Reservoir configuration constraint seismic attribute intelligent fusion and reservoir prediction method and system
By constructing a sand body structure model and using a multi-task learning method, combined with seismic forward modeling and well logging data, and optimizing the fusion of seismic attributes, the problem of low prediction accuracy when the sand body configuration is complex in the existing technology is solved, and the synchronous high-precision prediction of sand body structure and thickness and the enhancement of interpretability are realized.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA UNIV OF PETROLEUM (BEIJING)
- Filing Date
- 2025-06-11
- Publication Date
- 2026-05-22
AI Technical Summary
Existing seismic attribute fusion methods have low prediction accuracy when sand body configurations are complex, making it difficult to meet the needs of fine reservoir prediction, and lack effective characterization of sand body structure and model interpretability.
By constructing a sand body structure model, extracting seismic wavelets from the study area for forward modeling, optimizing sensitive seismic attributes, and employing a multi-task learning model combined with well logging interpretation data, the synchronous prediction of sand body thickness and structure is achieved. Attention mechanisms and Bayesian hyperparameter tuning algorithms are added to enhance the interpretability and accuracy of the model.
It improves the accuracy and interpretability of sand body prediction, realizes the synergistic prediction of sand body structure and thickness, reduces the uncertainty of reservoir prediction, and provides a more reliable geological basis for oil and gas exploration.
Smart Images
Figure CN120669312B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of seismic attribute analysis technology, and in particular to a method and system for intelligent fusion of seismic attributes constrained by reservoir configuration and for reservoir prediction. Background Technology
[0002] Seismic attribute analysis technology has been widely applied to reservoir prediction. Single seismic attributes extract and enhance information from one or more aspects of seismic data, typically improving reservoir prediction accuracy, but they suffer from significant limitations, losing a large amount of valuable information. Therefore, multi-attribute comprehensive analysis is an effective way to address these issues, with intelligent multi-attribute fusion being a current hot topic and cutting-edge area in attribute analysis.
[0003] Intelligent fusion of seismic attributes is an advanced seismic attribute analysis technique based on machine learning. It predicts the spatial distribution of reservoirs in inter-well regions by establishing a nonlinear mapping model between seismic data (especially seismic attributes) and well logging data. Compared to traditional RGB fusion and principal component analysis methods, this technique has significant advantages such as less subjective intervention, higher resolution, and stronger noise resistance, and has achieved good application results in the field of oil and gas reservoir prediction. However, existing technologies still have two main limitations: First, current methods are mainly based on supervised learning frameworks for well logging data and use shallow machine learning models (such as support vector machines or K-nearest neighbors) for modeling, but the model representation capabilities are limited, making it difficult to fully extract effective information from the data. Second, and more importantly, current research focuses on the macroscopic response relationship between sand body thickness and seismic attributes, neglecting the differentiated influence of different sand body structures on seismic response. Therefore, when the sand body configuration is complex, the accuracy of existing seismic attribute fusion and sand body prediction methods will be significantly reduced, making it difficult to meet the needs of fine reservoir prediction.
[0004] Multi-task learning offers a new approach to solving these problems. This technology, by constructing a shared feature extraction layer and dedicated task branches, can effectively integrate multi-source data information to achieve a comprehensive evaluation of multiple parameters such as reservoir thickness, structure, and saturation. Compared with traditional single-task learning methods, multi-task learning not only improves prediction accuracy and model robustness but also makes fuller use of key information in the data. However, this method also faces significant challenges: First, within the multi-task learning framework, each prediction task must maintain geological relevance. Besides sand body thickness, how to scientifically select other prediction targets as auxiliary monitoring information to enhance the model's generalization ability still requires further research. Second, although sand body structure affects seismic response characteristics, existing intelligent reservoir prediction methods still have significant limitations in characterizing sand body structure. Specifically, the following are the characteristics: (1) The simple thickness prediction model is not sensitive enough to changes in seismic waveforms and is difficult to effectively capture the seismic response of complex sand body structures such as thin interlayers; (2) The geological knowledge introduced in the prediction process is insufficient and the interpretability of the characteristic correlation of the model prediction process is poor, thus making it difficult to meet the needs of actual reservoir characterization; (3) Since the influence of sand body structure on seismic response is usually weaker than that of sand body thickness, previous seismic attribute analysis of sand body structure has mainly focused on profiles, and the prediction range is equivalent to a line on the plane. There is a lack of seismic attribute analysis methods for the planar distribution of sand body structure. Summary of the Invention
[0005] To address the aforementioned problems, the purpose of this invention is to provide a method and system for intelligent fusion of reservoir configuration-constrained seismic attributes and reservoir prediction, which can achieve multi-faceted and refined prediction of reservoirs, improve the accuracy of sand body interpretation, and enhance the interpretability of prediction results.
