radius of the local neighborhood Ω, such as 300 meters, to capture a larger range of co-locations. Alternatively, the integration range can be dynamically determined by analyzing characteristic points of the K(t) curve, such as peaks and inflection points, to reflect aggregation or dispersion patterns at different scales.

[0078] The system employs a minimization conjunction (min operation) to form the consistency index CI = min{O,B,T,C}, to avoid average dilution conflicts and ensure that poor performance of any component directly lowers the overall consistency index. In an alternative implementation, the system uses the product t-norm. Where the weight w q In step S106, the system is updated by online learning and human feedback, which allows the system to adjust the importance of each component based on the actual application scenario and expert experience.

[0079] w q The online learning mechanism can be based on human feedback from engineers.

[0080] An optional example: When an engineer marks a well layer as a reliable or unreliable sample in step S106, the system records the CI value of that sample and its four components O, B, T, and C. For reliable samples, the system increases the weights of those components that lead to high CI values; for unreliable samples, the system decreases the weights of those components that lead to high CI values ​​but are actually inconsistent. This can be implemented using methods such as gradient descent or reinforcement learning, for example, by defining a loss function. CI i It is the CI value calculated by the system, Label iIt's a manual judgment by the engineer, such as 0 or 1, and then... Update w q , where η is the learning rate.

[0081] The system outputs the CI and four-component details for each well layer as inputs for steps S104 and S105.

[0082] Step S104: Conflict Map and Minimum Recalculation Set Location. When the consistency index CI calculated in step S103 is lower than the preset threshold θ, it indicates that there are significant conflicts between the maps. At this time, the system will generate a conflict map and label the sources of differences item by item at the caliber layer, method layer, and data layer.

[0083] Differences at the aperture level may include inconsistencies in unit systems, mismatches in PVT apertures, or inconsistent time windows. Differences at the method level may involve differences in generation algorithm parameters such as interpolation kernels, filter half-width, and smoothing order. Differences at the data level may include RSM time steps, profile sampling density, and well-layer mapping rules.

[0084] To identify the key factors causing the conflict, the system performs local replay and differencing on a single factor. This means that the system re-executes the relevant calculations in a local region for a specific difference factor, such as adjusting the time window of the stress map, and observes the changes in CI. ΔCI represents the improvement contribution of this factor, i.e., the degree of improvement in CI after the factor adjustment.

[0085] It should be noted that ΔCI is calculated by comparing the changes in CI values ​​before and after recalculation. An optional example: for a conflict region r and a candidate recalculation operation c, the system first calculates the original CI value CI. orig (r). Then, the simulation executes operation c, which modifies the relevant parameters or data within a local region and recalculates the CI value. new (r). Then ΔCI(c,r)=CI new (r)-CI orig (r). To avoid the influence of noise, the CI value can be smoothed or averaged after multiple simulations.

[0086] Based on these ΔCI values, the system forms a sequence of factor importance.

[0087] Subsequently, the system abstracts the candidate recalculation operation into a set covering problem, and solves for the minimum recalculation set based on the improved threshold and cost function.

[0088] Furthermore, the optimization details of the minimum recalculation set MRS are as follows:

[0089] Minimize the total recalculation cost while ensuring that the CI improvement reaches the preset target. Formally, this can be summarized as:

[0090] MinimizeΣ c∈C cost(c)·x c ;

[0091] Subjectto:

[0092] For each conflicting regionr;

[0093]

[0094] Where c represents the candidate recalculation operation, C r It is the set of recalculation operations that can affect region r, x c These are binary variables, where 1 represents choosing operation c and 0 represents not choosing it. `cost(c)` is the computational cost of operation c, and `ΔCI(c,r)` is the contribution of operation c to the CI improvement of region `r`. target (r) is the CI target improvement value for region r.

[0095] In addition to CI improvement goals, resource constraints and dependency constraints can also be included. Optional examples of resource constraints include maximum computation time and the number of available CPU / GPU cores; dependency constraints include certain operations that can only be executed after other operations have been completed.

[0096] Considering the actual reservoir size and computational efficiency, approximate algorithms are usually adopted:

[0097] The algorithm selects the operation that maximizes the improvement in CI per unit cost each time, until the CI in all conflict regions reaches the target. This algorithm is simple and efficient, but it may not find the globally optimal solution.

