A method and system for three-dimensional seismic exploration prediction analysis

Abstract

The application discloses a three-dimensional seismic exploration prediction analysis method and system, and relates to the field of three-dimensional seismic exploration.The method comprises data preparation and preprocessing; construction of interpretation and establishment of a geological framework; well-seismic joint analysis and construction of a prior model; construction of a prior probability three-dimensional geological model through Markov chain simulation of lithofacies, Monte Carlo sampling to give elastic parameters, and generation of an adversarial network to expand pseudo-well samples; multi-source information fusion and probabilistic reservoir prediction; extraction of multi-evidence volumes and determination of weights; multi-target threshold decision and delineation of a favorable zone; high-precision target appearance and well site deployment.The method realizes logical judgment before appearance through prior probability modeling and multi-evidence weighted fusion, outputs a three-dimensional probability volume to quantify uncertainty, objectively delineates a favorable zone through multi-target decision, significantly improves reservoir prediction accuracy and micro-geological body recognition capability, and reduces exploration risk.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional seismic exploration, and in particular to a three-dimensional seismic exploration prediction and analysis method and system. Background Technology

[0002] Existing 3D seismic exploration reservoir prediction methods typically follow this technical process: First, data preparation and preprocessing are performed. Raw seismic data is acquired and static correction, noise attenuation, amplitude compensation, deconvolution, velocity analysis, and migration imaging are executed to obtain pre-stack and post-stack 3D seismic data volumes. Simultaneously, well logging data undergoes environmental correction and normalization. Next, structural interpretation and geological framework establishment are conducted. Target stratigraphic reflection characteristics are determined through well-seismic calibration, major stratigraphic interfaces and faults are traced and interpreted, and coherence volume or variance volume techniques are used to assist fault identification, constructing a 3D structural model. Based on this, well-seismic joint analysis and reservoir prediction are carried out. Seismic attributes such as amplitude and frequency are extracted from the seismic data for attribute analysis. Elastic parameter volumes such as wave impedance are obtained through post-stack or pre-stack inversion. The attribute volumes and inversion results are interpreted in conjunction with drilling data to qualitatively or semi-quantitatively predict the reservoir distribution range. Finally, favorable zones are delineated and well locations are deployed based on the prediction results. Interpreters select appropriate thresholds based on experience to delineate favorable areas, which are then displayed and well locations are optimized on a 3D visualization platform.

[0003] In the aforementioned existing methods, the essence of the predictive analysis stage is to first observe and then speculate. That is, first, geometric bodies with channel morphology or anomalous reflection characteristics are identified on the seismic profile, and then attribute analysis and inversion are used to infer whether the anomaly is a high-quality reservoir. This leads to a lack of one-to-one correspondence between anomalous reflection characteristics such as amplitude anomalies and coherence anomalies and reservoir development. Non-reservoir geological bodies such as mudstone, coal seams, and lithological pinch-outs can also produce similar seismic responses, resulting in many cases where observations are made but guesses are incorrect. At the same time, this method has limited ability to identify small geological bodies such as thin reservoirs, narrow channels, and small faults, and the prediction results heavily depend on the experience of the interpreters, with different interpreters potentially reaching significantly different conclusions. In view of the above, this invention provides a three-dimensional seismic exploration predictive analysis method and system to solve the above problems. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a three-dimensional seismic exploration prediction and analysis method and system. It solves the problem that existing methods essentially rely on observation followed by speculation in the prediction and analysis stage. This involves first identifying geometric bodies with channel morphology or anomalous reflection characteristics on seismic profiles, and then combining attribute analysis and inversion to infer whether the anomaly is a high-quality reservoir. This leads to a non-one-to-one correspondence between anomalous reflection characteristics such as amplitude anomalies and coherence anomalies and reservoir development. Non-reservoir geological bodies such as mudstone, coal seams, and lithological pinch-outs can also produce similar seismic responses, resulting in numerous cases of misjudgments despite observation. Furthermore, this method has limited ability to identify small geological bodies such as thin reservoirs, narrow channels, and small faults, and the prediction results heavily rely on the experience of interpreters, potentially leading to significantly different conclusions from different interpreters.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a three-dimensional seismic exploration prediction and analysis method and system, the method comprising: S1. Data preparation and preprocessing: Obtain the original 3D seismic data and drilling and logging data of the target work area, and generate a 3D seismic data volume suitable for subsequent interpretation and prediction through conventional processing procedures; S2. Structural Interpretation and Geological Framework Establishment: Using the post-stack seismic data and well logging data output in step S1, identify the main subsurface stratigraphic interfaces and faults, and construct a three-dimensional structural model; S3. Well-seismic joint analysis and prior model construction: Based on the limited well data and regional geological understanding in steps S1 and S2, the lithofacies sequence is simulated by Markov chain, elastic parameters are assigned by Monte Carlo sampling, and the pseudo-well samples are expanded by generative adversarial network. Finally, a three-dimensional geological model containing prior probabilities of lithology and physical properties is constructed. S4. Multi-source information fusion and probabilistic reservoir prediction: Extract conventional attributes, elastic parameters and five-dimensional anisotropy from the seismic data in the first three steps. Determine the weight of each piece of evidence through distance matching, sensitivity analysis and information entropy calculation. Use evidence theory to fuse the weighted evidence with the prior model and output a three-dimensional probabilistic body of lithology, physical properties and hydrocarbon content. S5. Multi-objective threshold decision and favorable zone delineation: For the three-dimensional probability volume output in step S4, multiple decision objectives are defined, and a multi-objective decision method is used to objectively determine the optimal probability threshold, thereby extracting the favorable zone with high confidence. S6. High-precision target visualization and well location deployment: Within the favorable zone determined in step S5, targeted attribute fusion and three-dimensional visualization are performed, and the most favorable drilling location is selected by combining geological, engineering and economic factors.

