A reservoir prediction method based on structure tensor

Abstract

The present invention discloses a reservoir prediction method based on structural tensors, which includes: Step 1, calculating a structural gradient tensor attribute volume on interpretive result data to obtain three eigenvalues λ1, λ2, and λ3; Step 2, scanning all sampling points to perform tensor thinning calculation to determine the starting mutations of fractures and cavities and the positions of primary fracture fragments; Step 3, performing seismic inversion with structural tensor attribute constraints to describe the characteristics of internal holes in the reservoir; Step 4, organically combining the structural tensor attributes for distinguishing and predicting the internal cave contours of the reservoir, the tensor thinning attributes of the starting mutations of fractures and cavities and the positions of primary fracture fragments, and the structural tensor-based inversion volume of the internal characteristics of the reservoir to perform multi-information three-dimensional stereoscopic carving display to clarify the spatial distribution characteristics of the reservoir. The present invention can quickly and effectively distinguish and predict different types of reservoirs with holes, fractures, and cavities, avoiding the defects of unscientific collocation and weak pertinence among various attributes in previous reservoir predictions.

Description

Technical Field

[0001] The present invention relates to a reservoir prediction method based on structure tensor, belonging to the technical field of seismic exploration. Background Art

[0002] In recent years, in the field of oil and gas exploration and development, evaluating the reservoir performance of carbonate reservoirs and predicting the internal spatial distribution law of reservoirs have become the key points and difficulties. Due to the fact that the pore structure of carbonate rocks is more complex in terms of morphology and distribution characteristics compared with conventional sandstone reservoirs, the reliability of conventional rock physics empirical relationships is reduced, and it is difficult to master the variation law of the key elastic parameter of longitudinal wave velocity, which affects the accuracy of reservoir velocity modeling, resulting in an increase in the difficulty of describing the internal characteristics of reservoirs and comprehensive prediction. The carbonate reservoir bodies in the LS area are mainly distributed in the Yijianfang Formation - Penglaiba Formation of the Ordovician System, and develop fracture-cave type reservoirs controlled by faults, with good reservoir performance. They form a good reservoir-caprock combination with overlying tight limestone, marlstone, mudstone and other caprocks in the upper Ordovician System. Industrial oil and gas flows have been obtained from multiple wells in the Yijianfang - Yingshan Formation. For example, Well LS1 deployed on the main fault zone encountered blowout and lost circulation, and obtained high-yield industrial oil and gas flows. Therefore, the fracture-cave type reservoirs controlled by faults in the Yijianfang - Penglaiba Formation of the Ordovician System have become the main type of carbonate reservoir prediction in the LS work area.

[0003] Regarding the fracture-cavity body reservoirs controlled by faults, many scholars have conducted a large number of studies, and gradually formed sensitive attributes and prediction methods for different types of reservoirs such as holes, caves and fractures, including AFE, Likelihood, amplitude change rate, coherent energy gradient, structure tensor, etc. Commonly used instantaneous energy, discontinuity and AFE are respectively used to conduct three-dimensional carving on holes, caves and fractures, and good results have been achieved. However, there are still problems such as the overly single attributes selected for reservoir prediction, the unscientific collocation among multiple attributes, and the insufficient close connection with the internal target. It is necessary to strengthen the research on comprehensive reservoir prediction technology, especially the in-depth research and scientific utilization of sensitive attributes, improve the technical pertinence and process rationality, so as to improve the prediction and description accuracy of the internal structure of reservoirs and effectively guide the well trajectory design.

