A sedimentary mineral prediction analysis method and system based on sedimentary microfacies and knowledge graph

Abstract

The application discloses a sedimentary mineral prediction analysis method and system based on sedimentary microfacies and a knowledge graph, relates to the technical field of mineral prediction analysis, and can recognize the time sequence evolution process of microfacies from lithological profiles, logging data, sequence structures and other multi-source data through a sedimentary evolution sequence extraction module, and construct a structured evolution matrix, thereby reconstructing the evolution logic of a sedimentary environment in the time dimension. In the original method, parameters such as porosity and permeability are usually processed separately in reservoir evaluation, and are discontinuous in space and not coupled between attributes. However, the system matches key reservoir parameters in seismic inversion and logging interpretation with a sedimentary evolution structure through a multi-dimensional reservoir attribute coding module, generates a three-dimensional attribute tensor RS, abstracts sedimentary evolution from artificially divided sequence units into an evolution structure matrix Mac, avoids subjective dependence of traditional qualitative description, establishes an evolution model in a data-driven manner, and enhances the restoration capability of a dynamic paleoenvironment.

Description

Technical Field

[0001] This invention relates to the field of mineral prediction and analysis technology, specifically to a method and system for predicting and analyzing sedimentary minerals based on sedimentary microfacies and knowledge graphs. Background Technology

[0002] In recent years, with the development of technologies such as high-resolution seismic exploration, multi-well logging interpretation, and geological modeling, sedimentary microfacies identification and reservoir prediction have played a core role in oil and gas exploration, coalbed methane development, and groundwater resource assessment. However, these processes typically involve the collaborative interpretation of multi-source heterogeneous data, the identification of multi-scale evolutionary characteristics, and the analysis and reasoning of complex semantic relationships between evolution and occurrence. Especially in areas with complex geological structures and intense sedimentary evolution, traditional methods struggle to efficiently characterize the spatial continuity and temporal evolution paths of reservoirs.

[0003] In current sedimentary mineral prediction processes, sedimentary microfacies are often modeled in isolation as static attributes, making it difficult to accurately express their temporal evolution paths and dynamic relationships with neighboring microfacies. Meanwhile, reservoir attributes (such as porosity, permeability, and effective thickness) can be obtained through seismic inversion and well logging analysis, but they usually lack logical binding and semantic association with the sedimentary process in prediction modeling. This results in prediction models relying only on limited geological priors for spatial interpolation or empirical rule regression, making it difficult to adaptively update and deduce the trend of sediment-reservoir co-evolution.

[0004] The aforementioned shortcomings mainly stem from the current system's lack of dynamic knowledge representation mechanisms and deductive capabilities. On the one hand, the evolution of sedimentary microfacies exhibits strong temporal fragmentation and regional inconsistency, making it difficult to reconstruct its multi-period evolution sequence solely based on spatial similarity. On the other hand, the formation of reservoir attributes is often a multi-factor coupled process controlled by sedimentary environment evolution, diagenesis, and subsequent geological disturbances. If a semantic chain between sedimentation and attributes cannot be established, the prediction will deviate from the actual geological process. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for predicting and analyzing sedimentary mineral resources based on sedimentary microfacies and knowledge graphs, thus solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution: a sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graph, including a sedimentary evolution sequence extraction module, a multi-dimensional reservoir attribute encoding module, a semantic graph construction module, a graph embedding reasoning module, a joint prediction block generation module, and a self-feedback graph optimization module; The sedimentary evolution sequence extraction module extracts the evolution sequence of microfacies types over time from multi-source geological data and generates a sedimentary evolution structure matrix Mac. The multidimensional reservoir attribute encoding module extracts reservoir attributes from seismic inversion, well test interpretation, and well logging analysis, and matches them with the sedimentary evolution structure matrix Mac to obtain the three-dimensional attribute tensor RS; The semantic graph construction module establishes semantic edge relationships between nodes based on the sedimentary evolution structure matrix Mac and the three-dimensional attribute tensor RS, forming a three-dimensional semantic graph structure G; The graph embedding inference module processes the 3D semantic graph structure G using an embedding algorithm to generate a predicted value Ψres for the micro-phase reservoir state of each grid point in the future time. The joint prediction block generation module maps the acquired microfacies reservoir state prediction values ​​Ψres back to the sedimentary profile to generate microfacies distribution maps, reservoir probability maps, and confidence index Qpred. The self-feedback-based map optimization module re-evaluates whether the three-dimensional semantic map structure G needs to be reconstructed based on the microfacies distribution map, reservoir probability map, and confidence index Qpred.

[0007] Preferably, the sedimentary evolution sequence extraction module includes a sedimentary microfacies identification and temporal calibration unit and a sedimentary sequence normalization processing unit; The sedimentary microfacies identification and temporal calibration unit is based on the spatial sequence and stratigraphic division results of geological data sources to obtain the sedimentary microfacies type fac and the time interval Δt between adjacent microfacies for each sedimentary period; The sedimentary microfacies type (FAC) is obtained by: based on the profile map, well logging curves and thin section lithological description, and by using the time slice mapping function to identify each geological layer as a sedimentary microfacies type (FAC), including the littoral sandbar microfacies, the floodplain muddy microfacies, the underwater distributary mouth microfacies and the muddy infill microfacies. The formula for the time slice mapping function is: fac(ti) = Fmap(Litho(ti), Curve(ti), Seq(ti)); In the formula, fac(ti) represents the sedimentary microfacies type at time point ti, Fmap represents the mapping function that outputs the microfacies type according to the matching rules of lithology-curve-sequence, Litho(ti) represents the lithology category at time point ti, Curve(ti) represents the logging curve shape at time point ti, and Seq(ti) represents the sequence position at time point ti. The time interval Δt between adjacent microfacies is calculated using a chronological sequence based on the time interval between sedimentary facies. The sedimentary sequence normalization processing unit performs outlier detection and normalization on the obtained sedimentary microfacies type fac and the time interval Δt between adjacent microfacies to generate the sedimentary evolution structure matrix Mac; Outlier detection uses a deposition time density gradient function to process outliers in the time interval Δt between adjacent microfacies. When the time interval Δt between adjacent microfacies is greater than a preset time threshold, it indicates the presence of geologically unreasonable isolated points, which are then cleaned up. The normalization process performs nonlinear normalization on the time interval Δt between adjacent microphases using a time scale transformation function; The processing method is as follows: For the time interval Δt between adjacent sedimentary microfacies, the square root is first taken to obtain its square root value; then this square root value is divided by the maximum value among all the square roots of the time intervals; the normalized time interval Δt between adjacent microfacies is obtained; its numerical range is compressed to between 0 and 1, which is used to reflect the relative position and evolution rate of the microfacies change in the overall sedimentary sequence. The sedimentary evolution structure matrix Mac is obtained using the following formula: ; In the formula, tn represents the nth time.

[0008] Preferably, the multidimensional reservoir attribute coding module includes a reservoir physical property parameter extraction and standardization unit and an attribute and sedimentary evolution structure matching and tensor generation unit; The reservoir property parameter extraction and standardization unit extracts reservoir attributes, including porosity Rpor, permeability Rpe, and reservoir thickness Rd, from seismic inversion, well test interpretation, and well logging analysis, and then fits them to obtain an initial three-dimensional attribute volume. The porosity Rpor is obtained as follows: inversion is performed from pre-stack seismic data to extract P-wave impedance data; curve fitting is performed based on the true porosity and the impedance value of the corresponding depth segment; the impedance data of the entire profile is substituted into the transformation relationship to calculate the continuous spatial porosity Rpor. The permeability Rpe is obtained as follows: First, obtain the well test curves and interpretation reports of the target well and adjacent areas, extract key well test parameters, including productivity, pressure and volume factor; then, use the well test theory formula to back-calculate the permeability. The reservoir thickness Rd is obtained by preprocessing the logging data in the well logging analysis, identifying the reservoir development section, analyzing each well section by section, identifying the effective sandstone section that meets the conditions, and finally performing vertical integration to obtain the total thickness. The attribute matching and tensor generation unit matches the acquired initial three-dimensional attribute volume with the sedimentary evolution structure matrix Mac; within the time window, for each time point ti, the corresponding attribute value in the initial three-dimensional attribute volume is searched. When there is an inconsistency in sampling frequency, the time points adjacent to the time point ti are used for representation; when multiple points fall within the time window, a weighted average is calculated, with the weights inversely proportional to the time distance. For multiple attribute values ​​existing in the time window, for each attribute value at time point t, calculate the time difference with time point ti, and use 1 / |t-ti| as the weight to calculate the weighted porosity Rpor, permeability Rpe and reservoir thickness Rd. For a given depositional evolution time point ti, within its time window there are n attribute sampling points with times t1, t2, ..., tn, and corresponding attribute values ​​V1, V2, ..., Vn, which can be any of porosity, permeability, or thickness. The weighted average value xV(ti) of this attribute at time point ti is calculated using the following formula: ; In the formula, xV(ti) represents the weighted average attribute value at the target time point ti, Vj represents the corresponding attribute value in the three-dimensional attribute volume at the j-th time point, and tj represents the sampling time corresponding to the attribute value; The weighted porosity Rpor, permeability Rpe, and reservoir thickness Rd are fitted with the sedimentary microfacies type fac to obtain the three-dimensional attribute tensor RS.