[0006] Preferably, in step S1, the main acquisition target is the original 3D seismic data of the work area, which is recorded by field shot collection. At the same time, existing drilling and logging data in the work area are collected, including lithological logging, logging curves with natural gamma, sonic transit time, density, resistivity, etc., as well as stratification data. Environmental corrections are performed on the logging curves, such as wellbore correction, mud invasion correction, and normalization processing, to eliminate systematic errors caused by different instruments and measurement conditions, and to obtain standardized curves that can be used for calibration and rock physical analysis. Finally, the processed pre-stack 3D seismic data volume, including offset and azimuth information, post-stack 3D seismic data volume, and standardized logging curves are output.

[0007] Preferably, the specific process in step S2 is as follows: S21. Well-seismic calibration: Using the standardized well logging curves obtained in step S1, a synthetic seismic record is generated. By comparing the synthetic record with the seismic traces near the well, the seismic reflection characteristics of the target layer are determined, and the time-depth calibration is completed. S22. Stratigraphic Interpretation: Stratigraphic interpretation is performed on the 3D seismic data volume along the calibrated reflection phase axis. A combination of automatic tracing and manual correction is used to obtain the stratigraphic grid of the main target layer and key marker layers. S23. Fault Interpretation: Perform coherence volume or variance volume attribute calculations on seismic data to enhance fault display, identify fault traces on the attribute volume, combine and close faults in conjunction with seismic profiles, and construct three-dimensional fault models by picking fault planes. S24. Construction of 3D Structural Model: Integrate the interpreted layer mesh and fault model to establish a closed 3D structural model.

[0008] Preferably, the specific process of simulating lithofacies sequences based on Markov chains in step S3 is as follows: input the well logging lithofacies interpretation results under the constraints of the structural model in step S2, statistically analyze the vertical transition probabilities of each lithofacies to construct a Markov chain model, and gradually generate the lithofacies sequence of virtual wells based on the transition probabilities from the initial lithofacies using a random simulation method. Repeat this process multiple times to obtain a large number of virtual well lithofacies columns that conform to the laws of geological statistics. During the simulation, the probability distribution of lithofacies thickness is considered so that the generated sequence has reasonable layer thickness characteristics. Subsequently, a rock physics template is established based on the well logging data. The probability distribution of elastic parameters for each type of rock facies is statistically analyzed and the correlation between parameters is analyzed. For each generated virtual rock facies, specific elastic parameter values ​​are randomly sampled from the corresponding probability distribution to generate a pseudo-well dataset containing depth, lithology, and elastic parameters. The specific steps of optimizing and constructing the 3D geological model using the generative adversarial network in step S3 are as follows: A generative adversarial network is constructed using the real well data and pseudo-well data as training samples. The generator is responsible for generating pseudo-seismic records, and the discriminator is responsible for distinguishing between generated and real data. An improved generative adversarial network architecture is adopted, and the network is forced to learn the real distribution of data through gradient penalty and latent variable constraints to generate high-fidelity and diverse pseudo-seismic records, further expanding the training sample library. Subsequently, the real and pseudo-well data are combined with the 3D structural model, and geostatistical interpolation methods are used to interpolate the entire work area. During the interpolation process, the structural model is used as the trend surface, and the well point data is used as the hard data to generate the lithological probability prior distribution and physical property parameter prior distribution for each grid node. Finally, the prior model is output as a 3D probability field.

[0009] Preferably, in step S4, before outputting the three-dimensional probability volume, it is necessary to extract various evidence bodies characterizing lithology, physical properties, and fluids, and determine the evidence weight of these evidence bodies. The specific process is as follows: S41. Extraction of multiple evidence sources: Extract conventional seismic attribute volumes from the post-stack seismic data volume output in step 1, including root mean square amplitude, instantaneous frequency, coherence volume, curvature volume, spectral decomposition tuning volume, etc., and perform pre-stack inversion from the pre-stack seismic data volume to obtain elastic parameter volumes, including P-wave impedance, S-wave impedance, P-wave / S-wave velocity ratio, Poisson's ratio, etc. If the work area has wide azimuth seismic data, extract five-dimensional seismic attribute volumes, including anisotropic gradient of amplitude with azimuth, fracture density, anisotropic intensity, etc., and perform statistical analysis on a certain attribute value and lithology at the well point to establish a mapping function of lithology probability with attribute value, and apply the mapping function to the entire work area to generate lithology probability attribute volumes; S42. Determining Evidence Weights: Each piece of evidence is assigned a weight coefficient, which represents the reliability and sensitivity of the evidence in reservoir prediction. The weight determination method includes at least one of the following: if the value of an evidence piece matches the standard template of a high-quality reservoir closely, the evidence piece is considered to have a higher weight; if rock physics forward modeling shows that an evidence piece is sensitive to changes in the target parameter, the evidence piece is considered to have a higher weight; if the information entropy of an evidence piece is low across the entire work area, indicating that it provides a large amount of information, the evidence piece is considered to have a higher weight; if multiple experts give consistent evaluations of the importance of an evidence piece and the scores are high, its subjective weight is calculated using the analytic hierarchy process (AHP); the final weight is a combination of the above methods and normalized. S43. Evidence Theory Fusion Reasoning: For each grid node in step S42, the numerical values ​​and weights of the evidence bodies are transformed into basic probability allocation functions. During the fusion process, the prior geological model obtained in step three serves as the basic evidence body for fusion, and its weight is set according to the confidence level of the prior model. The evidence theory combination rules are used to fuse the basic probability allocations of multiple evidence bodies. If the evidence bodies point in the same direction, the probability allocation value of the target category increases after fusion. If the evidence bodies point in conflict, the conflict coefficient increases. The fusion result reflects uncertainty. The probability allocation value of each underground grid node belonging to different reservoir categories is calculated, and finally three probability bodies are output: three-dimensional lithology probability body, three-dimensional physical property probability body, and three-dimensional hydrocarbon probability body.