[0004] In summary, reservoir prediction technology comprehensively infers the existence, spatial shape, and oil and gas content of unknown underground reservoirs from existing and limited seismic, logging, and drilling data, which directly affects trap description and target selection. However, due to the unscientific reservoir prediction attributes, low accuracy, and diversity of prediction results, the subsequent target selection and well location design cannot achieve ideal effects. Although traditional methods have higher accuracy, for different types of fracture-vug bodies such as holes, caves, and fractures, using conventional reservoir prediction technology usually can only predict the comprehensive type or single type response of the reservoir, but cannot predict and distinguish different types of hole-cave-fracture reservoirs. There are still defects such as single sensitive attributes, overly complex software, and unscientific attribute collocations, which cannot meet the need for rapid and effective discrimination and prediction of different types of hole-cave-fracture reservoirs. Especially in areas with a relatively fast progress of new 3D and rolling exploration, there is an urgent need for a rapid and effective method to distinguish and predict different types of hole-cave-fracture reservoirs, so as to achieve the purpose of fine description of traps and optimal selection of exploration targets. Summary of the Invention

[0005] The present invention provides a reservoir prediction method based on structure tensor. This method is based on the structural gradient tensor. On the basis of describing the internal contour of the reservoir, mathematical and inversion transformation calculations are performed on the structural gradient tensor to achieve the purpose of distinguishing and predicting the contours and internal structural characteristics of different types of hole-cave-fracture reservoirs.

[0006] The technical solution adopted by the present invention is: a reservoir prediction method based on structure tensor, which includes the following steps:

[0007] Step 1: On the interpretive result data, use the backbone profile to view the geological conditions and reservoir distribution laws of the entire study area, and calculate the structural gradient tensor attribute volume to obtain three eigenvalues, λ1, λ2, and λ3;

[0008] Step 2: On the basis of the structural gradient tensor attribute volume calculated from the seismic amplitude data volume, scan all sampling points and perform tensor thinning calculation to further determine the starting mutation of fractures and vugs and the positions of primary fracture fragments;

[0009] Step 3: Establish a low-frequency model with the structure tensor and tensor thinning as constraints, and then use this low-frequency model as a constraint condition to perform structural tensor attribute-constrained seismic inversion to describe the internal characteristics of the reservoir and describe the internal hole characteristics of the reservoir;

[0010] Step 4: Organically combine the structural tensor attributes for distinguishing and predicting the internal cave contours of the reservoir, the tensor thinning attributes of the starting mutation of fractures and vugs and the positions of primary fracture fragments, and the inversion body based on the structure tensor of the internal characteristics of the reservoir, and perform multi-information three-dimensional stereo carving display to clarify the spatial distribution characteristics of the reservoir and provide guidance for trap description and exploration target selection.

[0011] Further, the specific steps of the first step include:

[0012] (1) Gradient calculation: Perform gradient calculation on the interpretive result data to calculate the gradient vector of the energy at each point of the 3D seismic data volume;

[0013] (2) Construct the structure tensor: Smooth each component using a Gaussian window, characterize the texture features within a certain area with the average value, and suppress the sudden change of the structure tensor caused by noise;

[0014] (3) Calculate the result eigenvalues and eigenvectors: Select the key area calculation window of the smoothed average structure tensor attribute volume for test comparison and analysis to select a suitable calculation window; Use the optimized parameters to perform different vector calculations on the entire average structure tensor attribute volume to obtain three eigenvalues λ1, λ2, λ3 and eigenvectors.

[0015] Further, select the structure gradient tensor eigenvalue λ2 for geological target reservoir prediction to determine its contour features.

[0016] Further, the specific calculation of the step (2) is to calculate the variance first and then calculate the average tensor:

[0017] Define the variance E of the 3D seismic image as:

[0018]

[0019] Where l(x, y, z) is the 3D seismic amplitude function, and x, y, z are the line, trace position, and two-way travel time of the 3D seismic voxel respectively; w(x, y, z) is the Gaussian window function;

[0020] Given a small displacement, expand the function E according to the Taylor series, calculate and organize to obtain the structure tensor T1, and convolve the obtained gradient structure tensor with the Gaussian kernel function to obtain the regional average gradient structure tensor T2:

[0021]

[0022] Where T1 is the structure tensor, T2 is the average structure tensor, and g x 、g y 、g z Are the directional derivatives along the x, y, z directions respectively; * is the convolution operator; G is the Gaussian kernel function.

