A method for predicting the thickness distribution of carbonate fracture-vuggy reservoirs

Abstract

The present invention relates to the technical field of seismic exploration of oil and gas in complex exploration areas, and specifically to a method for predicting the thickness distribution of carbonate fracture-vuggy reservoirs. The prediction method comprises the following steps: (1) performing well-seismic calibration on the drilled reservoir of the carbonate fracture-vuggy reservoir; (2) performing multi-sensitivity attribute analysis on different reservoirs, and characterizing the time-domain spatial distribution of the reservoir through machine learning; (3) converting the time-domain spatial distribution results into a depth-domain distribution using a velocity model, performing thickness projection calculations at different coordinate positions on a plane, and forming a reservoir thickness distribution plane map; (4) performing relationship fitting between the actual drilled reservoir thickness at multiple sample points and the reservoir thickness data at the corresponding coordinates through machine learning, and correcting the reservoir thickness distribution plane map using a fitting algorithm. The present invention has the advantages of simple method and high accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas seismic exploration in complex exploration areas, and in particular to a method for predicting the thickness distribution of carbonate fracture-vuggy reservoirs. Background Art

[0002] Carbonate fracture-vuggy reservoirs are a type of reservoir. Exploration and development of these reservoirs, particularly in the Ordovician System of the Tarim Basin and the Permian System of the Sichuan Basin, have yielded excellent results. In recent years, advances in geophysical techniques and drilling technology have enabled exploration and development to continue at depth, with initial successes seen in the Cambrian System and deeper carbonate formations. With the depletion of shallow conventional oil and gas reservoirs and rising development costs, carbonate fracture-vuggy reservoirs are undoubtedly becoming a more critical area for future exploration and development.

[0003] However, as the drilling target continues to deepen, the difficulty of characterizing carbonate fracture-vuggy reservoirs continues to increase and the accuracy continues to decrease due to factors such as the severe and rapid decrease in high-frequency attenuation of seismic signals with increasing burial depth, the reservoir undergoing multiple stages of structural fractures and dissolution filling, and the strong heterogeneity of the carbonate rock itself.

[0004] As burial depth increases, the high-frequency attenuation of seismic signals becomes severe, resulting in a significant reduction in seismic quality and resolution. Furthermore, carbonate rocks are inherently brittle and highly heterogeneous, making them susceptible to heterogeneous dissolution and structural fracturing in the later stages of diagenesis. This makes the description and characterization of carbonate fracture-vuggy reservoirs very difficult, and conventional reservoir identification and characterization techniques struggle to accurately depict their distribution characteristics. Furthermore, with increasing drilling depths, drilling and other engineering costs are rapidly increasing, making the search and prediction of large-scale reservoirs a top priority. With the continuous advancement of seismic imaging technology, the use of slicing, attributes, and other technical methods can essentially characterize the lateral boundaries of fracture-vuggy bodies. However, due to factors such as high-frequency absorption and attenuation of seismic waves, seismic wavelet resolution, and the coherence of multiple geological bodies, the vertical resolution of fracture-vuggy bodies is insufficient, making the vertical thickness of fracture-vuggy anomalies the key to predicting reservoir size.

[0005] Previous reservoir thickness prediction methods have mostly relied on well logging, weighted averaging or grid interpolation of reservoir thickness, combined logging and seismic inversion, and attribute analysis. However, due to the strong vertical and lateral heterogeneity of fracture-vuggy carbonate reservoirs, these methods fail to effectively incorporate seismic information and cannot adequately reflect the spatial variation of the reservoir. Furthermore, due to the limitations of seismic resolution and different threshold settings for inversion results, reservoir thickness predictions are relatively accurate around wells, but exhibit significant errors further away from the wells. Beyond well logging and combined inversion, simply using attribute predictions, even with the inclusion of well-seismic calibration information, remains difficult to accurately predict the vertical thickness distribution of fracture-vuggy anomalies due to factors such as high-frequency absorption and attenuation of deep seismic waves and seismic wavelet resolution. Vuggy oil and gas reservoirs are undoubtedly becoming a more critical area for subsequent exploration and development.