[0010] Preferably, in step S5, the three-dimensional lithology probability volume, three-dimensional physical property probability volume, and three-dimensional hydrocarbon probability volume output in step S4 are input. By defining three quantitative indicators—boundary uncertainty, misclassification risk, and trap size—a decision matrix under different probability thresholds is constructed. Subsequently, the entropy weight method is used to automatically assign weights based on the degree of variation of the indicators under each threshold, eliminating subjective bias. The TOPSIS method is used to calculate the closeness of each threshold to the ideal scheme, and the one with the highest closeness is selected as the optimal probability threshold. This threshold is applied to perform binary segmentation and spatial connectivity analysis on the probability volume, and finally, high-confidence three-dimensional favorable zones are extracted, realizing the objective transformation from continuous probability to discrete target.

[0011] Preferably, the specific steps in step S6 are as follows: S61. Sensitive attribute fusion under probabilistic constraints: Select several seismic attributes that are sensitive to the heterogeneity of the reservoir only within the favorable zone. If the correlation coefficient of a certain attribute with the known reservoir quality is high, then the attribute is determined to have a larger weight during fusion. The selected attributes are fused using weighted linear superposition or evidence theory to generate a high-resolution reservoir fusion attribute body. This attribute body further distinguishes high-quality parts within the favorable zone. S62. 3D Visualization and Confidence Display: The 3D construction model from step S2, the 3D probability volume from step S4, the favorable zone from step S5, and the fused attribute volume generated in step S6 are loaded into a 3D visualization platform. To intuitively distinguish between high and low probabilities, a color mapping rule is set for the probability volume: if the probability value of a grid point is higher than a threshold, it is displayed in a warm color; if it is lower than the threshold, it is displayed in a cool color. In addition, to intuitively represent probability uncertainty, if the probability value of a region fluctuates around the threshold, the transparency of that region is increased. To intuitively represent the quality of the reservoir, a color mapping rule is set for the fused attribute volume: if the attribute value reflects good reservoir quality, it is displayed with high brightness or high saturation; if the quality is poor, it is displayed with low brightness or low saturation. S63. Well Location Optimization: In a 3D visualization environment, a comprehensive evaluation is conducted on locations within favorable zones. If the selected location is located in a high-probability core area and is far from fault fracture zones, the geological conditions are considered favorable. If the surface conditions at the selected location are suitable and the drilling trajectory is feasible, the engineering conditions are considered favorable. If the reserve estimation results at the selected location are good and the development costs are controllable, the economic conditions are considered favorable. Locations that simultaneously meet the geological, engineering, and economic favorable conditions are determined as preferred well locations, and the well location coordinates, design well depth, and geological recommendations are output to form a well location deployment plan.

[0012] Preferably, the system includes: Module 1: Data Preprocessing and Construction Modeling Module, including steps S1 and S2, is used to receive raw seismic and drilling data, perform routine seismic data processing and well logging preprocessing, complete well-seismic calibration, layer and fault interpretation, and construct a three-dimensional structural model to provide basic data and spatial constraints for subsequent analysis. Module 2: Prior geological model construction module, including step S3, which is used to construct a Markov chain model based on well logging lithofacies statistics to generate a virtual well, assign elastic parameters through Monte Carlo sampling, and construct a three-dimensional geological model containing lithological and physical property probabilistic priors by combining real well data with geostatistical interpolation. Module 3: Multi-source information fusion and probability prediction module, including step S4, which is used to extract multiple evidence bodies from seismic data and determine weights, use evidence theory to fuse weighted evidence with prior models, calculate reservoir category probability for each underground point, and output three-dimensional lithology, physical properties and hydrocarbon probability bodies; Module 4: Decision Analysis and Well Location Deployment Module, including steps S5 and S6, is used to define multiple objectives and solve for the optimal probability threshold, extract favorable zones, integrate sensitive attributes within the zone to generate reservoir fusion bodies, optimize well locations through three-dimensional visualization of comprehensive geological, engineering and economic conditions, and output deployment schemes.

[0013] The technical effects and advantages of this invention are as follows: This three-dimensional seismic exploration prediction and analysis method and system abandons the traditional technical logic of seeing first and then guessing, and constructs a prediction framework of logical judgment first and then manifestation: through step S3, a three-dimensional geological model containing the prior probability of lithology and physical properties is constructed. Combined with multi-source information fusion and probabilistic reasoning in step S4, the prediction process changes from passively identifying seismic anomalies to actively judging the probability of reservoir development. This method effectively solves the problem of the ambiguous correspondence between anomalous reflection characteristics and reservoir development in traditional technology, avoids misjudgment caused by similar seismic responses of non-reservoir geological bodies such as mudstone and coal seams, and significantly reduces the exploration risk of seeing but guessing wrong. This three-dimensional seismic exploration prediction and analysis method and system establishes a multi-evidence weighted fusion mechanism, overcoming the shortcomings of traditional methods that rely on single attributes and subjective selection. In step S4, objective weights are assigned to each evidence body through distance matching, sensitivity analysis, and information entropy calculation. The weighted evidence is fused with the prior model using evidence theory, and a three-dimensional lithology, physical properties, and hydrocarbon probability body is output. This method quantifies the differences in the contribution of different seismic attributes to reservoir prediction. When the evidence points to the same direction, the target probability increases. When the directions conflict, the fusion result reflects uncertainty. This realizes the leap from deterministic numerical values ​​to probabilistic expressions in prediction results, providing a quantitative risk basis for decision-making. This three-dimensional seismic exploration prediction and analysis method and system adopts a multi-objective threshold decision-making method, which solves the drawback of traditional favorable zone delineation relying on human experience. In step S5, three quantitative indicators are defined: boundary uncertainty, misclassification risk, and trap size. The weights of the indicators are automatically allocated by the entropy weight method, and the optimal probability threshold is objectively solved by the TOPSIS method, eliminating the difference in conclusions caused by different interpreters subjectively selecting thresholds. In step S6, sensitive attribute fusion and three-dimensional visualization are performed in the favorable zone. Differentiated display rules are set for probability level, uncertainty magnitude, and reservoir quality. Finally, the well location is optimized by comprehensively considering geological, engineering, and economic conditions, which significantly improves the identification accuracy and exploration success rate of small geological bodies such as thin reservoirs and narrow channels. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart illustrating the overall process of this invention; Figure 2 This is the logic diagram for the present invention; Figure 3 This is the overall system architecture diagram of the present invention. Detailed Implementation

