CN115577616A - Method and device for seismic characterization of carbonate fracture-caves based on deep learning - Google Patents

Abstract

The invention discloses a carbonatite fracture hole seismic characterization method and device based on deep learning, and belongs to the technical field of oil and gas exploration and development. The method comprises the following steps: acquiring seismic data; horizontally stacking the seismic data to obtain a post-stack migration seismic data volume; extracting a root-mean-square amplitude data volume and a seismic sweet-spot attribute data volume based on the post-stack migration seismic data volume; picking up learning points on a seismic section obtained by horizontally stacking seismic data; inputting training samples and learning targets into a learning vector machine for training to obtain a deep learning network, wherein the training samples are a post-stack migration seismic data body, square root amplitude data and seismic dessert attributes, and the learning targets are learning points; and obtaining a fracture-cave characteristic attribute data volume by utilizing a deep learning network based on the post-stack migration seismic data volume, the square root amplitude data and the seismic dessert attribute. The method of the invention not only can well predict the fracture and the hole, but also has good prediction effect on the underground corrosion hole.

Description

Method and device for seismic characterization of carbonate fracture-caves based on deep learning

technical field

The invention belongs to the technical field of oil and gas exploration and development, and in particular relates to a method and device for seismic characterization of carbonate fracture-caves based on deep learning.

Background technique

Carbonate oil and gas reservoirs are one of the main types of oil and gas reservoirs in my country. The formation of carbonate rock reservoirs is mainly affected by karst reformation and faulting in the later stage, forming secondary dissolved pores and fractures, and the storage space is secondary dissolved pores and caves. A system of fractures and cracks. Fracture-vuggy reservoirs in my country's oil and gas fields have the following characteristics: paleokarst has obvious vertical zoning, surface karst zone, vertical vadose zone and horizontal underflow zone are well developed; the reservoir space is mainly semi-filled or unfilled residual large karst caves formed by karstification Composed of dissolved pores, caves and fractures, the high-quality reservoir types are mainly fractures-dissolved pores-large karst caves, which are the most important reservoirs and main production layers for high and stable production in major oil and gas fields; the reservoirs are obviously controlled by paleokarst landforms and fault fractures, Karst slopes and fault development areas are the most favorable areas for reservoir development; secondary pores formed by buried organic dissolution are also important effective pores, and their development matches the formation, evolution, and migration and accumulation of hydrocarbons; supergene karst and buried organic The multi-stage superimposition and transformation of dissolution is the best combination mode for the formation of paleokarst reservoirs and oil and gas reservoirs.

At present, using seismic data to predict carbonatite fractures and caves mainly adopts the following technologies for prediction and identification: First, the anisotropic fracture prediction method is used to predict the development zone of fractures and caves. This prediction technology needs to collect seismic data in wide azimuths to overcome the impact of noise on the inversion results. However, the conventional seismic data are often not detailed enough, which makes the prediction of the fracture-cavity development zone difficult to meet the needs of development and production. The prediction is inaccurate and takes a long time, and there are many situations where the prediction of the fracture-vug system does not match the actual drilling production; Seismic characterization method of fractured-cavern structures in carbonate rocks. The feature of this method is mainly to use the lateral discontinuity characteristics of seismic data to predict the development area and scale of carbonate rocks on the basis of fault and fracture prediction. The advantages of this method It is better for the prediction of fractures and caves related to the development of fractures and fractures, but its disadvantage is that it is not good for the prediction of carbonatite caves formed by the dissolution of underground runoff and groundwater; the third is the method of estimating the fracture density around the well. It mainly uses imaging logging technology (imaging logging is mainly used for fracture prediction) and seismic attributes to establish training sets and test sets to predict fracture development areas. This technique is also effective in predicting fractures, but it is not effective in predicting dissolution pores.