[0023] Further, scan all sampling points, calculate the strike and dip of each sampling point along the structural layer, count all tensor attribute values within a certain step range, and only retain the maximum tensor attribute value for each sampling point, and set the non-maximum tensor value to 0.

[0024] Furthermore, the specific calculation steps are as follows:

[0025] (1) Perform coherence calculation on the interpretive result data to obtain the dip angle and azimuth angle of the seismic volume through coherence calculation;

[0026] (2) Calculate the structural gradient tensor of the seismic volume;

[0027] (3) Conduct tensor thinning calculation under the constraints of the structural dip angle and azimuth angle.

[0028] Furthermore, the calculated new data volume records the most likely positions of the fractures and caves and the tensor values of the fracture and cave development at these positions. This data volume is the tensor thinning attribute volume. Then, reservoir prediction is performed on the tensor thinning attribute volume to further determine the starting mutations of the fractures and caves and the positions of the primary fracture fragments, and to describe the internal fracture characteristics of the reservoir.

[0029] The beneficial effects of the present invention are as follows: This method makes full use of the OPN-GS software and the sensitive attributes of the structural gradient tensor, deeply explores the advantages of the structural gradient tensor, and performs mathematical and inversion transformation calculations on it. It can quickly and effectively distinguish and predict different types of reservoir with holes, fractures and caves, avoiding the defects of unscientific combination and weak pertinence among multiple attributes in previous reservoir prediction. Moreover, the operation software is single and has strong popularization, which has very important significance. The improvement and popularization of this method can effectively predict the fault-controlled fracture and cave body reservoir, guide the later development and well location design, which has very important significance and broad prospects.

[0030] Additional aspects and advantages of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is the flow chart of the reservoir prediction method based on structural tensor of the present invention;

[0032] Figure 2 is the comparative analysis of different calculation windows of the structural tensor;

[0033] Figure 3 is the calculation of three eigenvalues of the structural tensor;

[0034] Figure 4 is the overlay map of the original profile + structural tensor + tensor thinning;

[0035] Figure 5 is the comparison between the conventional inversion profile and the inversion profile constrained by the structural tensor;

[0036] Figure 6 is the three-dimensional stereoscopic carving map constrained by the structural tensor with multi-information;

[0037] Figure 7It is a three-dimensional stereo carving map with multi-information constrained by the structure tensor in the YS area. Specific implementation mode

[0038] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] Figure 1 As shown, a reservoir prediction method based on the structure tensor includes the following steps:

[0040] Step 1: On the interpretive result data, use the backbone profile to view the geological conditions and reservoir distribution laws of the entire study area, and calculate the structural gradient tensor attribute volume to obtain three eigenvalues, λ1, λ2, and λ3. Optimize the combination of structural gradient tensor eigenvalues for geological target reservoir prediction, and determine the internal cave contour of the reservoir.

[0041] (1) Perform gradient calculation on the interpretive result data to obtain the gradient vector (direction derivative) of the energy at each point of the 3D seismic data volume.

[0042] (2) Calculate the average gradient structure tensor, construct the structure tensor T, smooth each component using a Gaussian window, and use the average value to characterize the texture features within a certain area, and suppress the mutation of the structure tensor caused by noise.

[0043] The specific calculation is as follows: First, calculate the variance E based on the first-step direction derivative:

[0044]

[0045] In the formula, l(x, y, z) is the 3D seismic amplitude function, and x, y, z are the line, trace position, and two-way travel time of the 3D seismic voxel respectively; w(x, y, z) is the Gaussian window function. Given a small displacement, expand the attribute function I according to the Taylor series, calculate and organize to obtain the structure tensor T1, and convolve the obtained gradient structure tensor with the Gaussian kernel function to obtain the regional average gradient structure tensor T2:

[0046]

[0047] In the formula, T1 is the structure tensor, T2 is the average structure tensor, gx, gy, and gz are the direction derivatives along the x, y, and z directions respectively; * is the convolution operator; G is the Gaussian kernel function. The significance of calculating the average gradient structure tensor lies in the following two aspects:

[0048] a. Use the average gradient structure tensor to characterize the texture features within a certain area;

[0049] b. Used to suppress the mutation of the structure tensor caused by noise.