[0016] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0018] This invention discloses a three-dimensional seismic exploration prediction and analysis method and system, according to the appendix. Figure 1 As shown, the predictive analysis methods include: S1. Data preparation and preprocessing: Obtain the original 3D seismic data and drilling and logging data of the target work area, and generate a 3D seismic data volume suitable for subsequent interpretation and prediction through conventional processing procedures; S2. Structural Interpretation and Geological Framework Establishment: Using the post-stack seismic data and well logging data output from step S1, identify the main subsurface stratigraphic interfaces and faults, and construct a three-dimensional structural model; S3. Well-seismic joint analysis and prior model construction: Based on the limited well data and regional geological knowledge in steps S1 and S2, the lithofacies sequence is simulated by Markov chain, elastic parameters are assigned by Monte Carlo sampling, and the pseudo-well samples are expanded by generative adversarial network. Finally, a three-dimensional geological model containing prior probabilities of lithology and physical properties is constructed. S4. Multi-source information fusion and probabilistic reservoir prediction: Extract conventional attributes, elastic parameters and five-dimensional anisotropy from the seismic data in the first three steps. Determine the weight of each piece of evidence through distance matching, sensitivity analysis and information entropy calculation. Use evidence theory to fuse the weighted evidence with the prior model and output three-dimensional lithology, physical properties and hydrocarbon probability. S5. Multi-objective threshold decision and favorable zone delineation: For the three-dimensional probability volume output in step S4, multiple decision objectives are defined, and the optimal probability threshold is objectively determined by the multi-objective decision method, thereby extracting the favorable zone with high confidence. S6. High-precision target visualization and well location deployment: Within the favorable zone determined in step S5, targeted attribute fusion and three-dimensional visualization are performed, and the most favorable drilling location is selected by combining geological, engineering and economic factors.

[0019] According to the appendix Figure 1 As shown, further, in step S1, the main acquisition target is the original 3D seismic data of the work area, which is recorded by field shot collection. At the same time, existing drilling and logging data in the work area are collected, including lithological logging, logging curves with natural gamma, sonic transit time, density, resistivity, etc., as well as layer data. Environmental corrections are performed on the logging curves, such as well diameter correction, mud invasion correction and normalization processing, to eliminate systematic errors caused by different instruments and measurement conditions, and to obtain standardized curves that can be used for calibration and rock physical analysis. Finally, the processed pre-stack 3D seismic data volume containing offset and azimuth information, the post-stack 3D seismic data volume, and the standardized logging curves are output.

[0020] According to the appendix Figure 1 As shown, the specific process in step S2 is as follows: S21. Well-seismic calibration: Using the standardized well logging curves obtained in step S1, a synthetic seismic record is generated. By comparing the synthetic record with the seismic traces near the well, the seismic reflection characteristics of the target layer are determined, and the time-depth calibration is completed. S22. Stratigraphic Interpretation: Stratigraphic interpretation is performed on the 3D seismic data volume along the calibrated reflection phase axis. A combination of automatic tracing and manual correction is used to obtain the stratigraphic grid of the main target layer and key marker layers. S23. Fault Interpretation: Perform coherence volume or variance volume attribute calculations on seismic data to enhance fault display, identify fault traces on the attribute volume, combine and close faults in conjunction with seismic profiles, and construct three-dimensional fault models by picking fault planes. S24. Construction of 3D Structural Model: Integrate the interpreted layer mesh and fault model to establish a closed 3D structural model.

[0021] According to the appendix Figure 1 As shown, the specific process of simulating lithofacies sequences based on Markov chains in step S3 is as follows: input the well logging lithofacies interpretation results under the constraints of the structural model in step S2, statistically analyze the vertical transition probabilities of each lithofacies to construct a Markov chain model, and generate the lithofacies sequence of virtual wells step by step according to the transition probabilities from the initial lithofacies using a stochastic simulation method. Repeat this process multiple times to obtain a large number of virtual well lithofacies columns that conform to the laws of geological statistics. During the simulation, the probability distribution of lithofacies thickness is considered to make the generated sequence have reasonable layer thickness characteristics. Subsequently, a rock physics template was established based on the logging data. The probability distribution of elastic parameters for each rock facies was statistically analyzed and the correlation between parameters was analyzed. For each generated virtual rock facies, specific elastic parameter values ​​were randomly sampled from the corresponding probability distribution to generate a pseudo-well dataset containing depth, lithology, and elastic parameters. The specific steps of optimizing the generative adversarial network and constructing the 3D geological model in step S3 are as follows: a generative adversarial network is constructed using real well data and pseudo-well data as training samples. The generator is responsible for generating pseudo-seismic records, and the discriminator is responsible for distinguishing between generated data and real data. An improved generative adversarial network architecture is adopted, and the network is forced to learn the real distribution of data through gradient penalty and latent variable constraints to generate high-fidelity and diverse pseudo-seismic records, further expanding the training sample library. Then, real well and pseudo-well data are combined with the 3D structural model, and geostatistical interpolation methods are used to interpolate the entire work area. During the interpolation process, the structural model is used as the trend surface and the well point data is used as the hard data to generate the lithological probability prior distribution and physical property parameter prior distribution for each grid node. Finally, the prior model is output as a 3D probability field.