Generally speaking, the current carbonatite-salt fracture-cavity prediction technology has a good prediction effect on fault fractures and caves, but the prediction effect on dissolution-type pores is not good. However, in my country's carbonate rock oil and gas exploration, dissolution vugs are also one of the fields that are widely developed and have good exploration results. Therefore, the prediction technology of carbonate rock dissolution vugs can well solve the problem of carbonate rock dissolution vug oil and gas exploration.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the prior art and provide a method and device for seismic characterization of carbonate fracture-caves based on deep learning.

The purpose of the present invention is achieved through the following technical solutions:

According to the first aspect of the present invention, the carbonate fracture-vuggy seismic characterization method based on deep learning includes:

Obtain seismic data;

performing horizontal stacking on the seismic data to obtain a post-stack migration seismic data volume;

Extracting a root mean square amplitude data volume and a seismic sweet spot attribute data volume based on the post-stack migration seismic data volume;

Picking learning points on the seismic section obtained by horizontally stacking the seismic data;

The training sample and the learning target are input into the learning vector machine for training to obtain a deep learning network, the training sample is post-stack migration seismic data body, square root amplitude data and seismic sweet spot attribute, and the learning target is the learning point;

Based on the post-stack migration seismic data volume, square root amplitude data, and seismic sweet spot attributes, the fracture-vug characteristic attribute data volume is obtained by using the deep learning network.

Further, the carbonate fracture-cavity seismic characterization method includes:

Based on the characteristic attribute data volume of fractures and caves, the spatial distribution characteristics of carbonate fractures and caves are displayed in the display mode of section, plane or three-dimensional space.

Further, the extraction method of the root mean square amplitude data volume is:

Based on the post-stack migration seismic data volume, the root-mean-square amplitude data volume of post-stack migration seismic numbers is extracted by using the attribute extraction tool in the paradigm software.

Further, the extraction method of the seismic sweet spot attribute data body is:

Based on the post-stack migration seismic data volume, the sweet spot attribute data volume of the post-stack migration seismic data is extracted by using the attribute extraction tool in the paradigm software.

Further, the specified method of the learning point is:

Use the mouse to pick up learning points on the seismic section.

According to the second aspect of the present invention, the carbonate fracture-cavity seismic characterization device based on deep learning includes:

A data acquisition module for acquiring seismic data;

A data stacking module, configured to horizontally stack the seismic data to obtain a post-stack migration seismic data volume;

A data extraction module, configured to extract the RMS amplitude data volume and the seismic sweet spot attribute data volume based on the post-stack migration seismic data volume;

Learning point picking module, used to pick up learning points on the seismic profile;

The learning network construction module is used to input training samples and learning targets into the learning vector machine to train to obtain a deep learning network, the training samples are post-stack migration seismic data volume, square root amplitude data and seismic sweet spot attributes, and the learning targets for said learning point;

The fracture-vug characterization module is configured to use the deep learning network to obtain a fracture-vug characteristic attribute data volume based on the post-stack migration seismic data volume, square root amplitude data, and seismic sweet spot attributes.

Further, the carbonate fracture-cavity seismic characterization device also includes:

The display module is used to display the spatial distribution characteristics of carbonatite fractures and caves in a section, plane or three-dimensional display manner based on the fracture-cavity characteristic attribute data volume.

Further, the data extraction module is specifically configured to use the attribute extraction tool in the paradigm software to extract the RMS amplitude data volume and sweet spot attribute data volume of post-stack migration seismic data based on the post-stack migration seismic data volume.

Further, the learning point picking module is specifically configured to pick up learning points on the seismic section by means of mouse picking.

The beneficial effects of the present invention are: the present invention organically combines the post-stack migration seismic data volume, the root mean square amplitude data volume and the sweet spot attribute data volume to establish a deep learning network, and then uses the method of machine learning to visualize carbonate rock fractures and caves The spatial distribution characteristics and planar distribution rules of carbonatite fractures and caves are effectively predicted. Compared with the published prediction technology, the method of the present invention can not only predict fractured fractures and caves well, but also has a good prediction effect on underground dissolution pores and caves. The distribution of carbonate rock dissolution-type pores predicted by this method is consistent with Inoue has a high degree of agreement.