[0098] By combining optimization methods such as genetic algorithms and simulated annealing, a near-optimal solution can be searched within an acceptable timeframe.

[0099] For small-scale problems, an ILP model can be constructed and commercial solvers such as Gurobi and CPLEX can be used to obtain an exact optimal solution. For large-scale problems, techniques such as Lagrange relaxation or column generation can be used for decomposition.

[0100] Greedy algorithms typically have low complexity, approaching linear or logarithmic linearity, and converge quickly. ILP solvers exhibit exponential complexity growth with increasing problem size, but may perform well for problems with specific structures. The convergence of heuristic algorithms depends on the specific implementation and parameter settings, and is usually determined by the number of iterations or a threshold value for the objective function.

[0101] This minimum recalculation set is used to trigger incremental recalculation and block redrawing in step S107, thereby resolving the conflict with minimal computational cost.

[0102] Step S105, Bayesian fusion at the evidence level. This step aims to fuse evidence from multiple sources to obtain a reservoir potential assessment result with confidence. The system defines potential as a continuous potential field and applies a geologically compatible GMRF (Gaussian Markov Random Field) prior to it to express connectivity and anisotropy constraints, thus making the fused result more consistent with geological laws.

[0103] Maps of water aquifer, pressure, bubble point, and profile are used to construct specific observation factors, including monotonic constraints, segmental linearity, thresholding, and linear observation chains. These observation factors transform information from different types of maps into observational evidence of the potential potential field. The system is then jointly solved using maximum a posteriori (MAP) to obtain the posterior mean and confidence interval for each well layer.

[0104] To achieve unified evidence-oriented fusion, the system solves... Where Z represents the potential field, and logp(Z) is a GMRF prior term that encodes geological connectivity and anisotropy information. (k) For the observation of the k-th type of map, φ k The noise and response parameters corresponding to this map reflect the uncertainty of the observation and the sensitivity to the potential field.

[0105] For areas with low local consistency, i.e., areas with low CI scores in step S103, the system couples the low components of step S103 into the likelihood weights using a soft penalty to suppress the conflict amplification effect. This means that during the fusion process, the weight of observational evidence for maps with poor consistency will be appropriately reduced, thereby minimizing their negative impact on the final potential assessment results. To meet engineering timeliness requirements, the system employs approximation methods such as iterative conditional models, variational inference, or local INLA (Integrated Nested Laplace Approximation) for solution. The system ultimately outputs the posterior mean μ and variance σ for each well layer. 2 The confidence interval [L,U] is given to quantify the uncertainty of the potential assessment results.

[0106] Furthermore, the specific details of this step, which involves observing the family of factors, are as follows:

[0107] Observation factors p(Y) of each map (k) |Z;φ k The response function family and parameter calibration process are as follows:

[0108] The family of response functions includes:

[0109] Hydrographic map (Y) (含水)The response function typically exhibits a monotonically decreasing relationship with potential. It can be a sigmoid function or an exponentially decaying function, for example... Where f(Z) is a monotonic mapping function from potential Z to water content, such as linear or log-linear, φ 含水 ={f,σ 含水 Pressure diagram (Y) (压力) The response function typically exhibits a monotonically increasing relationship with potential. It can be a sigmoid function or a linear function, for example... Where g(Z) is a monotonic mapping function from potential Z to pressure, φ 压力 ={g,σ 压力 Bubble dot pattern (Y) (泡点) The relationship between bubble point pressure and formation pressure often involves a threshold effect. When the pressure is below the bubble point, a gas phase may occur, affecting potential. The response function can be a piecewise function or a function with a threshold parameter, such as p(Y). (泡点) |Z;φ 泡点 This can be modeled as when the actual pressure P > P0. b When the pressure is greater than the bubble point pressure, the impact on potential is small; when P ≤ P b At that time, it has a significant negative impact on potential. φ 泡点 ={P b ,σ 泡点}. Sectional view (Y (剖面) Profile data typically reflects property variations along the wellbore and can be viewed as a linear observation chain of local potential. The response function can be Gaussian likelihood, where the observed values ​​Y... (剖面) The projection of the potential field Z onto the wellbore path is linearly related, i.e., p(Y) (剖面) |Z;φ 剖面 )∝exp(-(Y (剖面) -HZ) T Σ -1 (Y (剖面) -HZ)), where H is the projection matrix, Σ is the observation noise covariance, and φ 剖面 ={H,Σ}.