[0022] In this embodiment, the three-dimensional prior geological model output in this step is used as one of the inputs to step S4, providing prior beliefs for subsequent multi-source information fusion and serving as the starting point for probabilistic reasoning.

[0023] According to the appendix Figures 1 to 2 As shown, specifically disclosed in step S4, before outputting the three-dimensional probability volume, it is necessary to extract various evidence bodies characterizing lithology, physical properties, and fluids, and determine the evidence weight of these evidence bodies. The specific process is as follows: S41. Extraction of multiple evidence sources: Extract conventional seismic attribute volumes from the post-stack seismic data volume output in step S1, including root mean square amplitude, instantaneous frequency, coherence volume, curvature volume, spectral decomposition tuning volume, etc. Pre-stack inversion is performed on the pre-stack seismic data volume to obtain elastic parameter volumes, such as P-wave impedance, S-wave impedance, P-wave / S-wave velocity ratio, Poisson's ratio, Lamé constant, etc. If the work area has wide-azimuth seismic data, then extract the five-dimensional seismic attribute volume, including the anisotropic gradient of amplitude as a function of azimuth, crack density, anisotropic intensity, etc. At the well point, statistics are performed on a certain attribute value and lithology to establish a mapping function of lithology probability as the attribute value changes, such as a probability density curve or a cumulative probability curve. This mapping function is then applied to the entire work area to generate a lithology probability attribute body, which directly expresses the indicative probability of seismic attributes to lithology. S42. Determining the weight of evidence: Each piece of evidence is assigned a weight coefficient, which represents the reliability and sensitivity of the evidence in reservoir prediction. This weight can be determined using one or more of the following methods in combination: Distance weighting method: Establish a standard template for high-quality reservoirs, such as the attribute value range of typical well points, and calculate the Euclidean or Mahalanobis distance between the value of each evidence body and the template. The smaller the distance, the higher the weight. Sensitivity weighting method: Through forward modeling of rock physics, the response change amplitude of each evidence body is analyzed when different lithologies, physical properties and fluids change. The greater the change amplitude, the stronger the sensitivity and the higher the weight. Entropy weighting method: For each piece of evidence, calculate its information entropy in the entire work area. The smaller the entropy value, the greater the amount of information provided by the evidence and the less uncertainty, and therefore assign it a higher weight. Expert experience weighting: Geological and geophysical experts were organized to score the importance of each piece of evidence, and the analytic hierarchy process was used to construct a judgment matrix and calculate the eigenvectors to obtain the subjective weights. The final weights can be a weighted combination of the above methods and normalized to the [0,1] interval; S43. Evidence Theory Fusion Reasoning: DS evidence theory fusion reasoning, its basic probability allocation construction: For each grid node, the value of the evidence body and its weight are transformed into a basic probability allocation function; for example, for evidence body E_i, according to the probability density of its value falling into different categories, such as "high-quality reservoir", "poor reservoir", "non-reservoir", combined with the weight w_i, BPA_i is obtained. The fusion rule is as follows: Using Dempster's combination rule, the BPAs of multiple pieces of evidence are fused. Let m1, m2, ..., mn be the basic probability assignments of n pieces of evidence. Then the fused BPAm is: m(A)=(1 / (1-K))*Σ_{∩Ai=A}Π_{i=1..n}mi(Ai), where K is the conflict coefficient, representing the degree of inconsistency between the pieces of evidence; During the fusion process, the prior geological model obtained in step S3 is used as the basic evidence body to participate in the fusion, and its weight can be set according to the confidence level of the prior model. For each grid node underground, the probability assignment value of its belonging to different categories is calculated, resulting in three probability volumes: Three-dimensional lithology probability volume: such as P_sand(x,y,t) Three-dimensional physical property probability volume: such as P_highporo(x,y,t), which means the probability that the porosity is greater than the commercial lower limit; A three-dimensional probabilistic volume representing the presence of oil or gas: such as P_hydrocarbon(x,y,t), which represents the probability of containing oil or gas.

[0024] In this embodiment, the three-dimensional probability volume output in this step provides basic data for the threshold decision in step S5. The probability volume not only provides the judgment of the favorable position, but also quantifies the uncertainty of the judgment. According to the appendix Figures 1 to 2 As shown, it is particularly important to emphasize that in step S5, the three-dimensional lithology probability volume, three-dimensional physical property probability volume, and three-dimensional hydrocarbon probability volume output from step S4 are input. By defining three quantitative indicators—boundary uncertainty, misclassification risk, and trap size—a decision matrix under different probability thresholds is constructed. Subsequently, the entropy weight method is used to automatically assign weights based on the degree of variation of the indicators under each threshold, eliminating subjective bias. The TOPSIS method is used to calculate the closeness of each threshold to the ideal scheme, and the one with the highest closeness is selected as the optimal probability threshold. This threshold is then used to perform binary segmentation and spatial connectivity analysis on the probability volume, and finally, high-confidence three-dimensional favorable zones are extracted, realizing the objective transformation from continuous probability to discrete target.