[0050] (3) Calculate the result eigenvalues and eigenvectors: Select the key area of the smoothed average structural tensor attribute volume and conduct test and comparative analysis using various calculation windows such as 3-3-3, 5-5-7, 7-7-9, 11-11-11, etc., and preferably select a suitable calculation window ( Figure 2 ); Use the preferably selected parameters to perform different vector calculations on the entire average structural tensor attribute volume to obtain three eigenvalues λ1, λ2, λ3 and eigenvectors, and preferably select the best eigenvalue that can finely describe the internal cave contour of the reservoir ( Figure 3 ).

[0051] Step 2: Based on the structural gradient tensor attribute volume calculated from the seismic amplitude data volume, scan all sampling points to perform tensor thinning calculation to further determine the starting mutation of fractures and cavities and the position of primary fracture fragments.

[0052] (1) Scan all sampling points, calculate the strike and dip of each sampling point along the tectonic layer, count all tensor attribute values within a certain step range, and only retain the maximum tensor attribute value for each sampling point, while setting non-maximum tensor values to 0. The specific calculation steps are as follows:

[0053] a. Conduct coherence calculation on the interpretive result data to obtain the dip and azimuth of the seismic volume through coherence calculation;

[0054] b. Calculate the structural gradient tensor of the seismic volume;

[0055] c. Conduct tensor thinning calculation under the constraints of the tectonic dip and azimuth.

[0056] The calculated new data volume is equivalent to recording the most likely positions of fractures and cavities and the tensor values of fracture and cavity development at these positions. This new data volume is the tensor thinning attribute volume. At the positions where fractures and cavities develop, the attributes of the tensor thinning volume are more refined than the tensor attributes ( Figure 4 ).

[0057] (2) Conduct reservoir prediction for the tensor thinning attribute volume to further determine the starting mutation of fractures and cavities and the position of primary fracture fragments, and describe the internal fracture characteristics of the reservoir

[0058] Step 3: Establish a low-frequency model with the structural tensor and tensor thinning as constraints, and then use this low-frequency model as a constraint condition to perform structural tensor attribute-constrained seismic inversion to describe the internal characteristics of the reservoir and the internal hole characteristics of the reservoir ( Figure 5 ).

[0059] Step 4: Combine the structural tensor attributes for differentiating the internal cave contours of the predicted reservoir, the tensor thinning attributes for differentiating the starting mutations of fractures and vugs and the positions of the primary fracture fragments, and the structure-tensor-based inversion volume of the internal characteristics of the reservoir, and perform multi-information three-dimensional stereoscopic carving display to clarify the spatial distribution characteristics of the reservoir and provide guidance for trap description and exploration target optimization. Figure 6 )

[0060] For example Figure 7 , to verify the application effect of the new method for reservoir prediction based on structural tensors of this patent, a comparative analysis is carried out on the single attributes in the past and the current structural tensors, tensor-constrained wave impedance inversion, and tensor thinning multi-attribute fusion stereoscopic carving in the YS study area. It can be seen that the structural tensors accurately depict the contour characteristics of the caves, the tensor-constrained inversion finely reflects the internal characteristics of the pores and caves, the tensor thinning shows the distribution characteristics of the fractures, and the internal structural characteristics of different types of reservoirs with pores, fractures, and caves described by the fused attributes are finer and more prominent than the single attributes in the past, and the internal connectivity of the reservoir is clearer. Combining the quantitative calculation results of the carved volume, the favorable development areas of faults and fracture-vug reservoirs are identified, the favorable reservoir development zones and exploration targets are optimized, and the well trajectory design is effectively guided.

[0061] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those of ordinary skill in the art should understand that the above embodiments do not limit the protection scope of the present invention in any form. Any technical solutions obtained by means of equivalent replacement and the like fall within the protection scope of the present invention. The parts not involved in the present invention are the same as or can be implemented by the prior art.