[0009] Preferably, the semantic graph construction module includes a graph node and edge type definition unit and an edge weight quantization and graph structure construction unit; The graph node and edge type definition unit uniquely identifies the sedimentary microfacies type fac in the sedimentary evolution structure matrix Mac and converts it into a semantic node: Porosity Rpor, permeability Rpe, and reservoir thickness Rd in the three-dimensional attribute tensor RS are used as reservoir attribute nodes; then, timestamp nodes are constructed using time points ti to connect different time levels. By transforming the sedimentary evolution structure matrix Mac and the three-dimensional attribute tensor RS, three types of semantic edges are established, including evolutionary edges Eu→u, coupling edges Eu→R, and temporal edges Et→R. Among them, the evolutionary edge Eu→u represents the evolutionary transformation between microfacies, which originates from the sedimentary evolution structure matrix Mac and the sequence of microfacies types ordered by time. Each pair of adjacent sedimentary microfacies types fac(ti) and fac(ti+1) constitute a candidate evolutionary edge. The steps to establish the structure are as follows: S1, traverse each pair of adjacent time nodes in the sedimentary evolution structure matrix Mac; S2. Record the sedimentary microfacies type fac(ti) at time point ti and the sedimentary microfacies type fac(ti+1) at time point ti+1. S3. Establish the evolutionary edge Eu→u=(fac(ti), fac(ti+1)); S4. Calculate the time interval Δt between adjacent microphases, which is used to assign edge weights; S5. The rationality of the boundary weights can be further enhanced by combining historical phase transition frequency statistics (such as existing regional profile data and research literature). The coupling edge Eu→R represents the co-occurrence relationship between microfacies and reservoir properties, which is derived from the sedimentary evolution structure matrix Mac and the three-dimensional property tensor RS. According to the time point ti, the sedimentary microfacies type fac is matched with the property values ​​porosity Rpor, permeability Rpe and reservoir thickness Rd at the same time point. The steps are as follows: For each time point ti, extract the sedimentary microfacies type fac from the sedimentary evolution structure matrix Mac, and extract the attribute values: porosity Rpor, permeability Rpe and reservoir thickness Rd from the three-dimensional attribute tensor RS. Establish the coupling edge Eu→R={(fac(ti),Rpor),(fac(ti),Rpe),(fac(ti),Rd)} If the attribute value is a weighted value, then the edge weight is assigned using a weight exponent (such as a coupling strength function); The temporal edge Et→R represents the evolution trend of the attribute over time, which comes from the attribute time series in the three-dimensional attribute tensor RS. That is, under the condition of fixed spatial location, the attribute values ​​at different times constitute the time series. The steps are as follows: fix a certain spatial location and extract the porosity Rpor, permeability Rpe, and reservoir thickness Rd at different times; For consecutive time points (ti, ti+1), establish a temporal edge Et→R={(ti→Rpor(ti+1)), (ti→Rpe(ti+1)), (ti→Rpe(ti+1))}; Local variance or derivative over a time period can be introduced as edge weights to express the rate of change or fluctuation trend; Using sedimentary microfacies type fac, time point ti, porosity Rpor, permeability Rpe, and reservoir thickness Rd as nodes, establish a node set GN={fac(ti), Rpor, Rpe, Rd, ti}; Using the evolutionary edge Eu→u, the coupling edge Eu→R, and the temporal edge Et→R as relation edges, establish the relation edge set GE={Eu→u, Eu→R, Et→R}.

[0010] Preferably, the edge weight quantization and graph structure construction unit quantifies each type of edge, constructs an edge weight function, including microfacies evolution edge weight Wu→u, microfacies and reservoir attribute coupling edge weight Wu→R, and time and reservoir attribute evolution edge weight Wt→R, and embeds it into the graph structure; By quantizing the evolution edge Eu→u, the micro-phase evolution edge weight Wu→u is obtained; The microfacies evolution boundary weight Wu→u is obtained as follows: First, take the time interval between adjacent sedimentary microfacies types fac(ti) and fac(ti+1), and then take the reciprocal of this value to represent the phase transition rate; then, calculate the evolution frequency from fac(ti) to fac(ti+1); finally, multiply the evolution frequency by the phase transition rate to obtain the microfacies evolution boundary weight Wu→u. By quantizing the coupling edge Eu→R, the coupling edge weight Wu→R between microfacies and reservoir properties is obtained. The method for obtaining the coupling weight Wu→R between microfacies and reservoir properties is as follows: First, the porosity Rpor and permeability Rpe at time point ti are added together to obtain the reservoir occurrence quality; then, the lithological change gradient in the vertical direction at the current location is extracted, its absolute value is taken, 1 is added, and then the square root is taken as the denominator to represent the degree of formation disturbance; finally, the reservoir occurrence quality is divided by the degree of formation disturbance to obtain the coupling weight Wu→R between microfacies and reservoir properties. By quantizing the temporal edge Et→R, the time-reservoir attribute evolution edge weight Wt→R is obtained; The method for obtaining the time-reservoir attribute evolution boundary weight Wt→R is as follows: First, select a time window centered on time point ti and extract the attribute value sequence within the window; then, calculate the standard deviation of the attribute values; finally, add a constant 1 to the standard deviation and take the reciprocal to obtain the time-reservoir attribute evolution boundary weight Wt→R. The obtained node set GN={fac(ti), Rpor, Rpe, Rd, ti} is combined with the quantized relation edge set GE={Eu→u:Wu→u, Eu→R:Wu→R, Et→R:Wt→R} to obtain the three-dimensional semantic graph structure G={GN, GE}.

[0011] Preferably, the graph embedding reasoning module processes the three-dimensional semantic graph structure G by using an embedding algorithm, transforms the node and edge relationships into vector form, unifies the embedding dimension to d, and outputs the embedding vector; Based on the embedded vector, a microfacies prediction function is constructed to predict the microfacies reservoir state at grid points at a future time point t+Δt, and the predicted value Ψres of the microfacies reservoir state is obtained. The predicted value Ψres of the microfacies reservoir state is obtained by the following formula: ; In the formula, Ψres(x, y, t+Δt) represents the predicted value of the microfacies reservoir state at grid point (x, y) at time point t+Δt, vfac(ti) represents the sedimentary microfacies type embedding vector at time point ti, vRpor(ti) represents the porosity embedding vector at time point ti, vRpe(ti) represents the permeability embedding vector at time point ti, vti+1-vti represents the semantic transition intensity of the time vector in the future, and ∇zLith(x, y) represents the lithological change gradient at grid point (x, y); Steps to obtain the lithological variation gradient ∇zLith: At each spatial grid point (x, y), extract the lithological data sequence in its vertical direction (i.e., the depth direction); The above lithological information is organized into structured data that varies with depth, that is, each depth layer corresponds to a lithological value. These lithological values ​​can be numerical (such as gamma value) or numbers mapped by lithological classification labels (such as mudstone=1, sandstone=2). On the same vertical profile, observe the degree of variation between adjacent lithological values ​​layer by layer; the greater the variation, the more drastic the lithological change at that point in the depth direction. At each depth point, the lithological differences between adjacent depth points are analyzed; the increase or decrease in lithological values ​​between three or five layers is compared using a sliding window method; the more dramatic the jump, the greater the lithological gradient. When local jumps or outliers caused by measurement noise occur in the depth layer, median filtering or statistical smoothing is used to smooth out local unrealistic fluctuations. The lithological variation index obtained at each grid point is assigned a value to form a two-dimensional spatially distributed lithological gradient layer, which is used as the lithological variation gradient ∇zLith. The obtained microfacies reservoir state prediction values ​​Ψres are analyzed to determine the quality of the mineral deposits. The judgment method is as follows: When the predicted value of microfacies reservoir state Ψres > 0.75, it indicates a preferred mineral deposit area; stable microfacies, good reservoir properties, and low lithological disturbance are recommended as key areas for mineral exploration. When 0.75 ≥ the predicted value of microfacies reservoir state Ψres > 0.55, it indicates a candidate mineral deposit area; the microfacies or reservoir performance is average, but the trend is still stable, and it can be used as a secondary target area. When 0.55 ≥ the predicted value Ψres of microfacies reservoir state, it indicates that the mineral deposit area is not recommended; the lithology changes drastically or the properties are discontinuous, and the prediction does not have the priority for mineral exploration.