[0006] To achieve the above objectives, in a first aspect, the technical solution adopted by the present invention is as follows: a method for intelligent fusion of reservoir configuration-constrained seismic attributes and reservoir prediction, comprising: constructing a sand body structure model; extracting actual seismic wavelets from the actual seismic data volume of the study area; and using a seismic forward modeling method to perform seismic forward modeling of the sand body structure model using the actual seismic wavelets of the study area; classifying the seismic forward modeling waveforms based on the seismic forward modeling simulation to obtain seismic forward modeling waveform classification results; and classifying the sand body structure in well logging interpretation according to the seismic forward modeling waveform classification results to obtain sand body structure classification results; selecting seismic attributes sensitive to sand body configuration and sand body thickness from the actual seismic data volume of the study area; inputting the selected seismic attributes, well logging interpretation of sand body thickness and sand body structure into a multi-task learning model to perform intelligent analysis of seismic attribute data of the study area based on multi-task learning, thereby realizing prediction of sand body thickness and sand body structure.
[0007] Furthermore, a sand body structure model is constructed, including:
[0008] Based on statistical analysis of core, well logging and seismic data of the study area, the characteristics of sand body thickness distribution and sand body stacking relationship of the study area were analyzed, and a sand body structure model was constructed.
[0009] Among them, the sand body structure model is a conceptual model that combines the sand body thickness and the relationship between sand body stacking.
[0010] Furthermore, the actual seismic wavelet of the study area is extracted from the actual seismic data volume of the study area. Specifically, the seismic statistical wavelet of the target layer is extracted as the actual seismic wavelet of the study area.
[0011] Furthermore, seismic forward modeling of the sand body structure model was performed using actual seismic wavelets from the study area, including:
[0012] By performing seismic forward modeling on sand body structure models, a seismic forward modeling response template covering different sand body thicknesses and stacking relationships is constructed, and geological models with similar seismic responses are grouped into the same category to obtain a sand body structure classification template.
[0013] Furthermore, sensitive seismic attributes are selected by: using cluster analysis to divide multiple seismic attributes into multiple categories, and then selecting the seismic attribute with the highest correlation between the reservoir and the seismic attribute in each category based on correlation analysis, as the sensitive seismic attribute.
[0014] Furthermore, multi-task learning models include:
[0015] Input seismic attributes, sand body thickness and sand body structure from well logging interpretation, and use sand body thickness and sand body structure as labels for multi-task learning;
[0016] The multi-task learning uses a shared hidden layer and sub-network structure layers for reservoir structure prediction and thickness prediction, while adding an attention mechanism at the connection points between the shared layer and the sub-network layers.
[0017] Among them, the shared hidden layer is a shared feature extraction layer used to learn the common features of seismic attributes; the sub-network structure layer is an independent task branch layer used to fit sand body thickness and sand body structure parameters respectively; the attention mechanism is used to dynamically allocate attribute weights to strengthen the correlation between sand body configuration, sand body thickness and seismic response.
[0018] In the multi-task learning process, it is necessary to minimize the sum of the loss functions of the two task branches, namely sand body thickness and sand body structure, and achieve global optimization for both.
[0019] Furthermore, the selected seismic attributes, well-logged interpreted sand body thickness, and sand body structure are input into a multi-task learning model to perform intelligent analysis of seismic attribute data in the study area based on multi-task learning, thereby achieving prediction of sand body thickness and sand body structure, including:
[0020] Based on forward modeling, sand body structure classification, and optimized seismic attributes, an input dataset is constructed, which includes sand body thickness interpreted from well logging, sand body structure classification results, and optimized seismic attributes.
[0021] The sand body thickness and sand body structure classification results interpreted from well logging are used as supervised data for multi-task learning, with sand body thickness as the regression target and sand body structure as the classification target. At the same time, the preferred seismic attribute values near the well access are extracted, and the seismic attribute values of the well access and the sand body thickness and sand body structure classification results interpreted from well logging are used as training datasets.
[0022] A multi-task learning model is used to fuse seismic attributes and predict reservoirs under sand body configuration constraints. By exploring the knowledge correlation between multiple tasks, the synchronous high-precision prediction of sand body structure and thickness is achieved.
[0023] A Bayesian hyperparameter tuning algorithm is used to adaptively balance the weights between tasks, achieving a deep integration of geological knowledge and data-driven approaches, and realizing the global optimization of sand body structure and thickness in sand body prediction.
[0024] Output the predicted results of the spatial structure and thickness distribution of sand bodies in the study area.
[0025] Secondly, the technical solution adopted by this invention is as follows: a reservoir configuration-constrained intelligent fusion and reservoir prediction system for seismic attributes, comprising: a forward modeling module, which constructs a sand body structure model, extracts actual seismic wavelets from the actual seismic data volume of the study area, and uses the seismic forward modeling method to perform seismic forward modeling simulation on the sand body structure model; a classification module, which classifies the seismic forward modeling waveforms based on the seismic forward modeling simulation, obtains the seismic forward modeling waveform classification results, and classifies the sand body structure in well logging interpretation according to the seismic forward modeling waveform classification results, obtains the sand body structure classification results; and a prediction module, which selects seismic attributes sensitive to sand body configuration and sand body thickness from the actual seismic data volume of the study area, inputs the selected seismic attributes, well logging interpretation of sand body thickness and sand body structure into a multi-task learning model to perform intelligent analysis of seismic attribute data of the study area based on multi-task learning, and realizes prediction of sand body structure and sand body thickness.