[0025] In this embodiment, the specific decision objective in step S5 is defined as follows: three evaluation indicators are set to measure the delineation effect under different probability thresholds; the first indicator is boundary uncertainty. If the probability value fluctuates in a small proportion near the threshold, the boundary uncertainty is considered to be small; the second indicator is misclassification risk. Based on the known well verification results, if the weighted sum of the false positive rate and the false negative rate is small, the misclassification risk is considered to be low; the third indicator is the degree of compliance of the delineation size. If the delineation area or volume reaches the lower limit of commercial development, the degree of compliance is considered to be high. Candidate threshold set generation: Generate a series of candidate probability thresholds within a reasonable range with a fixed step size; Target weight calculation: Calculate the three target values ​​corresponding to each candidate threshold to form a decision matrix; Standardize the matrix to eliminate the influence of dimensions; Calculate the information entropy of each target. If the value of a target differs greatly under different thresholds, its information entropy is small. It is determined that the target contributes more to distinguishing the merits of the thresholds and is given a higher weight. Optimal threshold solution: Construct a weighted normalization matrix to determine the ideal solution and the negative ideal solution; calculate the degree of closeness of each candidate threshold to the ideal solution and the degree of distance from the negative ideal solution. The larger the ratio of the two, the better the threshold; select the threshold with the largest ratio as the optimal probability threshold. Favorable zone extraction: The three-dimensional probability volume output from step four is binarized and segmented using the optimal probability threshold. All grid nodes whose probability values ​​reach or exceed the threshold are extracted. Spatial connectivity analysis is performed on these nodes. If adjacent nodes are connected, they are merged into the same region, ultimately forming several continuous three-dimensional spatial regions. These regions are determined to be favorable zones. Logically, the favorable zone output in this step is the direct target area for detailed analysis and well location deployment in the subsequent step S6. Its spatial range is determined by an objective decision-making method, avoiding the defects of arbitrary manual selection.

[0026] According to the appendix Figure 1 As shown, it is particularly important to emphasize that the specific steps in step S6 are as follows: S61. Sensitive attribute fusion under probabilistic constraints: Select several seismic attributes that are sensitive to the heterogeneity of the reservoir only within the favorable zone. If the correlation coefficient of a certain attribute with the known reservoir quality is high, then the attribute is determined to have a larger weight during fusion. The selected attributes are fused using weighted linear superposition or evidence theory to generate a high-resolution reservoir fusion attribute body. This attribute body further distinguishes high-quality parts within the favorable zone. S62. 3D Visualization and Confidence Display: The 3D structural model from step S2, the 3D probability volume from step S4, the favorable zone from step S5, and the fused attribute volume generated in step S6 are loaded into the 3D visualization platform. To intuitively distinguish between high and low probabilities, a color mapping rule is set for the probability volume: if the probability value of a grid point is higher than a threshold, it is displayed in a warm color; if it is lower than the threshold, it is displayed in a cool color. In addition, to intuitively represent probability uncertainty, if the probability value of a region fluctuates around the threshold, the transparency of that region is increased. To intuitively represent the quality of the reservoir, a color mapping rule is set for the fused attribute volume: if the attribute value reflects good reservoir quality, it is displayed with high brightness or high saturation; if the quality is poor, it is displayed with low brightness or low saturation. S63. Well Location Optimization: In a 3D visualization environment, a comprehensive evaluation is conducted on locations within favorable zones. If the selected location is located in a high-probability core area and is far from fault fracture zones, the geological conditions are considered favorable. If the surface conditions at the selected location are suitable and the drilling trajectory is feasible, the engineering conditions are considered favorable. If the reserve estimation results at the selected location are good and the development costs are controllable, the economic conditions are considered favorable. Locations that simultaneously meet the geological, engineering, and economic favorable conditions are determined as preferred well locations, and the well location coordinates, design well depth, and geological recommendations are output to form a well location deployment plan.

[0027] According to the appendix Figure 3 As shown, the system includes: Module 1: Data Preprocessing and Construction Modeling Module, including steps S1 and S2, is used to receive raw seismic and drilling data, perform routine seismic data processing and well logging preprocessing, complete well-seismic calibration, layer and fault interpretation, and construct a three-dimensional structural model to provide basic data and spatial constraints for subsequent analysis. Module 2: Prior geological model construction module, including step S3, which is used to construct a Markov chain model based on well logging lithofacies statistics to generate a virtual well, assign elastic parameters through Monte Carlo sampling, and combine real well data with geostatistical interpolation to construct a three-dimensional geological model containing lithological and physical property probabilistic priors. Module 3: Multi-source information fusion and probability prediction module, including step S4, which is used to extract multiple evidence volumes from seismic data and determine weights, use evidence theory to fuse weighted evidence with prior models, calculate reservoir category probability for each underground point, and output three-dimensional lithology, physical properties and hydrocarbon probability volumes; Module 4: Decision Analysis and Well Location Deployment Module, including steps S5 and S6, is used to define multiple objectives and solve for the optimal probability threshold, extract favorable zones, integrate sensitive attributes within the zone to generate reservoir fusion bodies, optimize well locations through three-dimensional visualization of comprehensive geological, engineering and economic conditions, and output deployment schemes.