[0012] Preferably, the joint prediction block generation module maps each microfacies reservoir state prediction value Ψres to a two-dimensional geological profile, locates it to a specific grid point (x, y), and constructs a microfacies distribution map and a reservoir probability map; The microfacies distribution map selects the corresponding microfacies type label based on the microfacies reservoir state prediction value Ψres; The method for constructing the microphase distribution map is as follows: First, obtain the predicted value of the microfacies reservoir state Ψres(x,y,t+Δt), the microfacies type identifier (e.g., channel facies=1, cross-bedding facies=2, colluvial facies=3...) at grid point (x,y) at time point t+Δt, and the sedimentary microfacies type fac(ti) at time point ti. For each grid point (x, y), read its microfacies reservoir state prediction value; map the score value to the corresponding microfacies type; Write the microphase type label of each grid point into the corresponding two-dimensional profile position, and use spatial interpolation (such as IDW, bilinear) to fill in the boundary ambiguity area in the valueless area; Different colors are assigned according to the microphase type number, and the images are rendered as a visual layer to form a microphase distribution map. The reservoir probability map maps the probability distribution of reservoir potential based on the microfacies reservoir state prediction value Ψres. The method for constructing a probability distribution plot is as follows: First, the predicted values ​​of the microphase reservoir state Ψres(x,y,t+Δt), porosity Rpor, and permeability Rpe are obtained at the grid point (x,y) at time point t+Δt. Normalize Ψres(x, y, t+Δt) to compress all values ​​to the range of 0~1; The predicted microfacies reservoir state value at each grid point is used as the heat value; the heat value is mapped to a color gradient, such as blue for low values ​​and red for high values; bilinear interpolation is used to make the layer smooth and continuous. Legend settings are set to 5 to 7 color levels (e.g., 0.0-0.2, 0.2-0.4, ..., 0.8-1.0); Set partition thresholds: for example, a value greater than 0.75 is a high-potential zone, and a value between 0.55 and 0.75 is a medium-potential zone, and obtain a probability distribution map; The predicted value of microfacies reservoir state Ψres is combined with the permeability Rpe to calculate the confidence index Qpred, which is then compared with the preset confidence threshold Tred to determine the confidence of the layer. The credibility index Qpred is obtained using the following formula: ; In the formula, Qpred(x,y) represents the confidence index at grid point (x,y), Rpe(x,y,t) represents the permeability at grid point (x,y) at time t, log represents the logarithmic function, and dfac(x,y) represents the spatial distance from grid point (x,y) to the center of the nearest known stable microphase region. The credibility of a layer is obtained through matching methods: When the confidence index Qpred ≥ the confidence threshold Tred, it indicates that the confidence is high and the microfacies reservoir state prediction value Ψres is effective; When the confidence index Qpred is less than the confidence threshold Tred, it indicates low confidence and the microfacies reservoir state prediction value Ψres is invalid.

[0013] Preferably, the self-feedback graph optimization module includes a graph structure variation detection and expansion triggering unit and a connectivity change evaluation and graph relearning unit; Comparison of newly added regions in microphase distribution maps and reservoir probability maps with the coverage of nodes in the three-dimensional semantic map structure G, and the expansion triggering unit of map structure variation detection and expansion triggering unit. If the newly added region is not covered by the existing node set GN or relation edge set GE in the three-dimensional semantic graph structure G, it is determined that the graph structure is insufficient. Based on new microfacies types, attribute distribution patterns, or new evolutionary paths appearing in the layer, the following are automatically added: new microfacies nodes; new reservoir attribute nodes; and new evolutionary edge relationships. Construct the supplemented three-dimensional semantic graph structure nG={GN+ΔN,GE+ΔE}; In the formula, ΔN represents the number of node additions, and ΔE represents the number of edge structure additions.

[0014] Preferably, the connectivity change assessment and graph relearning unit compares the degree of change in the connectivity between nodes between the three-dimensional semantic graph structure G and the supplemented three-dimensional semantic graph structure nG, and uses the graph connectivity index Cconn to evaluate the number of paths and the connectivity density between nodes in the graph, and calculates the connectivity change rate ΔCconn. The graph connectivity index Cconn is obtained as follows: S1. Arrange all nodes according to their numbers and establish the connection relationship between the nodes; if there is an edge between node A and node B (whether it is a direct edge or an indirect edge), mark the corresponding position in the adjacency matrix as "1" or record the edge weight. S2. Set a path length threshold k (usually 1~3) to indicate whether there is a connection within a k-step path; for example, if node A can be connected to node B through other nodes within 2 steps, then the two nodes are considered "connected". S3. Traverse all pairs of nodes in the graph and determine whether they are reachable within k steps. If they are reachable, they are marked as a valid connection. The total number of reachable pairs is the total number of connected paths. S4. Divide the number of effective connected node pairs by the total number of all possible node pairs to obtain the graph connectivity index Cconn, which ranges from 0 to 1. The closer it is to 1, the denser the graph structure and the higher the propagation efficiency; the lower it is, the more isolated regions or faults exist in the structure.

[0015] The rate of change of connectivity ΔCconn is obtained by the following formula: ; In the formula, nCconn represents the graph connectivity index of the supplemented three-dimensional semantic graph structure, and oCconn represents the graph connectivity index of the three-dimensional semantic graph structure; Analyze the connectivity change rate ΔCconn to determine whether the three-dimensional semantic graph structure G needs to be reconstructed; The judgment method is as follows: When the rate of change in connectivity ΔCconn < 0.15, it indicates that the change in connectivity is within the normal range and there is no need to reconstruct the graph structure. When the connectivity change rate ΔCconn≥0.15, it indicates abnormal connectivity change. The graph structure is reconstructed, and the embedding vector of each node is regenerated based on the supplemented three-dimensional semantic graph structure nG.

[0016] A sedimentary mineral resource prediction and analysis method based on sedimentary microfacies and knowledge graphs includes the following steps: Step 1: The sedimentary evolution sequence extraction module extracts the evolution sequence of microfacies types over time from multi-source geological data and generates a sedimentary evolution structure matrix Mac. Step 2: The multidimensional reservoir attribute encoding module extracts reservoir attributes from seismic inversion, well test interpretation, and well logging analysis, and matches them with the sedimentary evolution structure matrix Mac to obtain the three-dimensional attribute tensor RS. Step 3: The semantic graph construction module establishes semantic edge relationships between nodes based on the sedimentary evolution structure matrix Mac and the three-dimensional attribute tensor RS, forming a three-dimensional semantic graph structure G; Step 4: The graph embedding and reasoning module performs embedding algorithm processing on the three-dimensional semantic graph structure G to generate the micro-phase reservoir state prediction value Ψres for each grid point in the future time. Step 5: The joint prediction block generation module maps the obtained microfacies reservoir state prediction values ​​Ψres back to the sedimentary profile to generate a microfacies distribution map, a reservoir probability map, and a confidence index Qpred. Step 6: The self-feedback map optimization module re-determines whether the three-dimensional semantic map structure G needs to be reconstructed based on the microfacies distribution map, reservoir probability map, and confidence index Qpred.

[0017] This invention provides a method and system for predicting and analyzing sedimentary mineral deposits based on sedimentary microfacies and knowledge graphs, which has the following beneficial effects: (1) During system operation, through the sedimentary evolution sequence extraction module, the system can identify the temporal evolution process of microfacies from multi-source data such as lithological profiles, well logging data, and sequence structure, and construct a structured evolution matrix to reconstruct the evolution logic of the sedimentary environment in the time dimension. In the original method, parameters such as porosity and permeability are usually processed separately in reservoir evaluation, resulting in spatial discontinuity and no coupling between attributes. However, this system uses the "multi-dimensional reservoir attribute encoding module" to match key reservoir parameters in seismic inversion and well logging interpretation with the sedimentary evolution structure, generating a three-dimensional attribute tensor RS, thus establishing a spatial attribute coupling mapping between "sedimentation-reservoir".

[0018] Abstracting sedimentary evolution from artificially divided sequence units into an evolutionary structure matrix (Mac) avoids the subjective reliance on traditional qualitative descriptions and instead uses a data-driven approach to build evolutionary models, enhancing the ability to reconstruct paleoenvironmental dynamics.

[0019] (2) By synchronously and structurally expressing sedimentary microfacies types and reservoir physical parameters on the time axis, a three-dimensional map expression framework of three heterogeneous nodes of microfacies-reservoir-time is constructed for the first time. It can perform semantic reasoning and trend prediction on the relationship between the evolution of sedimentary environment and reservoir occurrence conditions, and no longer relies solely on two-dimensional static geological maps or single attribute analysis, thus overcoming the shortcomings of poor timeliness and weak correlation of existing methods.