[0026] Thirdly, the technical solution adopted by the present invention is: a computer-readable storage medium for storing one or more programs, wherein the one or more programs include instructions, which, when executed by a computing device, cause the computing device to perform any of the methods described above.
[0027] Fourthly, the technical solution adopted by the present invention is: a computing device comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for performing any of the methods described above.
[0028] The present invention has the following advantages due to the adoption of the above technical solutions:
[0029] 1. This invention introduces forward modeling and sand body structure classification templates to directly link seismic waveform characteristics with geological body characteristics, forming an intelligent attribute fusion method driven by both knowledge and data. This method not only improves the prediction accuracy of sand bodies but also takes into account the prediction of sand body structures. At the same time, it enhances the interpretability of prediction results and makes the prediction results more consistent with actual geological characteristics.
[0030] 2. This invention is the first to construct a seismic attribute fusion method based on a multi-task learning framework, which realizes the collaborative prediction of sand body thickness and structural characteristics. It overcomes the defects of decoupled modeling of the two in traditional methods, effectively reduces the uncertainty of reservoir prediction, and provides a more reliable geological basis for oil and gas exploration and development. Attached Figure Description
[0031] Figure 1 This is a flowchart illustrating the overall structure of the reservoir configuration-constrained seismic attribute intelligent fusion and reservoir prediction method in this embodiment of the invention.
[0032] Figure 2 This is a detailed flowchart of the intelligent fusion of reservoir configuration-constrained seismic attributes and reservoir prediction method in an embodiment of the present invention;
[0033] Figure 3 This is a schematic diagram of the main sand body structure development pattern in the example area of this invention.
[0034] Figure 4 This is a diagram showing the seismic forward modeling waveform response characteristics (i.e., seismic forward modeling results) of the main sand body structure in the example area of this invention embodiment;
[0035] Figure 5 This is a schematic diagram of the multi-task learning algorithm architecture in an embodiment of the present invention;
[0036] Figure 6 This is a schematic diagram of a multi-attribute intelligent fusion method based on multi-task learning in an embodiment of the present invention;
[0037] Figure 7 These are seismic forward modeling response characteristic diagrams under different sand body structures in embodiments of the present invention. Detailed Implementation
[0038] In the existing technology, sand body thickness prediction based on multi-attribute intelligent fusion is a reservoir characterization method that integrates seismic attribute analysis and computer technology, and is mainly used for quantitative prediction of sand body thickness. This method is based on seismic sedimentology theory and integrates multiple seismic attributes through machine learning to achieve quantitative prediction of sand body thickness and fine characterization of spatial distribution. Its technical process is as follows: (1) Seismic attribute extraction and standardization: extract multiple attributes such as amplitude, frequency and energy along the target layer and perform normalization to eliminate dimensional differences; (2) Seismic attribute optimization: select parameters that are significantly related to the sand body thickness at the well point from multiple seismic attributes such as amplitude, frequency and energy, and calculate the correlation coefficient matrix between attributes to remove redundant attributes; (3) Multi-attribute intelligent fusion modeling: use machine learning or deep learning algorithms to perform nonlinear fusion of the optimized attributes to construct a sand body thickness prediction model; (4) Fine characterization of sand body distribution: use the trained sand body thickness prediction model to fuse attribute surfaces to achieve clear division of sand body plane boundaries and quantitative characterization of thickness spatial changes. This method uses computer technology to construct a nonlinear prediction model between seismic attributes and sand body thickness, effectively realizing the quantitative prediction of sand body thickness.
[0039] However, this approach has the following shortcomings: The seismic property responses of sand bodies with the same thickness but different structures differ significantly. Existing methods only learn and model independently for sand body thickness, without considering the influence of sand body structure on seismic property responses. This leads to prediction results deviating from actual sedimentary patterns, especially in areas with complex geological structures, where the deviation is significantly greater. Furthermore, the correlation between the internal parameters of the existing model and geological features is difficult to trace, resulting in a lack of interpretability in the prediction results and failing to provide clear geological basis for adjusting development plans.
[0040] In the existing technology, the sand body structure prediction method based on geological dissection analysis is a qualitative-semi-quantitative analysis technique based on sedimentary facies analysis and combined with well logging, core and seismic data. It is mainly used to describe and predict the structural characteristics of sand bodies and their spatial distribution morphology. This method mainly relies on artificial dissection of sedimentary systems to identify the sand body stacking patterns and contact relationships, thereby determining reservoir connectivity and heterogeneity. Its technical process is as follows: (1) Under the guidance of high-resolution sequence stratigraphy theory, sedimentary microfacies units are divided using multi-source data such as core, well logging and seismic data; (2) Sand body types and their spatial distribution characteristics are identified through seismic inversion and seismic attribute fusion technology; (3) Based on the vertical and lateral stacking and contact relationships of sand bodies, a sand body structure model is constructed, including the vertical stacking pattern and lateral contact relationship, to clarify the sand body stacking relationship in the study area.