[0028] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method of predictive analysis for three-dimensional seismic exploration, characterized by, The method includes: S1. Data preparation and preprocessing: Obtain the original 3D seismic data and drilling and logging data of the target work area, and generate a 3D seismic data volume suitable for subsequent interpretation and prediction through conventional processing procedures; S2. Structural Interpretation and Geological Framework Establishment: Using the post-stack seismic data and well logging data output in step S1, identify the main subsurface stratigraphic interfaces and faults, and construct a three-dimensional structural model; S3. Well-seismic joint analysis and prior model construction: Based on the limited well data and regional geological understanding in steps S1 and S2, the lithofacies sequence is simulated by Markov chain, elastic parameters are assigned by Monte Carlo sampling, and the pseudo-well samples are expanded by generative adversarial network. Finally, a three-dimensional geological model containing prior probabilities of lithology and physical properties is constructed. S4. Multi-source information fusion and probabilistic reservoir prediction: Extract conventional attributes, elastic parameters and five-dimensional anisotropy from the seismic data in the first three steps. Determine the weight of each piece of evidence through distance matching, sensitivity analysis and information entropy calculation. Use evidence theory to fuse the weighted evidence with the prior model and output a three-dimensional probabilistic body of lithology, physical properties and hydrocarbon content. S5. Multi-objective threshold decision and favorable zone delineation: For the three-dimensional probability volume output in step S4, multiple decision objectives are defined, and a multi-objective decision method is used to objectively determine the optimal probability threshold, thereby extracting the favorable zone with high confidence. S6. High-precision target visualization and well location deployment: Within the favorable zone determined in step S5, targeted attribute fusion and three-dimensional visualization are performed, and the most favorable drilling location is selected by combining geological, engineering and economic factors.

2. The method of claim 1, wherein, In step S1, the main acquisition target is the original three-dimensional seismic data of the work area. At the same time, the existing drilling and logging data in the work area are collected. The logging curves are subjected to environmental correction and normalization to eliminate systematic errors caused by different instruments and measurement conditions. Standardized curves that can be used for calibration and rock physical analysis are obtained. Finally, the pre-stack three-dimensional seismic data volume, the post-stack three-dimensional seismic data volume, and the standardized logging curves are output.

3. The method of claim 2, wherein, The specific process in step S2 is as follows: S21. Well-seismic calibration: Using the standardized well logging curves obtained in step S1, a synthetic seismic record is generated. By comparing the synthetic record with the seismic traces near the well, the seismic reflection characteristics of the target layer are determined, and the time-depth calibration is completed. S22. Stratigraphic Interpretation: Stratigraphic interpretation is performed on the 3D seismic data volume along the calibrated reflection phase axis. A combination of automatic tracing and manual correction is used to obtain the stratigraphic grid of the main target layer and key marker layers. S23. Fault Interpretation: Perform coherence volume or variance volume attribute calculations on seismic data to enhance fault display, identify fault traces on the attribute volume, combine and close faults in conjunction with seismic profiles, and construct three-dimensional fault models by picking fault planes. S24. Construction of 3D Structural Model: Integrate the interpreted layer mesh and fault model to establish a closed 3D structural model.

4. The method of claim 3, wherein, The specific process of simulating lithofacies sequences based on Markov chains in step S3 is as follows: input the well logging lithofacies interpretation results under the constraints of the structural model in step S2, statistically analyze the vertical transition probabilities of each lithofacies to construct a Markov chain model, and generate a lithofacies sequence of virtual wells step by step according to the transition probabilities from the initial lithofacies using a stochastic simulation method. Repeat this process multiple times to obtain a large number of virtual well lithofacies columns that conform to the laws of geological statistics. During the simulation, the probability distribution of lithofacies thickness is considered so that the generated sequence has reasonable layer thickness characteristics. Subsequently, a rock physics template is established based on the well logging data. The probability distribution of elastic parameters for each type of rock facies is statistically analyzed and the correlation between parameters is analyzed. For each generated virtual rock facies, specific elastic parameter values ​​are randomly sampled from the corresponding probability distribution to generate a pseudo-well dataset containing depth, lithology, and elastic parameters.

5. The method of claim 4, wherein, The specific steps of optimizing and constructing the 3D geological model using the generative adversarial network in step S3 are as follows: A generative adversarial network is constructed using the real well data and pseudo-well data as training samples. The generator is responsible for generating pseudo-seismic records, and the discriminator is responsible for distinguishing between generated and real data. An improved generative adversarial network architecture is adopted, and the network is forced to learn the real distribution of data through gradient penalty and latent variable constraints to generate high-fidelity and diverse pseudo-seismic records, further expanding the training sample library. Subsequently, the real and pseudo-well data are combined with the 3D structural model, and geostatistical interpolation methods are used to interpolate the entire work area. During the interpolation process, the structural model is used as the trend surface, and the well point data is used as the hard data to generate the lithological probability prior distribution and physical property parameter prior distribution for each grid node. Finally, the prior model is output as a 3D probability field.

6. The three-dimensional seismic exploration prediction and analysis method according to claim 5, characterized in that, In step S4, before outputting the three-dimensional probability volume, it is necessary to extract various evidence volumes characterizing lithology, physical properties, and fluids, and determine the weight of each evidence volume. The specific process is as follows: S41. Extraction of multiple evidence sources: Extract conventional seismic attribute volume from the post-stack seismic data volume output in step 1, and obtain elastic parameter volume by pre-stack inversion from the pre-stack seismic data volume. If the work area has wide azimuth seismic data, extract the five-dimensional seismic attribute volume. Statistically analyze the relationship between a certain attribute value and lithology at the well point, establish a mapping function for the change of lithology probability with attribute value, and apply the mapping function to the entire work area to generate lithology probability attribute volume. S42. Determining Evidence Weights: Each piece of evidence is assigned a weight coefficient, which represents the reliability and sensitivity of the evidence in reservoir prediction. The weight determination method includes at least one of the following: if the matching distance between the value of an evidence and the standard template of a high-quality reservoir is small, the evidence is judged to have a higher weight; if the forward modeling of rock physics shows that an evidence is sensitive to changes in the target parameter, the evidence is judged to have a higher weight; if the information entropy of an evidence is small in the entire work area, it indicates that it provides a large amount of information, and the evidence is judged to have a higher weight; if multiple experts give consistent evaluations of the importance of an evidence and the scores are high, its subjective weight is calculated by the analytic hierarchy process; the final weight is a combination of the above methods and normalized.