[0006] Due to strong heterogeneity factors such as the significant reduction in seismic resolution and the diversity of reservoir types, the description and characterization of deep carbonate fracture-vuggy reservoirs is quite difficult. However, with the continuous advancement of seismic imaging technology, the lateral boundaries of fracture-vuggy bodies can basically be characterized using technical methods such as slicing and attributes. However, due to factors such as high-frequency absorption and attenuation of seismic waves, seismic wavelet resolution, and coherence of multiple geological bodies, the vertical resolution of fracture-vuggy bodies is insufficient. Therefore, the vertical thickness of fracture-vuggy anomalies becomes the key to reservoir size prediction.

[0007] Therefore, it is very necessary to develop a method for predicting the thickness distribution of carbonate fracture-vuggy reservoirs that can solve the above technical problems. Summary of the Invention

[0008] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a method for predicting the thickness distribution of carbonate fracture-vuggy reservoirs with a simple method and high accuracy.

[0009] The present invention is achieved through the following technical solutions:

[0010] A method for predicting the thickness distribution of carbonate fracture-vuggy reservoirs comprises the following steps:

[0011] (1) Conduct well-seismic calibration on the drilled reservoirs of carbonate fracture-cavity reservoirs;

[0012] (2) Analyze the multi-sensitive attributes of different reservoirs, integrate and analyze the multi-sensitive attributes of different reservoirs through machine learning, and then characterize the temporal and spatial distribution of the reservoirs;

[0013] (3) Using the velocity model, the time domain spatial distribution results of different reservoir bodies are converted into depth domain distribution, and then the thickness projection calculation at different coordinate positions is performed to form a reservoir thickness distribution plane map;

[0014] (4) Through machine learning, the relationship between the actual drilling reservoir thickness of multiple sample points and the reservoir thickness data at the corresponding coordinates is fitted, and the reservoir thickness distribution plane map is corrected using the fitting algorithm to achieve the thickness distribution prediction of the reservoir body.

[0015] Based on the diversity and strong heterogeneity of carbonate reservoir types, sensitive attribute analysis is carried out on different types of reservoirs such as caves, pores, and fractures on the basis of well-seismic calibration of drilled reservoirs. Through machine learning, the multi-sensitive attributes of reservoirs of different reservoir types are integrated and analyzed to form integrated attributes to characterize the reservoir aggregates formed by the three types of reservoirs in the study area, and complete the time domain and spatial distribution characterization of different types of reservoir aggregates.

[0016] Furthermore, in step (2), the sensitive attribute analysis of caves includes: wave impedance inversion, tensor attributes, energy attributes, and amplitude attributes; the sensitive attribute analysis of holes includes: wave impedance, root mean square amplitude, and chaotic attributes; and the sensitive attribute analysis of cracks includes: eigenvalue coherence, coherent AFE, and ant body.

[0017] Furthermore, the distribution characteristics of the screened sensitive attributes are used as models for machine learning, thereby using machine learning for integrated analysis.

[0018] Furthermore, the temporal spatial distribution characterization of the reservoir in step (2) is to establish a reservoir calibration result model based on the correspondence between the reservoir and the multi-sensitive attributes, and to predict the three-dimensional spatial sculpture of the same type of reservoir at other locations based on this model.

[0019] Furthermore, the velocity model is a depth domain imaging velocity model, and the specific method of converting the time domain spatial distribution results of different reservoirs into depth domain distribution is a map migration method.

[0020] Furthermore, the thickness projection calculation process is specifically as follows: thickness projection is performed by superimposing the thickness of sample points with the same X and Y values ​​but different Z values ​​in the three-dimensional space of the depth domain.

[0021] The velocity model is used to convert the spatial distribution results of different types of reservoir aggregates in the time domain into depth domain distribution. On this basis, the thickness projection calculation of different coordinate positions on the plane is completed to form a reservoir thickness distribution plane map described by integrated attributes.

[0022] Furthermore, the correction is to use the actual drilling thickness data to correct the predicted thickness to obtain the corrected true thickness of the reservoir; the specific method is: to obtain the correction coefficient at the well point, then to obtain the plane correction coefficient map, and finally to perform a unified correction in the study area.