[0041] However, the following shortcomings exist in this scheme: (1) It heavily relies on the experience of geological experts, and the analysis is highly subjective: The process of dissecting sand body structures requires a large amount of manual participation, and the modeling accuracy and reliability are highly dependent on the subjective judgment of geological personnel, lacking standardized and automated means. (2) It lacks data-driven modeling means: Traditional models cannot utilize deep correlation information in multi-source data such as well logging and seismic data, resulting in low information utilization efficiency and poor prediction generalization ability. (3) The structure and thickness are decoupled in the modeling, and there is a lack of collaborative prediction mechanism: The current method separates the sand body structure and thickness prediction, failing to reflect the coupling relationship between the two in the sedimentary system, resulting in the structure prediction results failing to effectively support the thickness prediction accuracy.
[0042] Therefore, this invention provides a method and system for intelligent fusion of reservoir configuration-constrained seismic attributes and reservoir prediction, which overcomes the technical defects existing in the prior art, realizes multi-faceted and refined prediction of reservoirs, and provides more reliable geological basis for oil and gas exploration and development.
[0043] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.
[0044] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.
[0045] In this invention, abbreviations and key terms are defined as follows:
[0046] Reservoir architecture: The morphology, scale, orientation, and stacking relationships of reservoir units of different levels, also known as sedimentary architecture. This concept reflects the differences in the spatial configuration and distribution of reservoir units and seepage barriers of different levels.
[0047] Sand body configuration: In clastic rocks, reservoirs are dominated by sand bodies. The reservoir configuration is also called sand body configuration, which mainly refers to the geometric shape, spatial distribution and superposition relationship of sand bodies. Some scholars also refer to it as sand body structure.
[0048] Earthquake attributes: Geometric, kinematic, dynamic, or statistical characteristics of seismic waves derived from seismic data through mathematical transformations, used to describe information such as underground geological structure, lithology, and fluid properties.
[0049] Seismic attribute optimization: For a specific problem to be solved in a specific region, the combination of attributes that is most sensitive to and effective for specific geological targets is selected from a large number of seismic attributes to improve interpretation efficiency and accuracy.
[0050] Seismic attribute fusion: This involves organically combining multiple seismic attributes through mathematical, statistical, or geological methods to form a new information body that comprehensively represents underground geological features, thereby improving the reliability and resolution of seismic interpretation.
[0051] Multi-task learning: Multi-task learning is a machine learning paradigm that aims to improve the generalization performance and prediction accuracy of a model by simultaneously learning shared representations of multiple related tasks. Its core idea is to leverage the correlation between tasks to achieve knowledge transfer and information complementarity during model training, thereby overcoming problems such as data sparsity or overfitting in single-task learning.
[0052] Mutual constraints and global optimum: In the optimization or modeling of complex systems, multiple variables, parameters, or objective functions form a dynamic equilibrium through mutual constraints, ultimately enabling the overall system performance to reach the theoretically optimal state.
[0053] In one embodiment of the present invention, a method for intelligent fusion of seismic attributes constrained by reservoir configuration and for reservoir prediction is provided to address the problems of insufficient accuracy and strong ambiguity in seismic reservoir prediction in existing technologies. Seismic attributes are the comprehensive response of geological bodies within a certain thickness range (resolution range), including information on sand body thickness and structure, among other aspects. According to seismic forward modeling results, sand body thickness and structure have the most significant impact on the seismic attribute response characteristics in terms of macroscopic reservoir features. However, previous studies often only used sand body thickness as the target data, leading to a mismatch between the target data and seismic attribute information. Therefore, in this embodiment, as... Figure 1 , Figure 2 As shown, the method includes the following steps:
[0054] 1) Construct a geological model including different sand body structures, i.e., a sand body structure model. Extract actual seismic wavelets from the actual seismic data volume of the study area, and use the seismic forward modeling method to perform seismic forward modeling on the sand body structure model using the actual seismic wavelets of the study area. Then, classify the sand body structures interpreted from well logging according to the differences in seismic waveform response, and obtain the classification results of the sand body structures.
[0055] 2) Based on seismic forward modeling, the seismic forward modeling waveforms are classified to obtain the seismic forward modeling waveform classification results. Based on the seismic forward modeling waveform classification results, the sand body structure in the well logging data is classified to obtain the sand body structure classification results.
[0056] 3) From the actual seismic data volume of the study area, select the seismic attributes that are sensitive to sand body configuration (i.e., sand body structure) and sand body thickness. Input the selected seismic attributes, sand body thickness and sand body structure interpreted by well logging into the multi-task learning model to perform intelligent analysis of seismic attribute data of the study area based on multi-task learning. This will achieve mutual constraint and global optimization of the two genetically related parameters of sand body structure and sand body thickness, and ultimately realize the prediction of sand body thickness and sand body structure.