[0020] By identifying the three key elements of lithology, well logging morphology, and sequence location for each sedimentary timepiece, precise identification and standardized mapping of sedimentary microfacies types were achieved, constructing a temporally continuous sedimentary microfacies evolution sequence. This addresses the issues of strong subjectivity and difficulty in quantitative modeling in traditional stratigraphic descriptions. Furthermore, a density gradient anomaly cleaning and time-scale nonlinear transformation mechanism were introduced into the time intervals of sedimentary microfacies, enhancing the ability to process temporally heterogeneous data in complex sedimentary contexts.

[0021] (3) By finely defining evolutionary edges, coupling edges, and temporal edges, and constructing an edge weight quantification mechanism, the digital expression of the evolution intensity between sedimentary microfacies, their correlation with reservoir occurrence quality, and the strength of the trend of attribute changes over time was realized, transforming the originally ambiguous geological evolution phenomena into quantifiable edge relationships. This three-edge joint expression mechanism constitutes a complete reasoning chain, enhancing the sensitivity and interpretability of subsequent map embedding and prediction models to structural changes, while avoiding the problems of microfacies attribute coupling breaks, discontinuities, or information redundancy in traditional methods.

[0022] (4) Based on the three-dimensional semantic graph structure G, multi-dimensional heterogeneous nodes (such as microfacies type, reservoir attributes, and timestamps) are transformed into vector expressions of a unified dimension through an embedding algorithm, thereby constructing a microfacies-reservoir joint prediction function for future time slices. This mechanism breaks through the limitations of traditional rule-based matching or single-time attribute mapping and has stronger spatiotemporal extrapolation capabilities. The microfacies reservoir state prediction value Ψres comprehensively considers the coupling changes of semantic time jump variables and local lithological disturbance gradient ∇zLith, which enables the system to maintain prediction stability and resolution when facing complex geological environments such as rapidly evolving zones and heterogeneous lithological zones, and significantly improves the ability to express anomalous sedimentary areas. Attached Figure Description

[0023] Figure 1 This is a schematic flowchart of a sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graphs according to the present invention. Figure 2 This is a schematic diagram illustrating the steps of a sedimentary mineral prediction and analysis method based on sedimentary microfacies and knowledge graphs according to the present invention. Figure 3 This is a schematic diagram of the three-dimensional semantic map structure acquisition process of the present invention; Figure 4 This is a trend chart of the credibility index of the present invention. Detailed Implementation

[0024] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0025] Example 1: This invention provides a sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graphs. Please refer to [link to relevant documentation]. Figures 1 to 4 It includes a sedimentary evolution sequence extraction module, a multi-dimensional reservoir attribute encoding module, a semantic map construction module, a map embedding and reasoning module, a joint prediction block generation module, and a self-feedback map optimization module; The sedimentary evolution sequence extraction module extracts the evolution sequence of microfacies types over time from multi-source geological data and generates a sedimentary evolution structure matrix Mac. The multidimensional reservoir attribute encoding module extracts reservoir attributes from seismic inversion, well test interpretation, and well logging analysis, and matches them with the sedimentary evolution structure matrix Mac to obtain the three-dimensional attribute tensor RS; The semantic graph construction module establishes semantic edge relationships between nodes based on the sedimentary evolution structure matrix Mac and the three-dimensional attribute tensor RS, forming a three-dimensional semantic graph structure G; The graph embedding inference module processes the 3D semantic graph structure G using an embedding algorithm to generate a predicted value Ψres for the micro-phase reservoir state of each grid point in the future time. The joint prediction block generation module maps the acquired microfacies reservoir state prediction values ​​Ψres back to the sedimentary profile to generate microfacies distribution maps, reservoir probability maps, and confidence index Qpred. The self-feedback-based map optimization module re-evaluates whether the three-dimensional semantic map structure G needs to be reconstructed based on the microfacies distribution map, reservoir probability map, and confidence index Qpred.

[0026] In this embodiment, through the sedimentary evolution sequence extraction module, the system can identify the temporal evolution process of microfacies from multi-source data such as lithological profiles, well logging data, and sequence stratigraphy, and construct a structured evolution matrix to reconstruct the evolutionary logic of the sedimentary environment in the time dimension. In existing methods for reservoir evaluation, parameters such as porosity and permeability are typically processed separately, resulting in spatial discontinuities and a lack of coupling between attributes. However, this system, through a "multi-dimensional reservoir attribute encoding module," matches key reservoir parameters from seismic inversion and well logging interpretation with the sedimentary evolution structure, generating a three-dimensional attribute tensor RS, thus establishing a spatial attribute coupling mapping between "sedimentation and reservoir."

[0027] The system introduces a semantic graph construction module, treating micro-phase nodes, attribute nodes, and time nodes as entities. These are connected through relationships such as "evolutionary edges," "attribute edges," and "time edges" to form a graph structure with spatiotemporal evolution and attribute logic, thus advancing the graph's expressive power from "spatial topology" to "semantic drive." Through the "graph embedding and reasoning module," the system uses graph neural networks or other embedding models to perform deep learning and state reasoning on the graph structure, outputting the micro-phase-reservoir joint state Ψres for each grid point at future moments, providing trend-based information for mineral exploration block prediction.

[0028] In the self-feedback graph optimization module, this system introduces the confidence index Qpred and the graph connectivity index Cconn to automatically determine whether new data causes significant changes in the graph structure, and performs graph expansion and edge weight updates accordingly, thereby achieving the system's evolutionary self-adaptation.

[0029] Abstracting sedimentary evolution from artificially divided sequence units into an evolutionary structure matrix (Mac) avoids the subjective reliance on traditional qualitative descriptions and instead uses a data-driven approach to build evolutionary models, enhancing the ability to reconstruct paleoenvironmental dynamics.

[0030] By matching the three-dimensional attribute tensor RS with the sedimentary evolution structure matrix Mac, the spatial distribution of reservoir physical parameters can be reconstructed. This helps to discover the aggregation patterns of high-potential reservoirs in the sedimentary evolution path, overcoming the deficiency of only looking at attributes without considering evolution. Instead of relying on black-box models for direct prediction, it uses semantic graphs to trace the results back to each edge, each evolution path, and the coupling logic of physical properties, making the prediction results traceable and interpretable.

[0031] The system can not only predict future states, but also achieve structural reconstruction and parameter self-learning through reliable scoring and structural connectivity feedback. It has the ability to perform long-term operation and maintenance and intelligent optimization, and can adapt to the new data and new scenarios that are constantly accumulating in the future.

[0032] Example 2: This example is an explanation of Example 1. Please refer to the example provided. Figure 3 Specifically: the sedimentary evolution sequence extraction module includes a sedimentary microfacies identification and temporal calibration unit and a sedimentary sequence normalization processing unit; The sedimentary microfacies identification and temporal calibration unit is based on the spatial sequence and stratigraphic division results of geological data sources to obtain the sedimentary microfacies type fac and the time interval Δt between adjacent microfacies for each sedimentary period; The sedimentary microfacies type (FAC) is obtained by: based on the profile map, well logging curves and thin section lithological description, and by using the time slice mapping function to identify each geological layer as a sedimentary microfacies type (FAC), including the littoral sandbar microfacies, the floodplain muddy microfacies, the underwater distributary mouth microfacies and the muddy infill microfacies. The formula for the time slice mapping function is: fac(ti) = Fmap(Litho(ti), Curve(ti), Seq(ti)); In the formula, fac(ti) represents the sedimentary microfacies type at time point ti, Fmap represents the mapping function that outputs the microfacies type according to the matching rules of lithology-curve-sequence, Litho(ti) represents the lithology category at time point ti, Curve(ti) represents the logging curve shape at time point ti, and Seq(ti) represents the sequence position at time point ti. The time interval Δt between adjacent microfacies is calculated using a chronological sequence based on the time interval between sedimentary facies. The sedimentary sequence normalization processing unit performs outlier detection and normalization on the obtained sedimentary microfacies type fac and the time interval Δt between adjacent microfacies to generate the sedimentary evolution structure matrix Mac; Outlier detection uses a deposition time density gradient function to process outliers in the time interval Δt between adjacent microfacies. When the time interval Δt between adjacent microfacies is greater than a preset time threshold, it indicates the presence of geologically unreasonable isolated points, which are then cleaned up. The normalization process performs nonlinear normalization on the time interval Δt between adjacent microphases using a time scale transformation function; The processing method is as follows: For the time interval Δt between adjacent sedimentary microfacies, first take the square root to obtain its square root value; then divide this square root value by the maximum value among all the square roots of the time intervals; obtain the normalized time interval Δt between adjacent microfacies. The sedimentary evolution structure matrix Mac is obtained using the following formula: ; In the formula, tn represents the nth time.