[0023] Machine learning is a multi-disciplinary interdisciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithmic complexity theory, and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their own performance. In this application, the geometric, physical, and other properties of various reservoir types are integrated, classified, and optimized using the comprehensive characteristics of probability theory, statistics, and other disciplines for different types of reservoirs, thereby achieving the purpose of attribute optimization. The process of integrated analysis is the process of machine learning. By establishing models of various types of reservoirs, machine recognition and attribute optimization are performed, thereby selecting the attributes and results with the best description effect.

[0024] Based on the interpretation of different types of reservoirs such as caves, holes, and fractures through drilling, measurement, and logging, this application determines the development position and basic response characteristics of the seismic profile reservoir through well depth calibration, extracts multiple types of attributes in a targeted manner, and integrates different targeted sensitive attributes using machine learning to complete the time domain spatial distribution characterization of different types of reservoir aggregates. The time domain spatial distribution morphology of different types of reservoir aggregates is converted into a depth domain distribution using a graph migration method using a depth domain imaging velocity model. On this basis, the thickness projection calculation at different coordinate positions on the plane is completed to form a reservoir thickness distribution plane map. Machine learning is then used to fit the relationship between the actual drilled reservoir thickness of multiple sample points and the reservoir thickness data characterized by the integration of different sensitive attributes at the corresponding coordinates. Finally, a fitting algorithm is used to convert the reservoir thickness distribution plane map characterized by the integration of different sensitive attributes using machine learning into a more accurate reservoir thickness distribution plane map, thereby realizing the prediction of reservoir thickness distribution in the study area.

[0025] The beneficial effects of the present invention are:

[0026] (1) The use of machine learning to integrate multiple attributes sensitive to different types of reservoirs, such as caves, pores, and fractures, avoids the limitations of single-attribute prediction and the manual selection of attributes that mainly characterize visible abnormal responses while ignoring more hidden information in seismic data. It can better describe the spatial distribution of different types of reservoir aggregates.

[0027] (2) Using the depth domain imaging velocity model, the time domain spatial distribution morphology of different types of reservoir aggregates is converted into depth domain distribution using the map migration method, avoiding the error caused by using the same replacement velocity conversion. The thickness model foundation of the converted reservoir aggregate is more superior.

[0028] (3) The machine learning method is used to fit the reservoir thickness data of multiple sample points and the reservoir thickness data characterized by the integration of multiple types of attributes. The fitting relationship is used to further realize the correction of the reservoir thickness distribution plane map, making the reservoir thickness description more accurate. It can be used as an important indicator for the scale evaluation of carbonate fracture-vuggy reservoirs in the study area. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 Schematic diagram of beaded reflection;

[0030] Figure 2 Schematic diagram of chaotic reflection;

[0031] Figure 3 Schematic diagram of linear weak reflection;

[0032] Figure 4 This is a schematic diagram of the spatial distribution of cave bodies;

[0033] Figure 5 Schematic diagram of the spatial distribution of pores;

[0034] Figure 6 Schematic diagram of the spatial distribution of fracture bodies;

[0035] Figure 7 Schematic diagram of the spatial distribution of cave + hole + crack fusion;

[0036] Figure 8 Schematic diagram of cave plane thickness. DETAILED DESCRIPTION

[0037] The present invention will be further described below with reference to specific embodiments, and the advantages and features of the present invention will become clearer as the description proceeds. However, these embodiments are merely exemplary and do not constitute any limitation to the scope of the present invention. It should be understood by those skilled in the art that the details and forms of the technical solutions of the present invention may be modified or replaced without departing from the spirit and scope of the present invention, and such modifications and replacements fall within the scope of protection of the present invention.

[0038] Example

[0039] An example of the proven reserves calculation process for a carbonate fracture-vuggy reservoir in a certain work area in northern Tarim:

[0040] This process is implemented in accordance with the latest revised National Reserves Commission's specifications for carbonate fracture-cavity reservoir reserves calculation using the carving method. The specific implementation is as follows:

[0041] (1) Clarify the target earthquake identification model in the study area: classify the earthquake phases in the work area and establish a model, and use statistics and calibration to identify the target earthquakes. Figure 1-3 Three high-probability seismic phases that can represent caves, holes, and cracks are used as samples for machine learning.