[0057] In step 1) above, the sand body structure model is constructed, specifically by statistically analyzing the sand body thickness distribution characteristics and sand body stacking relationships in the study area based on core, well logging, and seismic data. The sand body structure model is a conceptual model combining sand body thickness and sand body stacking relationships.
[0058] Taking a certain study area as an example, the target layer mainly develops two sets of sand bodies of varying thicknesses. In some areas, the sand bodies are closely superimposed (macroscopically resembling a single sand body), while in other areas, mudstone of varying thicknesses develops between the sand bodies, such as... Figure 3 As shown.
[0059] In step 1) above, the actual seismic wavelet of the study area is extracted from the actual seismic data volume of the study area. Specifically, the seismic statistical wavelet of the target layer is extracted as the actual seismic wavelet of the study area.
[0060] In step 1) above, based on seismic forward modeling, the seismic forward modeling waveforms are classified, including:
[0061] By performing seismic forward modeling on sand body structure models, seismic forward response templates covering different sand body thicknesses and stacking relationships are constructed. Geological models with similar seismic responses are grouped into the same category, resulting in sand body structure classification templates, such as... Figure 3 Different colored regions are represented. Specifically, sand body structure classification templates are generated through forward modeling and waveform similarity analysis.
[0062] In this embodiment, a suitable seismic forward modeling method (such as convolution model, wavefront velocity field scanning, ray theory, wave equation, etc.) is adopted, and the actual seismic wavelet of the study area is applied to perform seismic forward modeling simulation on the conceptual model of sand body thickness and structural combination in the study area.
[0063] In step 2) above, the sand body structure and thickness type in the well logging data are corrected and classified using the forward modeling classification results. Specifically, they can be divided into three main categories:
[0064] Category 1: The seismic response consists of two strong reflection axes, making it difficult to distinguish between the upper and lower sand bodies.
[0065] The second type: the seismic response has a weaker reflection axis between the two strong reflection axes, which can distinguish the two phases of sand bodies.
[0066] The third type: the seismic response is characterized by multiple strong radial axes, clearly distinguishing the two phases of sand bodies (e.g., Figure 4 As shown in the figure, this provides a calibration basis for subsequent multi-task learning.
[0067] In step 2) above, the seismic attributes that are sensitive to sand body configuration and sand body thickness are selected. Specifically, cluster analysis is used to divide various seismic attributes into multiple categories (in this embodiment, the seismic attributes are divided into several large categories with low correlation). Then, based on correlation analysis, the seismic attribute with the highest correlation between the reservoir and the seismic attribute is selected in each category as the sensitive seismic attribute.
[0068] In step 3) above, if Figure 5 As shown, the multi-task learning model includes:
[0069] Input seismic attributes, sand body thickness and sand body structure from well logging interpretation, and use sand body thickness and sand body structure as labels for multi-task learning;
[0070] The multi-task learning uses a shared hidden layer and separate sub-network structures for reservoir structure prediction and thickness prediction, while adding an attention mechanism at the connection points between the shared layer and the sub-network layers.
[0071] Among them, the shared hidden layer is a shared feature extraction layer used to learn the common features of seismic attributes; the sub-network structure layer is an independent task branch layer used to fit sand body thickness and sand body structure parameters respectively; the attention mechanism is used to dynamically allocate attribute weights to strengthen the correlation between sand body configuration and seismic response.
[0072] In the multi-task learning process, it is necessary to minimize the sum of the loss functions of the two task branches, namely sand body thickness and sand body structure, to achieve mutual constraints between the two causal correlation parameters, and ultimately achieve the global optimum of both.
[0073] In this embodiment, the parameters in the shared layer are shared by multiple tasks. The attention mechanism in the middle is used to focus on the more important common features between them, and to enhance the knowledge interaction and mutual constraints between multiple tasks. The individual sub-networks take into account the small differences between each task. Therefore, the last few layers use independent networks to perform data fitting, making the simulation more robust.
[0074] In step 3) above, the selected seismic attributes, well-logged interpreted sand body thickness, and sand body structure are input into the multi-task learning model to perform intelligent analysis of the seismic attribute data of the study area through multi-task learning, thereby achieving prediction of sand body configuration (sand body structure) and sand body thickness. Figure 6 As shown, it includes the following steps:
[0075] 3.1) Based on forward modeling, sand body structure classification, and optimized seismic attributes, an input dataset is constructed, which includes sand body thickness interpreted from well logging, sand body structure classification results, and optimized seismic attributes;
[0076] 3.2) The sand body thickness and sand body structure classification results interpreted from well logging are used as supervised data for multi-task learning, where sand body thickness is the regression target and sand body structure is the classification target; at the same time, the preferred seismic attribute values (all preferred seismic attributes) near the well access are extracted, and the seismic attribute values of the well access and the sand body thickness and sand body structure classification interpreted from well logging are used as training datasets.