[0110] The parameter calibration process is as follows:

[0111] Offline calibration: Combining the experience of reservoir engineers, qualitative judgments are made on the physical relationship between different maps and potential, such as monotonicity and threshold values. A large amount of historical well data is collected, including observations for each map and validated potential assessment results. Using historical data, regression analysis, such as linear regression and nonlinear regression, or machine learning models, such as support vector machines and neural networks, are used to fit the mapping relationship between observations and potential, and to estimate the noise parameter σ. kOr covariance Σ. For example, for a monotonic relationship, the specific form of f(Z) or g(Z) can be fitted. For a threshold relationship, the bubble point pressure P can be determined. b Empirical values ​​or statistical distributions of the calibration parameters. Use cross-validation techniques to evaluate the generalization ability of the calibration parameters.

[0112] Online Update: In step S106, when engineers are dissatisfied with the potential assessment results or rankings provided by the system, they can manually adjust the potential value of a well layer or mark it as a reliable sample. These manual feedbacks are then used as new observational evidence to update the observation factor parameter φ using a Bayesian inference framework. k The system calculates the posterior distribution of the water-cut well. For example, if an engineer marks a high-water-cut well as having high potential, the system adjusts the response function f(Z) of the water-cut map or its noise parameters to better tolerate the coexistence of high water cut and high potential. With the continuous accumulation of new data and human feedback, the system can continuously optimize parameters, enabling the model to better adapt to the complexity and dynamic changes of actual oil reservoirs.

[0113] Boundary case handling, such as the exception of high water content but high potential: For monotonic mapping functions like f(Z) or g(Z), boundary cases are specifically considered during calibration. For example, the exception of high water content but high potential might mean that the area has special geological conditions, such as high permeability channels, or injection-production measures, such as infiltrated water injection, resulting in high water content but still potential. In this case, the system introduces conditional dependencies or hybrid models, meaning that f(Z) is no longer a simple monotonic function but depends on additional geological or engineering features. Through online learning and human feedback, the system can identify these exception patterns and adjust φ accordingly. k The parameters allow the model to more accurately reflect these complex relationships.

[0114] INLA / ICM Selection and Convergence Diagnosis: For large-scale problems, INLA is typically used to approximate the posterior distribution due to its high computational efficiency and ability to provide good approximations of the posterior mean and variance. ICM is a simpler iterative optimization method with fast convergence, but it may converge to a local optimum. The system dynamically selects the appropriate approximation algorithm based on the problem's size, complexity, and real-time requirements.

[0115] Convergence Diagnosis: For Bayesian models, DIC (Deviance Information Criterion) and WAIC (Watanabe-Akaike Information Criterion) are commonly used metrics for model selection and convergence diagnosis. They measure the goodness of fit and complexity of the model; lower DIC / WAIC values ​​generally indicate a better model. Monitoring the posterior mean μ and variance σ is also important. 2The changes during the iteration process. When these values ​​change by less than a preset threshold in consecutive iterations, for example, a relative change of less than 0.001, the model is considered to have converged. In addition to the above statistical indicators, a maximum number of iterations can also be set as a stopping criterion to prevent the algorithm from getting stuck in an infinite loop.

[0116] Step S106, Candidate Well Ranking and Human-Machine Collaboration. This step aims to rank potential well layers and introduce a human-machine interaction mechanism to optimize the evaluation results. The system ranks the potential well layers based on the posterior mean μ obtained in step S105, initially determining their priority. To avoid samples with excessive uncertainty affecting the reliability of the ranking, the system uses a confidence penalty to backtrack the ranking of candidates with excessive uncertainty.

[0117] Specifically, the system employs a robust ranking score S = μ - λ·max{0, σ - σ0}, where σ0 is the uncertainty threshold and λ is the penalty coefficient. When the standard deviation σ of a well exceeds the threshold σ0, its posterior mean μ is penalized, thus lowering its ranking position. The values ​​of σ0 and λ can be set according to the project's risk preference and historical hit rate to balance potential and uncertainty.

[0118] Simultaneously, the system generates an evidence profile, including marginal evidence for each map element and bar charts of the CI components, to explain why a recommendation was made or why a conservative approach was adopted. This provides engineers with a transparent basis for decision-making and enhances the interpretability of the evaluation results.