Description of drawings

Fig. 1 is the flow chart of an embodiment of the carbonate fracture-cavity seismic description method in the present invention;

Figure 2 is a schematic diagram of a post-stack migration seismic data volume in one embodiment;

Figure 3 is a schematic diagram of an RMS amplitude data volume in one embodiment;

Fig. 4 is a schematic diagram of a dessert attribute data body in an embodiment;

Fig. 5 is a schematic diagram of a learning point in an embodiment;

Fig. 6 is the profile feature of the carbonate fracture-vug obtained by deep learning in one embodiment;

Fig. 7 is the plane feature of carbonate fracture-cavity obtained by deep learning in one embodiment;

Fig. 8 is the three-dimensional spatial characteristics of carbonate fractures and caves obtained by deep learning in one embodiment;

Fig. 9 is a composition block diagram of an embodiment of the carbonate fracture-cavity seismic characterization device in the present invention.

detailed description

The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Apparently, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts belong to the protection scope of the present invention.

Referring to Fig. 1-Fig. 9, the present invention provides a method and device for seismic characterization of carbonate fracture-caves based on deep learning:

The first aspect of the present invention provides a method for seismic characterization of carbonate fracture-caves based on deep learning. As shown in FIG. 1 , the method for seismic characterization of carbonate fracture-caves includes steps S100 to S600, which will be described in detail below.

S100. Acquiring seismic data.

S200. Perform horizontal stacking on the seismic data to obtain a post-stack migration seismic data volume.

As shown in Fig. 2, according to the post-stack migration seismic data volume, it can be known that the carbonate fracture-vuggy system generally presents a “string of beads” feature on the seismic section. Seismic amplitude enhancement, chaotic reflections, strong reflections, and beaded reflections in fracture-vug development areas on the seismic section.

Generally, the output result after horizontal stacking of seismic data is time section. In some embodiments, on the basis of the migration principle of the finite difference method and the Kirchhoff integral method, post-stack migration processing and analysis are performed on the geological model data with the help of Vista software; the analysis results show that the geological conditions are relatively simple and For seismic records with low signal-to-noise ratio, the effect of post-stack migration processing is better than that of pre-stack migration processing, so post-stack migration still has its advantages within a certain range.

S300. Based on the post-stack migration seismic data volume, extract a root mean square amplitude data volume and a seismic sweet spot attribute data volume.

In some embodiments, based on the post-stack migration seismic data volume, an RMS amplitude data volume of post-stack migration seismic numbers is extracted using an attribute extraction tool in the paradigm software.

In some embodiments, based on the post-stack migration seismic data volume, a sweet spot attribute data volume of post-stack migration seismic data is extracted using an attribute extraction tool in the paradigm software.

That is, in some embodiments, the RMS amplitude data volume and the seismic sweet spot attribute data volume are extracted from the post-stack migration seismic data volume by using the attribute extraction tool in the paradigm software. In this embodiment, by extracting the root mean square amplitude data volume and the seismic sweet spot attribute data volume from the post-stack migration seismic data volume, the number of learning samples is increased, which is beneficial to improve the accuracy of the deep learning network.

S400. Pick learning points on the seismic section obtained by horizontally stacking the seismic data.

In some embodiments, the learning point is picked on the seismic section by mouse picking, and the learning point has the characteristics of enhanced amplitude, strong reflection, beaded reflection, and low velocity on the seismic.

S500. Input the training sample and learning target into the learning vector machine for training to obtain a deep learning network, the training sample is post-stack migration seismic data volume, square root amplitude data and seismic sweet spot attribute, and the learning target is the learning point .

Specifically, the post-stack migration seismic data volume, square root amplitude data, and seismic sweet spot attributes are used as training samples, and the learning point is the learning target. The training samples and learning targets are input into the learning vector machine, and the depth is determined through iterative updating. Learning network; if the correlation is poor after iteration, return to iterate again until the preset effect is achieved.