[0077] 3.3) A multi-task learning model is used to fuse seismic attributes and predict reservoirs under sand body configuration constraints. By exploring the knowledge correlation between multiple tasks, the synchronous high-precision prediction of sand body structure and thickness is achieved.
[0078] This embodiment also includes a model reliability verification step: evaluating the model's credibility and generalization performance through cross-validation.
[0079] 3.4) The Bayesian hyperparameter tuning algorithm is used to adaptively balance the weights between tasks, so as to achieve a deep integration of geological knowledge and data-driven approaches, and to achieve the global optimum of sand body configuration and thickness in sand body prediction.
[0080] Specifically, when the model's credibility is >90% (or reaches the threshold set by the researcher), proceed to the next stage; otherwise, iteratively execute the reliability verification step and step 3.4).
[0081] This embodiment also includes blind test verification: the model's prediction accuracy and generalization ability are tested using blind well data. When the prediction accuracy is ≥90% (or reaches the threshold set by the researcher), the model is deemed qualified; otherwise, the reliability verification step and step 3.4 are re-executed for optimization.
[0082] 3.5) Inter-well prediction and results output: Output the predicted results of the spatial structure and thickness distribution of sand bodies in the study area by passing the validated multi-task learning model.
[0083] In this embodiment, a joint loss function is used to simultaneously optimize two types of prediction tasks, and a Bayesian hyperparameter tuning algorithm is combined to adaptively balance the weights between tasks, thereby achieving a deep fusion of geological knowledge and data-driven approaches. In sand body prediction, the global optimality of sand body structure and thickness is achieved, which not only improves the prediction accuracy of sand bodies, but also takes into account the prediction of sand body structure, and enhances the interpretability of the prediction results.
[0084] In summary, by adopting the technical solutions in the above embodiments, the present invention has at least the following beneficial effects: (1) It achieves the significant advantage of simultaneous prediction of sand body configuration (sand body structure) and sand body thickness, and has excellent computational efficiency, wherein the sand body structure mainly refers to the vertical combination relationship of multiple sand bodies and mudstones of different thicknesses within the target layer (e.g. Figure 7 (2) This invention addresses the problem of seismic response differences caused by vertical structural changes by combining forward modeling with seismic waveform feature analysis to construct a sand body structure classification template. (3) This invention addresses the problem of numerous seismic attributes and complex responses by proposing a method that integrates multi-attribute unsupervised clustering and correlation analysis to eliminate redundant information. (4) The sand body structure and sand body thickness coupled prediction method based on multi-task learning adopted in this invention takes sand body thickness and sand body structure as learning targets simultaneously. During the prediction process, it can autonomously extract sand body structure features and use structural information to constrain the thickness prediction space, achieving global optimization of both sand body thickness and sand body structure, thereby significantly improving prediction accuracy; at the same time, it also takes into account the fine prediction of the planar distribution of sand body thickness and sand body structure. (5) The deep learning network based on residual attention mechanism adopted in this invention has good generalization performance. Through dynamic analysis and adaptive weighting, it can effectively capture the coupling relationship between sand body thickness and structure, and significantly reduce the ambiguity of reservoir prediction through the collaborative constraint mechanism of the two.
[0085] In one embodiment of the present invention, a reservoir configuration-constrained seismic attribute intelligent fusion and reservoir prediction system is provided, comprising:
[0086] The forward modeling module constructs a sand body structure model, extracts actual seismic wavelets from the actual seismic data volume of the study area, and uses the seismic forward modeling method to perform seismic forward modeling simulation on the sand body structure model using the actual seismic wavelets of the study area.
[0087] The classification module, based on seismic forward modeling, classifies the seismic forward modeling waveforms to obtain seismic forward modeling waveform classification results, and classifies the sand body structure in well logging interpretation based on the seismic forward modeling waveform classification results to obtain the sand body structure classification results;
[0088] The prediction module selects seismic attributes sensitive to sand body configuration from the actual seismic data volume of the study area. The selected seismic attributes, sand body thickness interpreted from well logging, and sand body structure are input into a multi-task learning model to perform intelligent analysis of seismic attribute data of the study area based on multi-task learning. This achieves mutual constraints between sand body configuration and sand body thickness, and then enables sand body thickness prediction constrained by sand body configuration.
[0089] In the above embodiments, based on the statistical analysis of core, well logging and seismic data of the study area, the sand body thickness distribution characteristics and sand body stacking relationship of the study area are analyzed to construct a sand body structure model (sand body configuration model); wherein, the sand body structure model is a conceptual model combining sand body thickness and sand body stacking relationship.
[0090] In the above embodiments, the actual seismic wavelet of the study area is extracted from the actual seismic data volume of the study area. Specifically, the seismic statistical wavelet of the target layer is extracted as the actual seismic wavelet of the study area.