7. The three-dimensional seismic exploration prediction and analysis method according to claim 6, characterized in that, In step S4, the specific steps for outputting the three-dimensional probability volume are as follows: S43. Evidence Theory Fusion Reasoning: For each grid node in step S42, the numerical value and weight of the evidence body are transformed into a basic probability allocation function. During the fusion process, the prior geological model obtained in step three is used as the basic evidence body to participate in the fusion, and its weight is set according to the confidence level of the prior model. The evidence theory combination rules are adopted to fuse the basic probability allocations of multiple evidence bodies. If the evidence bodies point to the same direction, the probability allocation value of the target category increases after fusion. If the evidence bodies point to conflict, the conflict coefficient increases, and the fusion result reflects uncertainty. The probability allocation value of each underground grid node belonging to different reservoir categories is calculated, and finally three probability bodies are output: three-dimensional lithology probability body, three-dimensional physical property probability body, and three-dimensional hydrocarbon probability body.

8. The three-dimensional seismic exploration prediction and analysis method according to claim 7, characterized in that, In step S5, the three-dimensional lithology probability volume, three-dimensional physical property probability volume, and three-dimensional hydrocarbon probability volume output in step S4 are input. By defining three quantitative indicators—boundary uncertainty, misclassification risk, and trap size—a decision matrix under different probability thresholds is constructed. Subsequently, the entropy weight method is used to automatically assign weights based on the degree of variation of the indicators under each threshold, eliminating subjective bias. The TOPSIS method is used to calculate the closeness of each threshold to the ideal solution, and the one with the highest closeness is selected as the optimal probability threshold. This threshold is then used to perform binary segmentation and spatial connectivity analysis on the probability volume, and finally extracts the three-dimensional favorable zone with high confidence, realizing the objective transformation from continuous probability to discrete target.

9. A three-dimensional seismic exploration prediction and analysis method according to claim 8, characterized in that, The specific steps in step S6 are as follows: S61. Sensitive attribute fusion under probabilistic constraints: Select several seismic attributes that are sensitive to the heterogeneity of the reservoir only within the favorable zone. If the correlation coefficient of a certain attribute with the known reservoir quality is high, then the attribute is determined to have a larger weight during fusion. The selected attributes are fused using weighted linear superposition or evidence theory to generate a high-resolution reservoir fusion attribute body. This attribute body further distinguishes high-quality parts within the favorable zone. S62. 3D Visualization and Confidence Display: The 3D construction model from step S2, the 3D probability volume from step S4, the favorable zone from step S5, and the fused attribute volume generated in step S6 are loaded into a 3D visualization platform. To intuitively distinguish between high and low probabilities, a color mapping rule is set for the probability volume: if the probability value of a grid point is higher than a threshold, it is displayed in a warm color; if it is lower than the threshold, it is displayed in a cool color. In addition, to intuitively represent probability uncertainty, if the probability value of a region fluctuates around the threshold, the transparency of that region is increased. To intuitively represent the quality of the reservoir, a color mapping rule is set for the fused attribute volume: if the attribute value reflects good reservoir quality, it is displayed with high brightness or high saturation; if the quality is poor, it is displayed with low brightness or low saturation. S63. Well Location Optimization: In a 3D visualization environment, a comprehensive evaluation is conducted on locations within favorable zones. If the selected location is located in a high-probability core area and is far from fault fracture zones, the geological conditions are considered favorable. If the surface conditions at the selected location are suitable and the drilling trajectory is feasible, the engineering conditions are considered favorable. If the reserve estimation results at the selected location are good and the development costs are controllable, the economic conditions are considered favorable. Locations that simultaneously meet the geological, engineering, and economic favorable conditions are determined as preferred well locations, and the well location coordinates, design well depth, and geological recommendations are output to form a well location deployment plan.

10. A three-dimensional seismic exploration prediction and analysis system, comprising the three-dimensional seismic exploration prediction and analysis method according to any one of claims 1-9, characterized in that, The system includes: Module 1: Data Preprocessing and Construction Modeling Module, including steps S1 and S2, is used to receive raw seismic and drilling data, perform routine seismic data processing and well logging preprocessing, complete well-seismic calibration, layer and fault interpretation, and construct a three-dimensional structural model to provide basic data and spatial constraints for subsequent analysis. Module 2: Prior geological model construction module, including step S3, which is used to construct a Markov chain model based on well logging lithofacies statistics to generate a virtual well, assign elastic parameters through Monte Carlo sampling, and construct a three-dimensional geological model containing lithological and physical property probabilistic priors by combining real well data with geostatistical interpolation. Module 3: Multi-source information fusion and probability prediction module, including step S4, which is used to extract multiple evidence bodies from seismic data and determine weights, use evidence theory to fuse weighted evidence with prior models, calculate reservoir category probability for each underground point, and output three-dimensional lithology, physical properties and hydrocarbon probability bodies; Module 4: Decision Analysis and Well Location Deployment Module, including steps S5 and S6, is used to define multiple objectives and solve for the optimal probability threshold, extract favorable zones, integrate sensitive attributes within the zone to generate reservoir fusion bodies, optimize well locations through three-dimensional visualization of comprehensive geological, engineering and economic conditions, and output deployment schemes.

Images