S600. Based on the post-stack migration seismic data volume, square root amplitude data, and seismic sweet spot attributes, use the deep learning network to obtain a fracture-vug feature attribute data volume.

Specifically, the post-stack migration seismic data volume, square root amplitude data, and seismic sweet spot attributes are input into the deep learning network, and under the constraints of the training model, the fracture-vug feature attribute data volume is obtained through iterative updating.

In some embodiments, based on the characteristic attribute data volume of fractures and caves, the spatial distribution characteristics of carbonate fractures and caves are displayed in a section, plane or three-dimensional display.

Hereinafter, a case is used to illustrate the implementation method. Obtain seismic data; the post-stack migration seismic data volume obtained by horizontal stacking of the acquired seismic data is shown in Figure 2; the root mean square amplitude data volume extracted from the post-stack migration seismic data volume is shown in Figure 3 Figure 4 shows the seismic sweet spot attribute data volume extracted from the post-stack migration seismic data volume; Figure 5 shows the learning points picked up on the seismic section; input the training samples and learning targets into the learning vector machine for The deep learning network is obtained through training, the training samples are post-stack migration seismic data volume, square root amplitude data and seismic sweet spot attribute, and the learning target is the learning point; based on the post-stack migration seismic data volume, square root Based on the amplitude data and the seismic sweet spot attributes, the deep learning network is used to obtain the fracture-cavity characteristic attribute data body; based on the fracture-cavity characteristic attribute data body, the spatial distribution characteristics of carbonate fracture-caves are displayed in a section display mode, as shown in Figure 6 As shown; based on the fracture-cavity characteristic attribute data body, the spatial distribution characteristics of carbonatite fracture-caves are displayed in a planar display mode, as shown in Figure 7; based on the fracture-cavity characteristic attribute data body, the display mode in three-dimensional space The spatial distribution characteristics of carbonatite fractures and caves are displayed, as shown in Fig. 8.

The second aspect of the present invention provides a carbonatite fracture-cavity seismic characterization device based on deep learning. As shown in FIG. Learning point picking module, learning network building module and seam characterization module.

The data acquisition module is used for acquiring seismic data. In this embodiment, the data acquisition module can be used to execute step S100 shown in FIG. 1 , and for a specific description of the data acquisition module, refer to the description of step S100 .

The data stacking module is used to horizontally stack the seismic data to obtain a post-stack migration seismic data volume. In this embodiment, the data superposition module can be used to execute step S200 shown in FIG. 1 , and for a specific description of the data superposition module, refer to the description of step S200 .

The data extraction module is used to extract the root mean square amplitude data volume and the seismic sweet spot attribute data volume based on the post-stack migration seismic data volume. In this embodiment, the data extraction module can be used to execute step S300 shown in FIG. 1 , and for a specific description of the data extraction module, refer to the description of step S300 .

The learning point picking module is used to pick learning points on the seismic profile. In this embodiment, the learning point picking module can be used to execute step S400 shown in FIG. 1 , and for a specific description of the learning point picking module, refer to the description of step S400 .

The learning network construction module is used to input training samples and learning targets into the learning vector machine to train to obtain a deep learning network, the training samples are post-stack migration seismic data volume, square root amplitude data and seismic sweet spot attributes, and the learning targets For the learning point. In this embodiment, the learning network construction module can be used to execute step S500 shown in FIG. 1 , and for a specific description of the learning network construction module, refer to the description of step S500 .

The fracture-vug characterization module is configured to use the deep learning network to obtain a fracture-vug characteristic attribute data volume based on the post-stack migration seismic data volume, square root amplitude data, and seismic sweet spot attributes. In this embodiment, the crack and hole drawing module can be used to execute step S600 shown in FIG. 1 , and for a specific description of the crack and hole drawing module, refer to the description of step S600 .