[0033] The multidimensional reservoir attribute coding module includes a reservoir physical property parameter extraction and standardization unit and an attribute and sedimentary evolution structure matching and tensor generation unit; The reservoir property parameter extraction and standardization unit extracts reservoir attributes, including porosity Rpor, permeability Rpe, and reservoir thickness Rd, from seismic inversion, well test interpretation, and well logging analysis, and then fits them to obtain an initial three-dimensional attribute volume. The porosity Rpor is obtained as follows: inversion is performed from pre-stack seismic data to extract P-wave impedance data; curve fitting is performed based on the true porosity and the impedance value of the corresponding depth segment; the impedance data of the entire profile is substituted into the transformation relationship to calculate the continuous spatial porosity Rpor. The permeability Rpe is obtained as follows: First, obtain the well test curves and interpretation reports of the target well and adjacent areas, extract key well test parameters, including productivity, pressure and volume factor; then, use the well test theory formula to back-calculate the permeability. The reservoir thickness Rd is obtained by preprocessing the logging data in the well logging analysis, identifying the reservoir development section, analyzing each well section by section, identifying the effective sandstone section that meets the conditions, and finally performing vertical integration to obtain the total thickness. The attribute matching and tensor generation unit matches the acquired initial three-dimensional attribute volume with the sedimentary evolution structure matrix Mac; within the time window, for each time point ti, the corresponding attribute value in the initial three-dimensional attribute volume is searched. When there is an inconsistency in sampling frequency, the time points adjacent to the time point ti are used for representation; when multiple points fall within the time window, a weighted average is calculated, with the weights inversely proportional to the time distance. For multiple attribute values ​​existing in the time window, for each attribute value at time point t, calculate the time difference with time point ti, and use 1 / |t-ti| as the weight to calculate the weighted porosity Rpor, permeability Rpe and reservoir thickness Rd. The weighted porosity Rpor, permeability Rpe, and reservoir thickness Rd are fitted with the sedimentary microfacies type fac to obtain the three-dimensional attribute tensor RS.

[0034] In this embodiment, by synchronously and structurally expressing sedimentary microfacies types and reservoir physical parameters on the time axis, a three-dimensional map expression framework of three heterogeneous nodes—microfacies, reservoir, and time—is constructed for the first time. This framework enables semantic reasoning and trend prediction of the relationship between the evolution of the sedimentary environment and reservoir occurrence conditions, and no longer relies solely on two-dimensional static geological maps or single attribute analysis, thus overcoming the shortcomings of existing methods such as poor timeliness and weak correlation.

[0035] By identifying the three key elements of lithology, well logging morphology, and sequence location for each sedimentary timepiece, precise identification and standardized mapping of sedimentary microfacies types were achieved, constructing a temporally continuous sedimentary microfacies evolution sequence. This addresses the issues of strong subjectivity and difficulty in quantitative modeling in traditional stratigraphic descriptions. Furthermore, a density gradient anomaly cleaning and time-scale nonlinear transformation mechanism were introduced into the time intervals of sedimentary microfacies, enhancing the ability to process temporally heterogeneous data in complex sedimentary contexts.

[0036] The system systematically integrates data from three different sources: seismic inversion (for porosity), well test interpretation (for permeability), and well logging analysis (for thickness), making the extraction of reservoir attributes more targeted and accurate, and avoiding the one-sidedness of a single data source. In the process of matching reservoir attributes with sedimentary sequences, a sampling frequency compatibility processing and time-inverse weighting mechanism are introduced, effectively solving the common problem of inconsistent temporal resolution and enhancing the model's ability to fit the temporal changes of real geological information.

[0037] Example 3: This example is an explanation of Example 2. Please refer to the example provided. Figure 1 Specifically: the semantic graph construction module includes a graph node and edge type definition unit and an edge weight quantization and graph structure construction unit; The graph node and edge type definition unit uniquely identifies the sedimentary microfacies type fac in the sedimentary evolution structure matrix Mac and converts it into a semantic node: Porosity Rpor, permeability Rpe, and reservoir thickness Rd in the three-dimensional attribute tensor RS are used as reservoir attribute nodes; then, timestamp nodes are constructed using time points ti to connect different time levels. By transforming the sedimentary evolution structure matrix Mac and the three-dimensional attribute tensor RS, three types of semantic edges are established, including evolutionary edge Eu→u, coupling edge Eu→R, and temporal edge Et→R. Among them, the evolution edge Eu→u represents the evolutionary transition between microfacies, which comes from the sedimentary evolution structure matrix Mac and the sequence of microfacies types ordered by time. Each pair of adjacent sedimentary microfacies types fac(ti) and fac(ti+1) constitute a candidate evolution edge. The coupling edge Eu→R represents the co-occurrence relationship between microfacies and reservoir properties, which is derived from the sedimentary evolution structure matrix Mac and the three-dimensional property tensor RS. According to the time point ti, the sedimentary microfacies type fac is matched with the property values ​​porosity Rpor, permeability Rpe and reservoir thickness Rd at the same time point. The temporal edge Et→R represents the evolution trend of the attribute over time, which comes from the attribute time series in the three-dimensional attribute tensor RS. That is, under the condition of fixed spatial location, the attribute values ​​at different times constitute the time series. Using sedimentary microfacies type fac, time point ti, porosity Rpor, permeability Rpe, and reservoir thickness Rd as nodes, establish a node set GN={fac(ti), Rpor, Rpe, Rd, ti}; Using the evolutionary edge Eu→u, the coupling edge Eu→R, and the temporal edge Et→R as relation edges, establish the relation edge set GE={Eu→u, Eu→R, Et→R}.

[0038] The edge weight quantization and graph structure construction unit quantifies each type of edge, constructs edge weight functions, including microfacies evolution edge weight Wu→u, microfacies and reservoir attribute coupling edge weight Wu→R, and time and reservoir attribute evolution edge weight Wt→R, and embeds them into the graph structure. By quantizing the evolution edge Eu→u, the micro-phase evolution edge weight Wu→u is obtained; The microfacies evolution boundary weight Wu→u is obtained as follows: First, take the time interval between adjacent sedimentary microfacies types fac(ti) and fac(ti+1), and then take the reciprocal of this value to represent the phase transition rate; then, calculate the evolution frequency from fac(ti) to fac(ti+1); finally, multiply the evolution frequency by the phase transition rate to obtain the microfacies evolution boundary weight Wu→u. By quantizing the coupling edge Eu→R, the coupling edge weight Wu→R between microfacies and reservoir properties is obtained. The method for obtaining the coupling weight Wu→R between microfacies and reservoir properties is as follows: First, the porosity Rpor and permeability Rpe at time point ti are added together to obtain the reservoir occurrence quality; then, the lithological change gradient in the vertical direction at the current location is extracted, its absolute value is taken, 1 is added, and then the square root is taken as the denominator to represent the degree of formation disturbance; finally, the reservoir occurrence quality is divided by the degree of formation disturbance to obtain the coupling weight Wu→R between microfacies and reservoir properties. By quantizing the temporal edge Et→R, the time-reservoir attribute evolution edge weight Wt→R is obtained; The method for obtaining the time-reservoir attribute evolution boundary weight Wt→R is as follows: First, select a time window centered on time point ti and extract the attribute value sequence within the window; then, calculate the standard deviation of the attribute values; finally, add a constant 1 to the standard deviation and take the reciprocal to obtain the time-reservoir attribute evolution boundary weight Wt→R. The obtained node set GN={fac(ti), Rpor, Rpe, Rd, ti} is combined with the quantized relation edge set GE={Eu→u:Wu→u, Eu→R:Wu→R, Et→R:Wt→R} to obtain the three-dimensional semantic graph structure G={GN, GE}.

[0039] In this embodiment, by deeply fusing the sedimentary evolution structure matrix Mac with the three-dimensional attribute tensor RS, sedimentary microfacies, reservoir properties, and time-series data are transformed into unified semantic nodes. A graph structure is then constructed using three types of semantic edges, establishing a multi-dimensional integrated expression framework for sedimentary evolution, reservoir properties, and temporal changes. This approach overcomes the limitations of traditional geological data's single-dimensional recording and lack of semantic relationship modeling. It enables the systematic modeling and encoding of temporal relationships and spatial attribute changes in complex geological evolution processes, thereby enhancing the ability to structurally describe complex geological scenarios.

[0040] By precisely defining evolutionary edges, coupling edges, and temporal edges, and constructing an edge weight quantification mechanism, this method achieves a digital expression of the evolutionary intensity between sedimentary microfacies, their correlation with reservoir occurrence quality, and the strength of the trend of attribute changes over time. This transforms the originally ambiguous geological evolution phenomena into quantifiable edge relationships. This three-edge joint expression mechanism constitutes a complete inference chain, enhancing the sensitivity and interpretability of subsequent map embedding and prediction models to structural changes, while avoiding the problems of broken coupling, discontinuity, or information redundancy of microfacies attributes in traditional methods.