[0119] The system allows users to select conflicting areas or label samples requiring recalculation and reliable samples on the map. These manual labels will be used to update the weights w in step S103 online. q This is used for calculating the consistency index CI, and is related to the observation factor parameter φ in step S105. k This is used for Bayesian fusion. This human-machine collaborative governance mechanism allows the system to continuously learn and optimize from the engineer's experience, improving the accuracy and adaptability of the evaluation.

[0120] Step S107, Incremental Recalculation and Block Redrawing. This step is the actual operation for resolving conflicts and updating the map. Based on the minimum recalculation set obtained in step S104, the system only performs recalculation on necessary parts and local areas, instead of performing a full recalculation. This greatly saves computing resources and time.

[0121] The rendering engine employs tile-based and version management technologies. Tile-based rendering breaks down the entire map into small, independent tiles. When a local area changes, only the affected differential tiles need to be re-rendered and replaced, without redrawing the entire map. Version management ensures the traceability of map updates. The system also generates version differential heatmaps to quantify the CI improvements and post-acceptance convergence brought about by updates, visually demonstrating the effects of recalculation. Tile size, differential threshold, and version strategy can be flexibly configured according to the deployment environment and time requirements.

[0122] Furthermore, the specific details of the adaptive tile granularity in this step are as follows:

[0123] The adaptive tile granularity in this implementation is based on the variation amplitude of the Consistency Index (CI) on the co-located grid and the conflict density of the conflict map. First, the spatial fluctuations of CI are assessed locally, and then the conflict distribution in the corresponding area is statistically analyzed. When CI changes are significant or conflict distribution is concentrated within a region, a finer tile granularity is preferred; when CI changes are gradual and conflicts are dispersed, a coarser tile granularity is used; when the two determinations differ, the finer granularity takes precedence, and a smooth transition is set for adjacent areas to avoid discontinuous mosaicking.

[0124] Spatial organization uses quadtrees or octrees to divide the study area level by level. The parent node stores the sum of CI and conflict. When a node does not meet the stability criterion, it continues to be subdivided. When the stability criterion is met, the subdivision stops and leaf node tiles are generated. The partition boundary is aligned with the co-located grid and boundary set to maintain the consistency between the block and the data index.

[0125] The rendering side maintains a tile cache with layers, scaling levels, rows, columns, and version numbers as keys. Upon receiving a request, it first checks the cache; if a match is found, it returns the result directly; otherwise, it triggers calculation and rendering and writes back to the cache. When differential updates arrive and the version number changes, old entries are marked as expired and cleaned up according to a strategy. Calculation and rendering are executed in parallel using separate task queues, with processing priorities determined by conflict density and CI fluctuations. Mutual exclusion is set for the same tile to avoid duplicate calculations. Incremental recalculation only overwrites affected tiles, and upon completion, the rendering side is notified via message to refresh the cache by tile ID and version number.

[0126] Version number management is integrated throughout the generation and release process. Algorithm or caliber adjustments trigger version jumps, and data updates trigger version increments. The rendering end decides to refresh, reuse, or roll back based on a version consistency strategy, and version mapping relationships and decision records are uniformly written into the method metadata field set. In cases of insufficient local statistics, data gaps, or resource constraints, the generation is temporarily completed with fixed granularity and conservative rules, and the downgraded status is marked. Once the statistics are supplemented, the adaptive strategy is restored, and all judgments, sources, and time tags form a complete tracking record.

[0127] Step S108, outputting the evidence interface with EPBP. This step serves as the interface for integrating this method with external systems. The system synchronizes key information such as standardized map tiles, CI decomposition, confidence intervals, conflict maps, and minimum recalculation recommendations to the EPBP (Enhanced Production Business Process) work panel via parsing and write-back channels. This allows reservoir engineers to view and utilize the evaluation results of this method within a unified interface.

[0128] Simultaneously, the system incorporates the potential posterior, specifically the posterior mean and confidence interval, along with the CI, as preliminary evidence input for subsequent measure selection. This achieves the integration of map analysis and injection-production optimization, enabling reservoir development decisions to be based on more comprehensive evidence.

[0129] Example 2: This application also provides examples of specific application scenarios:

[0130] This embodiment takes the consistency assessment and conflict resolution of well W12-layer L3 as an example to explain in detail the application of this method.