The above descriptions are only preferred embodiments of the present invention, and it should be understood that the present invention is not limited to the forms disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and Modifications can be made within the scope of the ideas described herein, by virtue of the above teachings or skill or knowledge in the relevant art. However, changes and changes made by those skilled in the art do not depart from the spirit and scope of the present invention, and should all be within the protection scope of the appended claims of the present invention.

Claims (9)

1. The carbonate fracture-cavity seismic characterization method based on deep learning, characterized in that it includes: Obtain seismic data; performing horizontal stacking on the seismic data to obtain a post-stack migration seismic data volume; Extracting a root mean square amplitude data volume and a seismic sweet spot attribute data volume based on the post-stack migration seismic data volume; Picking learning points on the seismic section obtained by horizontally stacking the seismic data; The training sample and the learning target are input into the learning vector machine for training to obtain a deep learning network, the training sample is post-stack migration seismic data body, square root amplitude data and seismic sweet spot attribute, and the learning target is the learning point; Based on the post-stack migration seismic data volume, square root amplitude data, and seismic sweet spot attributes, the fracture-vug characteristic attribute data volume is obtained by using the deep learning network. 2. the carbonate fracture-cavity seismic characterization method based on deep learning according to claim 1, is characterized in that, the carbonate fracture-cavity seismic characterization method comprises: Based on the characteristic attribute data volume of fractures and caves, the spatial distribution characteristics of carbonate fractures and caves are displayed in the display mode of section, plane or three-dimensional space. 3. the carbonate fracture-cavity seismic description method based on deep learning according to claim 1, is characterized in that, the extracting method of described root mean square amplitude data body is: Based on the post-stack migration seismic data volume, the root-mean-square amplitude data volume of post-stack migration seismic numbers is extracted by using the attribute extraction tool in the paradigm software. 4. the carbonate fracture-cavity seismic description method based on deep learning according to claim 1, is characterized in that, the extraction method of described seismic sweet spot attribute data body is: Based on the post-stack migration seismic data volume, the sweet spot attribute data volume of the post-stack migration seismic data is extracted by using the attribute extraction tool in the paradigm software. 5. the carbonate fracture-cavity seismic description method based on deep learning according to claim 1, is characterized in that, the appointed method of described learning point is: Use the mouse to pick up learning points on the seismic section. 6. The carbonate fracture-cavity seismic characterization device based on deep learning, characterized in that it includes: A data acquisition module for acquiring seismic data; A data stacking module, configured to horizontally stack the seismic data to obtain a post-stack migration seismic data volume; A data extraction module, configured to extract the RMS amplitude data volume and the seismic sweet spot attribute data volume based on the post-stack migration seismic data volume; Learning point picking module, used to pick up learning points on the seismic profile; The learning network construction module is used to input training samples and learning targets into the learning vector machine to train to obtain a deep learning network. The training samples are post-stack migration seismic data bodies, square root amplitude data and seismic sweet spot attributes, and the learning targets are for said learning point; The fracture-vug characterization module is configured to use the deep learning network to obtain a fracture-vug feature attribute data volume based on the post-stack migration seismic data volume, square root amplitude data, and seismic sweet spot attributes. 7. The carbonate fracture-cavity seismic characterization device based on deep learning according to claim 6, wherein the carbonate fracture-cavity seismic characterization device also includes: The display module is used to display the spatial distribution characteristics of carbonatite fractures and caves in a section, plane or three-dimensional display manner based on the fracture-cavity characteristic attribute data volume. 8. The carbonate fracture-cavity seismic characterization device based on deep learning according to claim 6, wherein the data extraction module is specifically used to utilize the attributes in the paradigm software based on the post-stack migration seismic data body The extraction tool extracts the RMS amplitude data volume and sweet spot attribute data volume of post-stack migration seismic numbers. 9. The deep learning-based seismic characterization device for carbonate rock fractures and caves according to claim 6, wherein the learning point picking module is specifically used to pick up learning points on the seismic section by means of mouse picking.

Images