[0041] The calculation of edge weights incorporates key factors such as "phase transition rate," "reservoir occurrence quality / formation disturbance degree," and "attribute temporal fluctuation degree," making each edge no longer a static connection but possessing dynamic meaning and physical interpretation. The existence of edge weights allows the graph to not only record structure but also convey intensity and trends, enabling dynamic adjustment and relearning of the subsequent graph structure. This mechanism provides a foundation for adaptive processing of the graph in subsequent embedded reasoning and self-feedback evolution, overcoming the problem that static knowledge graphs struggle to represent complex spatiotemporal structures, thereby enhancing the system's ability to predict and judge future sedimentary states and reservoir change trends.

[0042] Example 4: This example is an explanation of Example 3. Please refer to the example provided. Figure 3 and Figure 4 Specifically: The graph embedding reasoning module processes the three-dimensional semantic graph structure G by using an embedding algorithm, transforms the node and edge relationships into vector form, unifies the embedding dimension to d, and outputs the embedding vector; Based on the embedded vector, a microfacies prediction function is constructed to predict the microfacies reservoir state at grid points at a future time point t+Δt, and the predicted value Ψres of the microfacies reservoir state is obtained. The predicted value Ψres of the microfacies reservoir state is obtained by the following formula: ; In the formula, Ψres(x, y, t+Δt) represents the predicted value of the microfacies reservoir state at grid point (x, y) at time point t+Δt, vfac(ti) represents the sedimentary microfacies type embedding vector at time point ti, vRpor(ti) represents the porosity embedding vector at time point ti, vRpe(ti) represents the permeability embedding vector at time point ti, vti+1-vti represents the semantic transition intensity of the time vector in the future, and ∇zLith(x, y) represents the lithological change gradient at grid point (x, y); The obtained microfacies reservoir state prediction values ​​Ψres are analyzed to determine the quality of the mineral deposits. The judgment method is as follows: When the predicted value of microfacies reservoir state Ψres > 0.75, it indicates a preferred mineral deposit area; When 0.75 ≥ the predicted value of microfacies reservoir state Ψres > 0.55, it indicates a candidate mineral deposit area; When 0.55 ≥ the predicted value Ψres of the microfacies reservoir state, it indicates that the mineral deposit area is not recommended.

[0043] The joint prediction block generation module maps each microfacies reservoir state prediction value Ψres to a two-dimensional geological profile, locates it to a specific grid point (x, y), and constructs a microfacies distribution map and a reservoir probability map. The microfacies distribution map selects the corresponding microfacies type label based on the microfacies reservoir state prediction value Ψres; The reservoir probability map maps the probability distribution of reservoir potential based on the microfacies reservoir state prediction value Ψres. The predicted value of microfacies reservoir state Ψres is combined with the permeability Rpe to calculate the confidence index Qpred, which is then compared with the preset confidence threshold Tred to determine the confidence of the layer. The credibility index Qpred is obtained using the following formula: ; In the formula, Qpred(x,y) represents the confidence index at grid point (x,y), Rpe(x,y,t) represents the permeability at grid point (x,y) at time t, log represents the logarithmic function, and dfac(x,y) represents the spatial distance from grid point (x,y) to the center of the nearest known stable microphase region. The credibility of a layer is obtained through matching methods: When the confidence index Qpred ≥ the confidence threshold Tred, it indicates that the confidence is high and the microfacies reservoir state prediction value Ψres is effective; When the confidence index Qpred is less than the confidence threshold Tred, it indicates low confidence and the microfacies reservoir state prediction value Ψres is invalid.

[0044] In this embodiment, based on the three-dimensional semantic graph structure G, an embedding algorithm is used to transform multi-dimensional heterogeneous nodes (such as microfacies type, reservoir attributes, and timestamps) into vector representations of a unified dimension, thereby constructing a joint microfacies-reservoir prediction function for future time slices. This mechanism overcomes the limitations of traditional rule-based matching or single-moment attribute mapping, and possesses stronger spatiotemporal extrapolation capabilities. The predicted microfacies reservoir state value Ψres comprehensively considers the coupled changes of semantic time jump variables and local lithological disturbance gradient ∇zLith. This enables the system to maintain prediction stability and resolution even when facing complex geological environments such as rapidly evolving zones and heterogeneous lithological zones, significantly improving its ability to express anomalous sedimentary areas.

[0045] By mapping the predicted microfacies reservoir state value Ψres to a two-dimensional geological profile, the system generates a microfacies distribution map and a reservoir probability map. Combined with the credibility index Qpred, a spatial credibility evaluation mechanism is constructed, thereby realizing the automated screening and visual labeling of mineral-rich areas within the sedimentary space. This helps to assist in the selection of mineral exploration schemes and the deployment of blocks.

[0046] This scheme not only characterizes reservoir endowment capacity based on permeability Rpe, but also introduces spatial distance to reflect spatial proximity information, thereby constructing a dynamically adjusted credibility evaluation function to support the validity judgment and screening control of prediction results. The system sets up multi-level judgment rules (preferred, alternative, not recommended) based on the microfacies reservoir state prediction value Ψres, and is supplemented by the control logic of the credibility index Qpred, supporting multi-strategy parallel prediction and block screening, providing a decision-making basis for subsequent mineral exploration deployment and exploration deployment, and improving overall work efficiency.

[0047] Example 5: This example is an explanation of Example 4. Please refer to the example provided. Figure 1 and Figure 3 Specifically: the self-feedback graph optimization module includes a graph structure variation detection and expansion triggering unit and a connectivity change evaluation and graph relearning unit; Comparison of newly added regions in microphase distribution maps and reservoir probability maps with the coverage of nodes in the three-dimensional semantic map structure G, and the expansion triggering unit of map structure variation detection and expansion triggering unit. If the newly added region is not covered by the existing node set GN or relation edge set GE in the three-dimensional semantic graph structure G, it is determined that the graph structure is insufficient. Based on new microfacies types, attribute distribution patterns, or new evolutionary paths appearing in the layer, the following are automatically added: new microfacies nodes; new reservoir attribute nodes; and new evolutionary edge relationships. Construct the supplemented three-dimensional semantic graph structure nG={GN+ΔN,GE+ΔE}; In the formula, ΔN represents the number of node additions, and ΔE represents the number of edge structure additions.

[0048] The connectivity change assessment and graph relearning unit compares the degree of change in inter-node connectivity between the three-dimensional semantic graph structure G and the supplemented three-dimensional semantic graph structure nG, and evaluates the number of inter-node paths and connectivity density in the graph by using the graph connectivity index Cconn, and calculates the connectivity change rate ΔCconn. The rate of change of connectivity ΔCconn is obtained by the following formula: ; In the formula, nCconn represents the graph connectivity index of the supplemented three-dimensional semantic graph structure, and oCconn represents the graph connectivity index of the three-dimensional semantic graph structure; Analyze the connectivity change rate ΔCconn to determine whether the three-dimensional semantic graph structure G needs to be reconstructed; The judgment method is as follows: When the rate of change in connectivity ΔCconn < 0.15, it indicates that the change in connectivity is within the normal range and there is no need to reconstruct the graph structure. When the connectivity change rate ΔCconn≥0.15, it indicates abnormal connectivity change. The graph structure is reconstructed, and the embedding vector of each node is regenerated based on the supplemented three-dimensional semantic graph structure nG.

[0049] In this embodiment, the system can actively identify structural cover defects of the three-dimensional semantic map structure G based on the newly emerging sedimentary microfacies type or reservoir attribute pattern in the prediction results, and automatically trigger the node and edge relationship supplementation mechanism, so that the map structure has dynamic update capability and can continuously adapt to new evolution paths and new occurrence types in complex geological environments.

[0050] By calculating the rate of change of connectivity ΔCconn between the newly added map structure and the original map structure, and setting a reasonable threshold for judgment, the system can accurately identify whether the changes in map structure are sufficient to affect the effectiveness of the embedded inference model, avoid prediction bias caused by the accumulation of structural changes, and thus ensure the semantic consistency and accuracy of subsequent microfacies reservoir state prediction results.

[0051] Example 6: A method for predicting and analyzing sedimentary mineral deposits based on sedimentary microfacies and knowledge graphs. Please refer to... Figure 2 Specifically, it includes the following steps: Step 1: The sedimentary evolution sequence extraction module extracts the evolution sequence of microfacies types over time from multi-source geological data and generates a sedimentary evolution structure matrix Mac. Step 2: The multidimensional reservoir attribute encoding module extracts reservoir attributes from seismic inversion, well test interpretation, and well logging analysis, and matches them with the sedimentary evolution structure matrix Mac to obtain the three-dimensional attribute tensor RS. Step 3: The semantic graph construction module establishes semantic edge relationships between nodes based on the sedimentary evolution structure matrix Mac and the three-dimensional attribute tensor RS, forming a three-dimensional semantic graph structure G; Step 4: The graph embedding and reasoning module performs embedding algorithm processing on the three-dimensional semantic graph structure G to generate the micro-phase reservoir state prediction value Ψres for each grid point in the future time. Step 5: The joint prediction block generation module maps the obtained microfacies reservoir state prediction values ​​Ψres back to the sedimentary profile to generate a microfacies distribution map, a reservoir probability map, and a confidence index Qpred. Step 6: The self-feedback map optimization module re-determines whether the three-dimensional semantic map structure G needs to be reconstructed based on the microfacies distribution map, reservoir probability map, and confidence index Qpred.