[0131] In reservoir dynamics analysis, engineers need to assess the potential of well W12 in layer L3. The system first completes strong primary key registration between well W12 and layer L3 under a unified grid and time step in step S101, and registers the generation parameters and methods for maps such as water cut, pressure, bubble point, and profile. For example, the water cut map may use Kriging interpolation, the pressure map may use inverse distance weighting, and each may have different filter half-widths and time windows.

[0132] Subsequently, in step S102, the system performs sub-pixel co-location of the water aquifer and pressure maps to ensure their precise spatial alignment. Simultaneously, the profile map is back-mapped to voxel slices based on the well W12 trajectory, enabling spatial comparison with the isosurface map. After boundary assimilation, the system calculates the Hausdorff distance between the boundaries of the water aquifer and pressure maps to be 60m, with a tolerance scale τ = 100m. According to the formula, the corresponding boundary similarity B = exp(-60 / 100) ≈ 0.55. This low boundary similarity already indicates potential conflicts between the maps.

[0133] In step S103, the system identifies the hotspot set of the L3 region of well W12 with a neighborhood radius of 300m, and calculates four consistency components:

[0134] The spatial overlap component O≈0.64 indicates that the hotspot areas have some overlap, but there may be scale differences.

[0135] The boundary consistency component B = 0.55. Directly using the result of step S102 shows poor boundary consistency.

[0136] The trend correlation component T = 0.58 indicates that the correlation between the two charts in the local area is generally low.

[0137] The abnormal isotope component C = 0.60, indicating that the degree of overlap of the abnormal regions is also generally low.

[0138] Ultimately, the system uses the minimum conjunctive equation to calculate the consistency index CI = min{0.64, 0.55, 0.58, 0.60} = 0.55. Since this CI value is lower than the preset alarm threshold of 0.65, the system triggers the conflict handling procedure.

[0139] In step S104, the system generates a conflict map and identifies the sources of discrepancies:

[0140] In the direction of the diameter layer, the time window for the pressure map is [t-7,t] days, while the time window for the water aquifer map is [t-3,t] days, indicating an inconsistency in the time windows.

[0141] In the method layer direction, Kriging interpolation is used on the pressure side with weak smoothing, while other interpolation methods or different degrees of smoothing may be used on the water-bearing side.

[0142] The system performs local replay and difference analysis on the above factors. For example, after unifying the time window of the pressure map to [t-3, t] days and aligning the interpolation kernel and smoothing parameters, it was found that the total ΔCI was improved by about 0.13. Based on this, the system selects the two operations of unifying the time window and aligning the interpolation kernel and smoothing as the minimum recalculation set and triggers step S107.

[0143] In step S107, the system recalculates and redraws only the affected differential tiles around well W12 based on the minimum recalculation set, instead of performing a full recalculation of the entire reservoir. After the recalculation is complete, the system executes step S103 again to perform a consistency assessment, obtaining the updated components:

[0144] Spatial overlap O = 0.70; Boundary consistency B = 0.68; Trend correlation T = 0.66; Anomaly co-location C = 0.63;

[0145] At this point, the consistency index rose to 0.63. Although still below 0.65, it had improved significantly, indicating that the conflict had been effectively mitigated.

[0146] In step S105, the system performs maximum a posteriori fusion using GMRF priors and multiple observation factors, such as water content, pressure, bubble point, and profile, to obtain the potential posterior mean μ = 72 and variance σ of well W12-layer L3. 2 =36, and the confidence interval [60, 84]. This indicates that the well potential estimate is 72, and there is a 95% probability that it falls between 60 and 84.

[0147] In step S106, the system calculates the robust ranking score S = μ - λ·max{0,σ - σ0} = 72 - 0.5·max{0,6 - 5} = 72 - 0.5·1 = 71.5 with σ0 = 5 and λ = 0.5. This score is used to rank the well layers and generate evidence profiles, showing the marginal evidence and CI components of each map, explaining the basis for the potential assessment of the well layer.

[0148] Finally, in step S108, the system writes back the updated tiles, CI decomposition, confidence intervals, conflict maps, and minimum recalculation recommendations to the EPBP working panel for direct use in subsequent optimization. This process, while ensuring caliber and geometric consistency, achieves rapid improvement in local consistency with minimal recalculation cost and provides interpretable ranking results with confidence levels.