[0042] (2) Using machine learning to integrate multiple attributes sensitive to different types of reservoirs, such as caves, holes, and fractures, it avoids the limitations of single-attribute prediction and avoids manual selection of attributes that mainly characterize visible abnormal responses while ignoring more hidden information in seismic data. Machine learning uses multiple attributes to carve out various seismic phases and depict spatial distributions such as Figure 4-7 shown.

[0043] (3) According to the three-dimensionally carved reservoir body, the time domain spatial distribution morphology of different types of reservoir aggregates is converted into depth domain distribution by using the depth domain imaging velocity model and the map migration method. The seismic-geological bridge is established through the porosity curve, and the porosity body is constructed by the phase-controlled inversion method.

[0044] (4) Realize the vertical thickness and average porosity of various reservoirs, build a model based on the above reservoir characteristics and bring it into machine learning, use machine learning to fit the reservoir thickness data of multiple sample points and the reservoir thickness data characterized by the integration of multiple types of attributes, and use the fitting relationship to further realize the correction of the reservoir thickness distribution plane map. This method uses the method of multiple attributes and weighted functions to obtain the plane attribute thickness, which is different from the traditional structural map compiled by plane structure. Figure 8 As shown in the figure, taking the cave thickness map as an example, the calibration accuracy of the actual drilling is 89%.

[0045] The above detailed description is a specific description of one feasible embodiment of the present invention. This embodiment is not intended to limit the patent scope of the present invention. Any equivalent implementation or modification that does not depart from the present invention should be included in the scope of the technical solution of the present invention.

Claims

1. A method for predicting the thickness distribution of carbonate fracture-vuggy reservoirs, characterized in that: The steps include: (1) Conduct well-seismic calibration on the drilled reservoirs of carbonate fracture-vuggy reservoirs; (2) Analyze the sensitive attributes of different reservoirs, integrate and analyze the sensitive attributes of different reservoirs through machine learning, and then characterize the temporal and spatial distribution of the reservoir; (3) Using the velocity model, the spatial distribution results of different reservoir bodies in the time domain are converted into depth domain distribution. Then, the thickness projection calculation is performed at different coordinate positions on the plane to form a reservoir thickness distribution plane map. The seismic-geological bridge is established through the porosity curve, and the porosity volume is constructed through the phase-controlled inversion method. (4) The vertical thickness and average porosity of various reservoirs are realized. The model constructed with the above reservoir characteristics is brought into machine learning. The relationship between the actual drilling reservoir thickness of multiple sample points and the reservoir thickness data at the corresponding coordinates is fitted through machine learning. The reservoir thickness distribution plane map is corrected using the fitting algorithm to realize the thickness distribution prediction of the reservoir. The correction steps in step (4) are specifically as follows: obtaining the correction coefficient at the well point, then obtaining the plane correction coefficient map, and finally performing a unified correction in the study area.

2. The prediction method according to claim 1, characterized in that The types of reservoirs described in step (1) include caves, pores and fractures.

3. The prediction method according to claim 2, characterized in that In step (2), the sensitive attribute analysis of caves includes: wave impedance inversion, tensor attributes, energy attributes, and amplitude attributes; the sensitive attribute analysis of holes includes: wave impedance, root mean square amplitude, and chaotic attributes; the sensitive attribute analysis of cracks includes: eigenvalue coherence, coherent AFE, and ant body.

4. The prediction method according to claim 1, wherein: The temporal spatial distribution characterization of the reservoir in step (2) is to establish a reservoir calibration result model based on the correspondence between the reservoir and the multi-sensitive attributes, and to predict the three-dimensional spatial sculpture of the same type of reservoir at other locations based on this model.

5. The prediction method according to claim 1, wherein: The velocity model in step (3) is a depth domain imaging velocity model; the specific method for converting the time domain spatial distribution results of different reservoirs into depth domain distribution is a map migration method.

6. The prediction method according to claim 1, characterized in that The thickness projection calculation process described in step (3) is as follows: the thickness of the sample points with the same X and Y values ​​but different Z values ​​in the three-dimensional space of the depth domain is superimposed, which is the thickness projection.

Images