[0052] In this embodiment, by constructing a sedimentary evolution structure matrix and a three-dimensional attribute tensor, the evolutionary timeline of sedimentary microfacies and the multidimensional attributes of the reservoir are uniformly encoded in a spatiotemporal manner, significantly enhancing the information organization of sedimentary geological processes. Compared to traditional methods based solely on two-dimensional profile diagrams or static lithological distributions, this approach organizes geological data into structured map nodes and semantic edge relationships, constructing a three-dimensional semantic map structure suitable for inference. This structure not only preserves the dynamic evolutionary characteristics of sedimentary facies changes but also precisely couples continuous attributes such as porosity, permeability, and reservoir thickness with microfacies types, improving the dimensionality of expression for complex geological evolution relationships.

[0053] This method uses inference functions to predict the microfacies reservoir state at each grid location at future time points, thereby enabling the spatial division of preferred mining areas, candidate mining areas, and unrecommended areas. Furthermore, a credibility index, Qpred, is constructed by combining microfacies prediction values ​​and reservoir permeability, providing a verification mechanism for the prediction results and enhancing the decision-making credibility of the predicted blocks. Compared to traditional methods that only interpolate and extrapolate on known data points or rely on rule-based judgments, this method, supported by a geographic map, can perform forward-looking predictions, semantic relationship inference, and result credibility measurement, improving the foresight and refinement of the analysis.

[0054] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graphs, characterized in that: It includes a sedimentary evolution sequence extraction module, a multi-dimensional reservoir attribute encoding module, a semantic map construction module, a map embedding and reasoning module, a joint prediction block generation module, and a self-feedback map optimization module; The sedimentary evolution sequence extraction module extracts the evolution sequence of microfacies types over time from multi-source geological data and generates a sedimentary evolution structure matrix; The multidimensional reservoir attribute encoding module extracts reservoir attributes from seismic inversion, well test interpretation, and well logging analysis, and matches them with the sedimentary evolution structure matrix to obtain three-dimensional attribute tensors; The semantic graph construction module establishes semantic edge relationships between nodes based on the sedimentary evolution structure matrix and the three-dimensional attribute tensor, forming a three-dimensional semantic graph structure. The graph embedding and reasoning module processes the 3D semantic graph structure using an embedding algorithm to generate a predicted value of the micro-phase reservoir state for each grid point at a future time. The joint prediction block generation module maps the acquired microfacies reservoir state prediction values ​​back to the sedimentary profile to generate microfacies distribution maps, reservoir probability maps, and confidence indices. The self-feedback-based map optimization module determines whether the three-dimensional semantic map structure needs to be reconstructed based on the microfacies distribution map, reservoir probability map, and confidence index.

2. The sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graph as described in claim 1, characterized in that: The sedimentary evolution sequence extraction module includes a sedimentary microfacies identification and temporal calibration unit and a sedimentary sequence normalization processing unit; The sedimentary microfacies identification and temporal calibration unit is based on the spatial sequence and stratigraphic division results of geological data sources to obtain the sedimentary microfacies type fac and the time interval Δt between adjacent microfacies for each sedimentary period; The sedimentary microfacies type (FAC) is obtained by: based on the profile map, well logging curves and thin section lithological description, and by using the time slice mapping function to identify each geological layer as a sedimentary microfacies type (FAC), including the littoral sandbar microfacies, the floodplain muddy microfacies, the underwater distributary mouth microfacies and the muddy infill microfacies. The formula for the time slice mapping function is: fac(ti) = Fmap(Litho(ti), Curve(ti), Seq(ti)); In the formula, fac(ti) represents the sedimentary microfacies type at time point ti, Fmap represents the mapping function that outputs the microfacies type according to the matching rules of lithology-curve-sequence, Litho(ti) represents the lithology category at time point ti, Curve(ti) represents the logging curve shape at time point ti, and Seq(ti) represents the sequence position at time point ti. The time interval Δt between adjacent microfacies is calculated using a chronological sequence based on the time interval between sedimentary facies. The sedimentary sequence normalization processing unit performs outlier detection and normalization on the obtained sedimentary microfacies type fac and the time interval Δt between adjacent microfacies to generate the sedimentary evolution structure matrix Mac; Outlier detection uses a deposition time density gradient function to process outliers in the time interval Δt between adjacent microfacies. When the time interval Δt between adjacent microfacies is greater than a preset time threshold, it indicates the presence of geologically unreasonable isolated points, which are then cleaned up. The normalization process performs nonlinear normalization on the time interval Δt between adjacent microphases using a time scale transformation function; The processing method is as follows: For the time interval Δt between adjacent sedimentary microfacies, first take the square root to obtain its square root value; then divide this square root value by the maximum value among all the square roots of the time intervals; obtain the normalized time interval Δt between adjacent microfacies. The sedimentary evolution structure matrix Mac is obtained using the following formula: ； In the formula, tn represents the nth time.

3. The sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graph as described in claim 2, characterized in that: The multidimensional reservoir attribute coding module includes a reservoir physical property parameter extraction and standardization unit and an attribute and sedimentary evolution structure matching and tensor generation unit; The reservoir property parameter extraction and standardization unit extracts reservoir attributes, including porosity Rpor, permeability Rpe, and reservoir thickness Rd, from seismic inversion, well test interpretation, and well logging analysis, and then fits them to obtain an initial three-dimensional attribute volume. The porosity Rpor is obtained as follows: inversion is performed from pre-stack seismic data to extract P-wave impedance data; curve fitting is performed based on the true porosity and the impedance value of the corresponding depth segment; the impedance data of the entire profile is substituted into the transformation relationship to calculate the continuous spatial porosity Rpor. The permeability Rpe is obtained as follows: First, obtain the well test curves and interpretation reports of the target well and adjacent areas, extract key well test parameters, including productivity, pressure and volume factor; then, use the well test theory formula to back-calculate the permeability. The reservoir thickness Rd is obtained by preprocessing the logging data in the well logging analysis, identifying the reservoir development section, analyzing each well section by section, identifying the effective sandstone section that meets the conditions, and finally performing vertical integration to obtain the total thickness. The attribute matching and tensor generation unit matches the acquired initial three-dimensional attribute volume with the sedimentary evolution structure matrix Mac; within the time window, for each time point ti, the corresponding attribute value in the initial three-dimensional attribute volume is searched. When there is an inconsistency in sampling frequency, the time points adjacent to the time point ti are used for representation; when multiple points fall within the time window, a weighted average is calculated, with the weights inversely proportional to the time distance. For multiple attribute values ​​existing in the time window, for each attribute value at time point t, calculate the time difference with time point ti, and use 1 / |t-ti| as the weight to calculate the weighted porosity Rpor, permeability Rpe and reservoir thickness Rd. The weighted porosity Rpor, permeability Rpe, and reservoir thickness Rd are fitted with the sedimentary microfacies type fac to obtain the three-dimensional attribute tensor RS.

4. The sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graph as described in claim 3, characterized in that: The semantic graph construction module includes a graph node and edge type definition unit and an edge weight quantization and graph structure construction unit; The graph node and edge type definition unit uniquely identifies the sedimentary microfacies type fac in the sedimentary evolution structure matrix Mac and converts it into a semantic node: Porosity Rpor, permeability Rpe, and reservoir thickness Rd in the three-dimensional attribute tensor RS are used as reservoir attribute nodes; then, timestamp nodes are constructed using time points ti to connect different time levels. By transforming the sedimentary evolution structure matrix Mac and the three-dimensional attribute tensor RS, three types of semantic edges are established, including evolutionary edge Eu→u, coupling edge Eu→R, and temporal edge Et→R. Among them, the evolution edge Eu→u represents the evolutionary transition between microfacies, which comes from the sedimentary evolution structure matrix Mac and the sequence of microfacies types ordered by time. Each pair of adjacent sedimentary microfacies types fac(ti) and fac(ti+1) constitute a candidate evolution edge. The coupling edge Eu→R represents the co-occurrence relationship between microfacies and reservoir properties, which is derived from the sedimentary evolution structure matrix Mac and the three-dimensional property tensor RS. According to the time point ti, the sedimentary microfacies type fac is matched with the property values ​​porosity Rpor, permeability Rpe and reservoir thickness Rd at the same time point. The temporal edge Et→R represents the evolution trend of the attribute over time, which comes from the attribute time series in the three-dimensional attribute tensor RS. That is, under the condition of fixed spatial location, the attribute values ​​at different times constitute the time series. Using sedimentary microfacies type fac, time point ti, porosity Rpor, permeability Rpe, and reservoir thickness Rd as nodes, establish a node set GN={fac(ti), Rpor, Rpe, Rd, ti}; Using the evolutionary edge Eu→u, the coupling edge Eu→R, and the temporal edge Et→R as relation edges, establish the relation edge set GE={Eu→u, Eu→R, Et→R}.