[0149] Example 3: This example also discloses the corresponding system architecture, as follows:

[0150] Reference Figure 2 This embodiment of the invention provides a geological map scene implementation system based on measures for potential tapping. The system includes the following core modules: a data access and registration module 100, a registration and assimilation module 110, a consistency assessment module 120, a conflict localization module 130, a Bayesian fusion module 140, a sorting and human-machine collaborative governance module 150, an incremental recalculation and redrawing module 160, and an interface output module 170. These modules work collaboratively to implement the aforementioned method flow. The system also includes a data storage layer 180 and a user interface layer 190, and interacts with an external EPBP work panel 200.

[0151] The data access and registration module 100 is responsible for receiving multi-source data and maps from various sources, such as reservoir simulators, field monitoring equipment, and geological modeling software. Its core function is to establish a unified data management framework, including the construction of strong primary key indexes, such as well ID, layer ID, grid cell, and time step; the standardization of data calibers, such as unit system, PVT caliber, and time step; and the registration of method metadata, such as map generation algorithms, interpolation kernels, filter half-width, smoothing order, sampling scale, and profile sampling density.

[0152] This module is typically deployed on a data server and connects to external data sources via a data acquisition agent or API interface. Internally, it includes a data parser, data converter, and metadata management components. Normalization rules and caliber conversion genealogies are stored in data storage layer 180.

[0153] The registration and assimilation module 110 is responsible for resolving spatial inconsistencies between multi-source maps. This includes projecting various maps onto a multi-resolution pyramid for sub-pixel-level alignment, mapping profiles back to voxel slices based on well trajectories and stratigraphic control points, extracting and vectorizing map boundaries, and achieving precise boundary alignment through an elastic registration algorithm. Simultaneously, it calculates the boundary similarity index B and the recording scale factor and displacement field.

[0154] This module typically includes image processing libraries, geometry processing libraries, and optimization algorithm libraries. The construction of multi-resolution pyramids, sampling and interpolation of profile back mapping, morphological processing, curvature regularization, and iterative optimization of elastic registration are all performed within this module. Registration parameters, such as α, β, and γ, can be obtained from data storage layer 180.

[0155] The consistency assessment module 120 is responsible for quantitatively assessing the consistency of different maps in local areas. Within the local neighborhood Ω of candidate wells and layers, it calculates the spatial overlap component O, the boundary consistency component B, the trend correlation component T, and the anomaly isotope component C. Finally, a consistency index CI is formed based on these components.

[0156] Implementation: This module includes spatial analysis, statistical analysis, and anomaly detection components. Hotspot region extraction, MAD threshold calculation, Spearman rank correlation calculation, and Ripley-K function analysis, including boundary correction and integration range selection, are all implemented in this module. CI calculations, such as min operations or product t-norm calculations, are also performed here.

[0157] The conflict localization module 130 is used to generate a conflict map when the consistency index CI is lower than a preset threshold θ, and to label the sources of difference item by item at the caliber, method, and data layers. By performing local replay and differencing on a single factor, ΔCI is calculated and a factor importance sequence is formed. Finally, the candidate recalculation operation is abstracted into a set coverage problem, and the minimum recalculation set is solved.

[0158] This module comprises a rules engine, a simulation executor, and an optimization solver. The rules engine triggers conflict analysis based on the CI threshold. The simulation executor is responsible for simulating parameter or data modifications in local regions. The optimization solver, such as a greedy algorithm or an ILP solver, determines the minimum recalculation set based on the cost function and the CI improvement objective.

[0159] The Bayesian fusion module 140 is responsible for fusing evidence from multiple source maps to obtain reservoir potential assessment results with confidence levels. Potential is defined as a continuous potential field, a geologically compatible GMRF prior is applied, and specific observation factors for maps such as water cut, pressure, bubble point, and profile are constructed. Through joint maximum a posteriori (MAP) solution, the posterior mean μ and variance σ for each well layer are obtained. 2 And the confidence interval [L,U].

[0160] This module is the core computing engine, typically accelerated using high-performance computing clusters or GPUs. Internally, it contains GMRF models, observation factor models, and Bayesian inference algorithms such as INLA, ICM, and variational inference. The observation factor parameter φ... k The GMRF prior parameters are stored in data storage layer 180 and support online updates.