Claims

1. A reservoir prediction method based on structure tensor, characterized in that It includes the following steps: Step 1: On the interpretive result data, use the backbone profile to view the geological conditions and reservoir distribution laws of the entire study area, and calculate the structural gradient tensor attribute volume to obtain three eigenvalues, namely λ1, λ2, and λ3; Step 2: Based on the structural gradient tensor attribute volume calculated from the seismic amplitude data volume, scan all sampling points to perform tensor thinning calculation to further determine the starting mutation of fractures and cavities and the positions of primary fracture fragments; Step 3: Establish a low-frequency model with the structural tensor and tensor thinning as constraints, and then use this low-frequency model as a constraint condition to perform structural tensor attribute-constrained seismic inversion to describe the pore characteristics inside the reservoir; Step 4: Combine the structural tensor attributes that distinguish the internal cave contours of the predicted reservoir, the tensor thinning attributes of the starting mutation of fractures and cavities and the positions of primary fracture fragments, and the inversion volume based on the structural tensor of the internal characteristics of the reservoir organically, and perform multi-information three-dimensional stereo carving display to clarify the spatial distribution characteristics of the reservoir, providing guidance for trap description and exploration target optimization.

2. The method for reservoir prediction based on structural tensor according to claim 1, wherein The specific steps of the said Step 1 include: (1) Gradient calculation: Perform gradient calculation on the interpretive result data to calculate the gradient vector of the energy at each point of the 3D seismic data volume; (2) Construct the structural tensor: Smooth each component with a Gaussian window, use the average value to characterize the texture features within a certain area, and suppress the mutation of the structural tensor caused by noise; (3) Calculate the result eigenvalues and eigenvectors: Select the key area calculation window of the smoothed average structural tensor attribute volume for test and comparative analysis to select a suitable calculation window; Use the optimized parameters to perform different vector calculations on the entire average structural tensor attribute volume to obtain three eigenvalues, namely λ1, λ2, and λ3, and eigenvectors.

3. A structural tensor-based reservoir prediction method according to claim 1 or 2, characterized in that Select the structural gradient tensor eigenvalue λ2 for geological target reservoir prediction to determine its contour features.

4. The method for reservoir prediction based on structure tensor according to claim 2, wherein The specific calculation of the said step (2) is to calculate the variance first and then the average tensor: Define the variance E of the 3D seismic image as: ， where l(x, y, z) is the 3D seismic amplitude function, and x, y, z are the line, trace position, and two-way travel time of the 3D seismic voxel respectively; w(x, y, z) is the Gaussian window function; Given a small displacement, expand the function E according to the Taylor series, calculate and organize to obtain the structural tensor T1, and convolve the obtained gradient structural tensor with the Gaussian kernel function to obtain the regional average gradient structural tensor T2: ， where T1 is the structure tensor, T2 is the average structure tensor, and g x , g y , g z are the directional derivatives along the x, y, and z directions, respectively; * is the convolution operator; and G is the Gaussian kernel function.

5. A reservoir prediction method based on structure tensor according to claim 1, characterized in that, In the said Step 2, scan all sampling points, calculate the strike and dip of each sampling point along the tectonic layer, count all tensor attribute values within a certain step range, and only retain the maximum tensor attribute value for each sampling point, and set the non-maximum tensor value to 0.

6. The method for reservoir prediction based on structure tensor according to claim 5, wherein The specific calculation steps are as follows: (1) Perform coherence calculation on the interpretive result data to obtain the dip and azimuth of the seismic volume through coherence calculation; (2) Calculate the structural gradient tensor of the seismic volume; (3) Perform tensor thinning calculation under the constraints of the tectonic dip and azimuth.

7. A reservoir prediction method based on structure tensor according to claim 6, characterized in that, The calculated new data volume records the most likely positions of the fractures and vugs and the tensor values of the fracture and vug development at these positions. This data volume is the tensor thinning attribute volume. Then, reservoir prediction is carried out for the tensor thinning attribute volume to further determine the starting mutations of the fractures and vugs and the positions of the primary fracture fragments, and to describe the internal fracture characteristics of the reservoir.

Images