[0091] In the above embodiments, the application of actual seismic wavelet waves from the study area to perform seismic forward modeling of the sand body structure model includes:
[0092] By performing seismic forward modeling on sand body structure models, a seismic forward modeling response template covering different sand body thicknesses and stacking relationships is constructed, and geological models with similar seismic responses are grouped into the same category to obtain a sand body structure classification template.
[0093] In the above embodiments, seismic attributes sensitive to sand body configuration are preferably selected, including: using cluster analysis to divide multiple seismic attributes into multiple categories, and then selecting the seismic attribute with the highest correlation between the reservoir and the seismic attribute in each category based on correlation analysis, as the sensitive seismic attribute.
[0094] In the above embodiments, the multi-task learning model includes:
[0095] Input seismic attributes, sand body thickness and sand body structure from well logging interpretation, and use sand body thickness and sand body structure as labels for multi-task learning;
[0096] The multi-task learning uses a shared hidden layer and separate sub-network structures for reservoir structure prediction and thickness prediction, while adding an attention mechanism at the connection points between the shared layer and the sub-network layers.
[0097] Among them, the shared hidden layer is a shared feature extraction layer used to learn the common features of seismic attributes; the sub-network structure layer is an independent task branch layer used to fit sand body thickness and sand body structure parameters respectively; the attention mechanism is used to dynamically allocate attribute weights to strengthen the correlation between sand body configuration and seismic response.
[0098] In the multi-task learning process, it is necessary to minimize the sum of the loss functions of the two task branches, namely sand body thickness and sand body structure, and achieve global optimization for both.
[0099] In the above embodiments, the selected seismic attributes, well-logged interpreted sand body thickness, and sand body structure are input into a multi-task learning model to perform intelligent analysis of seismic attribute data in the study area based on multi-task learning, thereby achieving prediction of sand body thickness and sand body structure, including:
[0100] Based on forward modeling, sand body structure classification, and optimized seismic attributes, an input dataset is constructed, which includes sand body thickness interpreted from well logging, sand body structure classification results, and optimized seismic attributes.
[0101] The sand body thickness and sand body structure classification results interpreted from well logging are used as supervised data for multi-task learning, with sand body thickness as the regression target and sand body structure as the classification target. At the same time, the preferred seismic attribute values near the well access are extracted, and the seismic attribute values of the well access and the sand body thickness and sand body structure classification results interpreted from well logging are used as training datasets.
[0102] A multi-task learning model is used to fuse seismic attributes and predict reservoirs under sand body configuration constraints. By exploring the knowledge correlation between multiple tasks, the synchronous high-precision prediction of sand body structure and thickness is achieved.
[0103] A Bayesian hyperparameter tuning algorithm is used to adaptively balance the weights between tasks, achieving a deep integration of geological knowledge and data-driven approaches, and realizing the global optimization of sand body configuration and thickness in sand body prediction.
[0104] Output the predicted results of the spatial structure and thickness distribution of sand bodies in the study area.
[0105] The system provided in this embodiment is used to execute the above-described method embodiments. For specific processes and details, please refer to the above embodiments, which will not be repeated here.
[0106] In one embodiment of the present invention, a computing device is provided. This computing device can be a terminal and may include a processor, a communication interface, memory, a display screen, and an input device. The processor, communication interface, and memory communicate with each other via a communication bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. When the computer programs are executed by the processor, they implement the methods described in the above embodiments. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals. Wireless communication can be achieved through Wi-Fi, a management network, NFC (Near Field Communication), or other technologies. The display screen can be a liquid crystal display (LCD) or an e-ink display. The input device can be a touch layer covering the display screen, or buttons, a trackball, or a touchpad mounted on the casing of the computing device, or an external keyboard, touchpad, or mouse. The processor can call logical instructions stored in the memory.
[0107] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0108] In one embodiment of the present invention, a computer program product is provided, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, and when the program instructions are executed by a computer, the computer is able to perform the methods provided in the above-described method embodiments.
[0109] In one embodiment of the present invention, a non-transitory computer-readable storage medium is provided, which stores server instructions that cause a computer to perform the methods provided in the above embodiments.
[0110] The computer-readable storage medium provided in the above embodiments has a similar implementation principle and technical effect to the above method embodiments, and will not be described again here.