[0161] Sorting and Human-Machine Collaboration Module 150:

[0162] This system is responsible for ranking potential well layers and providing a user-friendly interface to support engineer feedback. Ranking is based on the posterior mean μ output by the Bayesian fusion module, and candidates with high uncertainty are backranked with a confidence penalty to calculate a robust ranking score S. Simultaneously, an evidence profile is generated to interpret the ranking results.

[0163] This module is tightly integrated with the user interface layer 190. The ranking algorithm is implemented here, and visualization tools are provided to display the potential distribution, CI components, and evidence profile. Users select conflict areas or label samples through the interface; this feedback data is collected and used to update the consistency assessment weights online. q and observation factor parameter φ k .

[0164] The incremental recalculation and redrawing module 160 is used to perform recalculation only on necessary steps and local areas based on the minimum recalculation set provided by the conflict location module 130. The rendering engine uses tile-based and version number management technology to replace only differential tiles and generate version differential heatmaps.

[0165] This module contains a distributed computing framework and a tile rendering engine. The computing framework is responsible for scheduling and executing local recalculation tasks. The tile rendering engine manages tile caching, version control, and concurrent rendering. An adaptive tile granularity adjustment strategy is implemented in this module to optimize recalculation and rendering efficiency.

[0166] The interface output module 170 is responsible for synchronizing key information such as standardized map tiles, CI decomposition, confidence intervals, conflict maps, and minimum recalculation suggestions generated by this system to the external EPBP work panel 200 through the parsing and write-back channel.

[0167] This module provides standardized API interfaces, such as RESTful API, supports data format conversion, such as JSON, and implements security auditing functions such as data encryption, access control, operation logging, and data verification.

[0168] Data storage layer 180 is used to store all raw data, standardized data cubes, method metadata, caliber conversion genealogy, registration parameters, CI evaluation results, Bayesian fusion model parameters, potential evaluation results, user feedback data, and tile data, etc.

[0169] Distributed file systems, such as HDFS, relational databases, such as PostgreSQL, NoSQL databases, such as MongoDB, or object storage, such as Min IO, can be used in combination to meet the storage needs and access performance requirements of different types of data.

[0170] User interface layer 190 provides a graphical user interface for reservoir engineers to perform operations such as data browsing, map visualization, parameter configuration, conflict viewing, manual feedback, and result analysis.

[0171] It can be implemented using web front-end technologies such as React and Vue.js, or desktop application frameworks such as Qt and Electron, and interact with back-end modules through APIs.

[0172] The EPBP work panel 200 is used by external collaboration platforms to receive evidence-based information output by this system, enabling reservoir engineers to select the best measures and make decisions.

[0173] This system interacts with the EPBP work panel 200 through the interface output module 170 to achieve seamless integration.

[0174] Finally, 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 them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some 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.

Claims

1. A method for implementing geological map scenarios based on measures to tap potential, characterized in that, Includes the following steps: Acquire multi-source maps and numerical sequences from both the model and field sides, establish strong primary key indexes for well ID, layer ID, grid cell, and time step under a unified reference system, and complete the specification of unit system and PVT caliber and generate method metadata field set record items. The calculation projects the map onto a multi-resolution pyramid and performs co-location registration, extracts contour lines and vectorizes them into a boundary set, minimizes the symmetric Hausdorff distance with elastic transformation, records the scale factor, displacement field and registration evidence and establishes a co-location raster. The four components of spatial overlap, boundary consistency, trend correlation, and anomalous co-location in the local neighborhood of the candidate well layer and the co-located grid are calculated and synthesized into a consistency index CI according to the grid sampling and concatenation rules. The component values ​​and identification field sets are recorded and associated with a unified number and time step label. Determine the relationship between the consistency index CI and the threshold. If it is lower than the threshold, generate a conflict map of the caliber layer, method layer and data layer, construct a candidate recalculation operation set, and form a factor importance sequence by single-factor local replay and difference and solve the minimum recalculation set. The incremental recalculation and block redrawing results are generated under the constraints of the minimum recalculation set. Differential tiles are replaced by tileization and version number management. The standardized map tiles, CI decomposition and potential a posteriori are synchronously sent to the external working panel through the parsing and write-back channel.