5. The sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graph as described in claim 4, characterized in that: The edge weight quantization and graph structure construction unit quantifies each type of edge, constructs edge weight functions, including microfacies evolution edge weight Wu→u, microfacies and reservoir attribute coupling edge weight Wu→R, and time and reservoir attribute evolution edge weight Wt→R, and embeds them into the graph structure. By quantizing the evolution edge Eu→u, the micro-phase evolution edge weight Wu→u is obtained; The microfacies evolution boundary weight Wu→u is obtained as follows: First, take the time interval between adjacent sedimentary microfacies types fac(ti) and fac(ti+1), and then take the reciprocal of this value to represent the phase transition rate; then, calculate the evolution frequency from fac(ti) to fac(ti+1); finally, multiply the evolution frequency by the phase transition rate to obtain the microfacies evolution boundary weight Wu→u. By quantizing the coupling edge Eu→R, the coupling edge weight Wu→R between microfacies and reservoir properties is obtained. The method for obtaining the coupling weight Wu→R between microfacies and reservoir properties is as follows: First, the porosity Rpor and permeability Rpe at time point ti are added together to obtain the reservoir occurrence quality; then, the lithological change gradient in the vertical direction at the current location is extracted, its absolute value is taken, 1 is added, and then the square root is taken as the denominator to represent the degree of formation disturbance; finally, the reservoir occurrence quality is divided by the degree of formation disturbance to obtain the coupling weight Wu→R between microfacies and reservoir properties. By quantizing the temporal edge Et→R, the time-reservoir attribute evolution edge weight Wt→R is obtained; The method for obtaining the time-reservoir attribute evolution boundary weight Wt→R is as follows: First, select a time window centered on time point ti and extract the attribute value sequence within the window; then, calculate the standard deviation of the attribute values; finally, add a constant 1 to the standard deviation and take the reciprocal to obtain the time-reservoir attribute evolution boundary weight Wt→R. The obtained node set GN={fac(ti), Rpor, Rpe, Rd, ti} is combined with the quantized relation edge set GE={Eu→u:Wu→u, Eu→R:Wu→R, Et→R:Wt→R} to obtain the three-dimensional semantic graph structure G={GN, GE}.

6. The sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graph as described in claim 5, characterized in that: The graph embedding reasoning module processes the three-dimensional semantic graph structure G using an embedding algorithm, transforms the node and edge relationships into vector form, unifies the embedding dimension to d, and outputs the embedding vector. Based on the embedded vector, a microfacies prediction function is constructed to predict the microfacies reservoir state at grid points at a future time point t+Δt, and the predicted value Ψres of the microfacies reservoir state is obtained. The predicted value Ψres of the microfacies reservoir state is obtained by the following formula: ； In the formula, Ψres(x, y, t+Δt) represents the predicted value of the microfacies reservoir state at grid point (x, y) at time point t+Δt, vfac(ti) represents the sedimentary microfacies type embedding vector at time point ti, vRpor(ti) represents the porosity embedding vector at time point ti, vRpe(ti) represents the permeability embedding vector at time point ti, vti+1-vti represents the semantic transition intensity of the time vector in the future, and ∇zLith(x, y) represents the lithological change gradient at grid point (x, y); The obtained microfacies reservoir state prediction values ​​Ψres are analyzed to determine the quality of the mineral deposits. The judgment method is as follows: When the predicted value of microfacies reservoir state Ψres > 0.75, it indicates a preferred mineral deposit area; When 0.75 ≥ the predicted value of microfacies reservoir state Ψres > 0.55, it indicates a candidate mineral deposit area; When 0.55 ≥ the predicted value Ψres of the microfacies reservoir state, it indicates that the mineral deposit area is not recommended.

7. The sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graph as described in claim 6, characterized in that: The joint prediction block generation module maps each microfacies reservoir state prediction value Ψres to a two-dimensional geological profile, locates it to a specific grid point (x, y), and constructs a microfacies distribution map and a reservoir probability map. The microfacies distribution map selects the corresponding microfacies type label based on the microfacies reservoir state prediction value Ψres; The reservoir probability map maps the probability distribution of reservoir potential based on the microfacies reservoir state prediction value Ψres. The predicted value of microfacies reservoir state Ψres is combined with the permeability Rpe to calculate the confidence index Qpred, which is then compared with the preset confidence threshold Tred to determine the confidence of the layer. The credibility index Qpred is obtained using the following formula: ； In the formula, Qpred(x,y) represents the confidence index at grid point (x,y), Rpe(x,y,t) represents the permeability at grid point (x,y) at time t, log represents the logarithmic function, and dfac(x,y) represents the spatial distance from grid point (x,y) to the center of the nearest known stable microphase region. The credibility of a layer is obtained through matching methods: When the confidence index Qpred ≥ the confidence threshold Tred, it indicates that the confidence is high and the microfacies reservoir state prediction value Ψres is effective; When the confidence index Qpred is less than the confidence threshold Tred, it indicates low confidence and the microfacies reservoir state prediction value Ψres is invalid.

8. The sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graph as described in claim 7, characterized in that: The self-feedback graph optimization module includes a graph structure variation detection and expansion triggering unit and a connectivity change evaluation and graph relearning unit. Comparison of newly added regions in microphase distribution maps and reservoir probability maps with the coverage of nodes in the three-dimensional semantic map structure G, and the expansion triggering unit of map structure variation detection and expansion triggering unit. If the newly added region is not covered by the existing node set GN or relation edge set GE in the three-dimensional semantic graph structure G, it is determined that the graph structure is insufficient. Based on new microfacies types, attribute distribution patterns, or new evolutionary paths appearing in the layer, the following are automatically added: new microfacies nodes; new reservoir attribute nodes; and new evolutionary edge relationships. Construct the supplemented three-dimensional semantic graph structure nG={GN+ΔN,GE+ΔE}; In the formula, ΔN represents the number of node additions, and ΔE represents the number of edge structure additions.

9. The sedimentary mineral prediction and analysis system based on sedimentary microfacies and knowledge graph as described in claim 8, characterized in that: The connectivity change assessment and graph relearning unit compares the degree of change in inter-node connectivity between the three-dimensional semantic graph structure G and the supplemented three-dimensional semantic graph structure nG, and evaluates the number of inter-node paths and connectivity density in the graph by using the graph connectivity index Cconn, and calculates the connectivity change rate ΔCconn. The rate of change of connectivity ΔCconn is obtained by the following formula: ； In the formula, nCconn represents the graph connectivity index of the supplemented three-dimensional semantic graph structure, and oCconn represents the graph connectivity index of the three-dimensional semantic graph structure; Analyze the connectivity change rate ΔCconn to determine whether the three-dimensional semantic graph structure G needs to be reconstructed; The judgment method is as follows: When the rate of change in connectivity ΔCconn < 0.15, it indicates that the change in connectivity is within the normal range and there is no need to reconstruct the graph structure. When the connectivity change rate ΔCconn≥0.15, it indicates abnormal connectivity change. The graph structure is reconstructed, and the embedding vector of each node is regenerated based on the supplemented three-dimensional semantic graph structure nG.

10. A sedimentary mineral resource prediction and analysis method based on sedimentary microfacies and knowledge graph, applied to the sedimentary mineral resource prediction and analysis system based on sedimentary microfacies and knowledge graph as described in any one of claims 1 to 9, characterized in that: Includes the following steps: Step 1: The sedimentary evolution sequence extraction module extracts the evolution sequence of microfacies types over time from multi-source geological data and generates a sedimentary evolution structure matrix Mac. Step 2: The multidimensional reservoir attribute encoding module extracts reservoir attributes from seismic inversion, well test interpretation, and well logging analysis, and matches them with the sedimentary evolution structure matrix Mac to obtain the three-dimensional attribute tensor RS. Step 3: The semantic graph construction module establishes semantic edge relationships between nodes based on the sedimentary evolution structure matrix Mac and the three-dimensional attribute tensor RS, forming a three-dimensional semantic graph structure G; Step 4: The graph embedding and reasoning module performs embedding algorithm processing on the three-dimensional semantic graph structure G to generate the micro-phase reservoir state prediction value Ψres for each grid point in the future time. Step 5: The joint prediction block generation module maps the obtained microfacies reservoir state prediction values ​​Ψres back to the sedimentary profile to generate a microfacies distribution map, a reservoir probability map, and a confidence index Qpred. Step 6: The self-feedback map optimization module re-determines whether the three-dimensional semantic map structure G needs to be reconstructed based on the microfacies distribution map, reservoir probability map, and confidence index Qpred.

Images