[0111] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0112] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0113] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for intelligent fusion of reservoir configuration-constrained seismic attributes and reservoir prediction, characterized in that, include: A sand body structure model was constructed. The actual seismic wavelet of the study area was extracted from the actual seismic data volume of the study area. The seismic forward modeling method was used to perform seismic forward modeling on the sand body structure model. Based on seismic forward modeling, the seismic forward modeling waveforms are classified to obtain seismic forward modeling waveform classification results. Based on the seismic forward modeling waveform classification results, the sand body structure in well logging interpretation is classified to obtain the sand body structure classification results. From the actual seismic data volume of the study area, seismic attributes sensitive to sand body configuration and thickness are selected. The selected seismic attributes, sand body thickness and sand body structure interpreted by well logging are input into a multi-task learning model to perform intelligent analysis of seismic attribute data of the study area based on multi-task learning, so as to realize the prediction of sand body thickness and sand body structure. Multi-task learning models include: Input seismic attributes, sand body thickness and sand body structure from well logging interpretation, and use sand body thickness and sand body structure as labels for multi-task learning; The multi-task learning uses a shared hidden layer and separate sub-network structures for reservoir structure prediction and thickness prediction, while adding an attention mechanism at the connection points between the shared layer and the sub-network layers. Among them, the shared hidden layer is a shared feature extraction layer used to learn the common features of seismic attributes; the sub-network structure layer is an independent task branch layer used to fit sand body thickness and sand body structure parameters respectively; the attention mechanism is used to dynamically allocate attribute weights to strengthen the correlation between sand body configuration, sand body thickness and seismic response. In the multi-task learning process, it is necessary to minimize the sum of the loss functions of the two task branches, namely sand body thickness and sand body structure, and achieve global optimization for both. The selected seismic attributes, well-logged interpreted sand body thickness, and sand body structure are input into a multi-task learning model to perform intelligent analysis of seismic attribute data in the study area based on multi-task learning, thereby enabling the prediction of sand body thickness and structure, including: Based on forward modeling, sand body structure classification, and optimized seismic attributes, an input dataset is constructed, which includes sand body thickness interpreted from well logging, sand body structure classification results, and optimized seismic attributes. The sand body thickness and sand body structure classification results interpreted from well logging are used as supervised data for multi-task learning, with sand body thickness as the regression target and sand body structure as the classification target. At the same time, the preferred seismic attribute values near the well access are extracted, and the seismic attribute values of the well access and the sand body thickness and sand body structure classification results interpreted from well logging are used as training datasets. A multi-task learning model is used to fuse seismic attributes and predict reservoirs under sand body configuration constraints. By exploring the knowledge correlation between multiple tasks, the synchronous high-precision prediction of sand body structure and thickness is achieved. A Bayesian hyperparameter tuning algorithm is used to adaptively balance the weights between tasks, achieving a deep integration of geological knowledge and data-driven approaches, and realizing the global optimization of sand body structure and thickness in sand body prediction. Output the predicted results of the spatial structure and thickness distribution of sand bodies in the study area.
2. The method for intelligent fusion of reservoir configuration-constrained seismic attributes and reservoir prediction as described in claim 1, characterized in that, Constructing a sand body structure model includes: Based on statistical analysis of core, well logging and seismic data of the study area, the characteristics of sand body thickness distribution and sand body stacking relationship of the study area were analyzed, and a sand body structure model was constructed. Among them, the sand body structure model is a conceptual model that combines the sand body thickness and the relationship between sand body stacking.
3. The method for intelligent fusion of reservoir configuration-constrained seismic attributes and reservoir prediction as described in claim 1, characterized in that, From the actual seismic data volume of the study area, the actual seismic wavelet of the study area is extracted. Specifically, the seismic statistical wavelet of the target layer is extracted as the actual seismic wavelet of the study area.
4. The method for intelligent fusion of reservoir configuration-constrained seismic attributes and reservoir prediction as described in claim 1, characterized in that, Based on seismic forward modeling, seismic forward modeling waveforms are classified, including: By performing seismic forward modeling on sand body structure models, a seismic forward modeling response template covering different sand body thicknesses and stacking relationships is constructed, and geological models with similar seismic responses are grouped into the same category to obtain a sand body structure classification template.
5. The method for intelligent fusion of reservoir configuration-constrained seismic attributes and reservoir prediction as described in claim 1, characterized in that, The seismic attributes sensitive to sand body configuration and thickness are selected by: using cluster analysis to divide multiple seismic attributes into multiple categories, and then selecting the seismic attributes with the highest correlation between sand body structure and sand body thickness and seismic attributes in each category based on correlation analysis, as sensitive seismic attributes.
6. A reservoir configuration-constrained intelligent fusion and reservoir prediction system, used to implement the reservoir configuration-constrained intelligent fusion and reservoir prediction method as described in any one of claims 1 to 5, characterized in that, include: The forward modeling module constructs a sand body structure model, extracts actual seismic wavelets from the actual seismic data volume of the study area, and uses the seismic forward modeling method to perform seismic forward modeling simulation on the sand body structure model using the actual seismic wavelets of the study area. The classification module, based on seismic forward modeling, classifies the seismic forward modeling waveforms to obtain seismic forward modeling waveform classification results, and classifies the sand body structure in well logging interpretation based on the seismic forward modeling waveform classification results to obtain the sand body structure classification results; The prediction module selects seismic attributes sensitive to sand body configuration and thickness from the actual seismic data volume of the study area. The selected seismic attributes, sand body thickness and sand body structure interpreted by well logging are input into a multi-task learning model to perform intelligent analysis of seismic attribute data of the study area based on multi-task learning, so as to realize the prediction of sand body thickness and sand body structure.
7. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods described in claims 1 to 5.
8. A computing device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods described in claims 1 to 5.