2. The method for implementing a geological map scenario based on potential tapping measures as described in claim 1, characterized in that: Data access and map registration establish a strong primary key index consisting of well ID, layer ID, grid cell, and time step. This index places multi-source maps and numerical sequences from both the model and field sides into a unified grid and reference system. It also registers the unit system and PVT caliber specifications, generates an index mapping table and a reverse index table, executes key conflict detection and conflict replacement strategies, saves index verification summaries and serialization snapshots, and sets the index version number.

3. The method for implementing a geological map scenario based on potential tapping of measures according to claim 2, characterized in that: Establish a caliber standard dictionary, a field alias table, and a measurement conversion table. Use the caliber standard dictionary to calibrate multi-source maps and numerical sequences item by item, record the caliber, conversion path, and batch number to the method metadata field set, and set the storage format and version number identifier. Establish a source data registration table to record the data source identifier, time step range, file path, and verification summary, and establish a reference relationship with the strong primary key index.

4. The method for implementing a geological map scenario based on measures to tap potential, as described in claim 1, is characterized in that; The multi-resolution pyramid extracts contour lines layer by layer and vectorizes them into a boundary set. Morphological thinning and curvature regularization are performed. The scale factor and displacement field are obtained by minimizing the symmetric Hausdorff distance using elastic transformation. The co-location raster and registration evidence are output, and the level number, node sequence number, sampling interval and boundary length threshold are recorded. The registration diagnosis log and parameter snapshot are generated and associated with the method metadata field set.

5. The method for implementing a geological map scenario based on potential tapping measures according to claim 4, characterized in that: Establish the correspondence between boundary sets and grid cells on the co-located grid, store node coordinates, orientation and topology labels, record operation logs, timestamps and operator identifiers for topology connection, cutting and merging, generate verification summaries and archive them, and register the topology change batch and audit trail number, as well as the associated boundary set version number and co-located grid version number in the method metadata field set.

6. The method for implementing a geological map scenario based on potential tapping of measures according to claim 1, characterized in that: The candidate well layer local neighborhood is limited by well ID and layer ID, and the neighborhood radius and maximum neighborhood depth are set. The grid sampling step size and sampling start point are set on the co-position grid. The calculation order, window size and cache structure of the four components of spatial overlap, boundary consistency, trend correlation and abnormal co-position are established. The parameter source, default value and coverage relationship are recorded in the method metadata field set, and the persistent key name and calculation task number are set.

7. The method for implementing a geological map scenario based on potential tapping measures as described in claim 6, characterized in that: Traverse the co-position grids at a fixed step size, obtain the values ​​of the four components: spatial overlap, boundary consistency, trend correlation, and abnormal co-position, and synthesize them into a consistency index (CI) cell by cell according to the concatenation rule. Record the sampling position, the values ​​of the four components and the CI number, and write the values ​​and numbers into the identifier field set and associate them with the well ID, layer ID and time step label, and save the snapshot and data verification summary.

8. The method for implementing a geological map scenario based on potential tapping of measures according to claim 1, characterized in that: The threshold is configured in the method metadata field set and has a threshold number. When determining the relationship between the consistency index CI and the threshold, a conflict map of the caliber layer, method layer and data layer is generated, and a conflict list table and a conflict grouping table are established. A sorting key and priority field are set, and a conflict index is generated for reference by the candidate recalculation operation set. The conflict cell number and the well layer object to which it belongs are recorded.

9. A method for implementing a geological map scenario based on potential tapping of measures, as described in claim 8, characterized in that: Based on the conflict map, a set of candidate recalculation operations is established. For a single factor, local replay is performed in the local neighborhood of the candidate well layer, and the difference is taken to form a factor importance sequence. The factor importance sequence is associated with the well ID and layer ID and stored. A persistent identifier and timestamp are generated. A foreign key association and retrieval key are established between the factor importance sequence and the well layer object, and the execution order and difference range are registered.

10. A geological map scene implementation system based on potential tapping of measures, used to implement the geological map scene implementation method based on potential tapping of measures as described in any one of claims 1-9, characterized in that, include: The memory and processor, wherein the memory stores instructions, and the processor implements the steps of the method of claim 1 when executing the instructions; the data structure of the instructions includes at least a strong primary key index, a co-location grid, a boundary set, a conflict map, a minimum recalculation set, and a tiled and version number management field, and associates the well ID, layer ID, grid cell, and time step with the strong primary key index and keeps them consistent with the unified reference system.

Images