Mud shale fracture earthquake identification method based on 3D U-Net convolutional neural network combined with ant tracking

Abstract

The invention relates to a shale fracture earthquake identification method based on a 3D U-Net convolutional neural network combined with ant tracking. According to the technical scheme, a geological model of shale layer cracks is established based on single well data, and forward modeling is conducted on the geological model to determine the dominant frequency band of shale crack recognition; the method comprises the following steps: obtaining a dominant frequency band data body by using frequency spectrum peak value frequency division, carrying out 3D U-Net convolutional neural network combined ant tracking calculation on the dominant frequency band data body to obtain a 3D U-Net ant body, carrying out verification by using microseismic data, extracting attributes of the 3D U-Net ant body along a layer, and determining regional plane distribution characteristics of shale cracks. The method has the advantages that the fracture is predicted by combining a single well and seismic data, and the exploration cost reduction effect is remarkable; seismic forward modeling, spectrum peak decomposition and dominant frequency band data calculation are comprehensively combined, so that the seismic resolution of the shale cracks is remarkably improved; and horizon slices are extracted by using the data volume, so that the crack identification range is expanded.

Description

Technical Field

[0001] The present invention relates to the field of seismic attribute recognition of shale fractures, and in particular to a seismic recognition method for shale fractures based on a 3D U-Net convolutional neural network combined with ant tracking. Background Art

[0002] At present, the identification of shale fractures is still a key problem in shale risk exploration, which seriously restricts the exploration and development of shale gas resources. In addition to core CT scanning, the prediction accuracy of other methods for identifying shale fractures, such as single attribute prediction, anisotropy, fracture parameter inversion and other technologies, is not ideal. CT scanning technology relies on drilling cores and can only identify fractures in a small range of drilling cores. For shale fractures in the entire area, its results are tantamount to seeing the leopard through the tube. In recent years, geologists have successfully realized the identification of small faults using convolutional neural network image processing technology in the biomedical field, but no effective research on the identification of shale fractures has been conducted. Ant tracking is extremely sensitive to the difference in fractures in the data body, and is usually combined with coherence bodies, curvature and other attributes to identify fractures. It has achieved accurate identification of carbonate faults with high fracture reflection characteristics, but in the identification of shale fractures with unclear fracture reflection characteristics, its results have too much clutter and interference, and the effect of fracture identification needs to be improved. Therefore, developing a new method that can efficiently and accurately identify shale fractures is of great significance for promoting the deep exploration and development of shale gas resources.

[0004] Since the Longmaxi Formation is a shale layer with a burial depth of more than 3500m, the resolution of seismic data is low, the seismic response of fractures is weak, and conventional methods cannot identify small faults and fractures; in addition, the fracture prediction technology of the early post-stack and pre-stack earthquakes in the Zi-215 well area is single, the fracture identification accuracy is limited, and the verification standard has not yet been established, which makes it difficult to support the optimization design of fracturing, hindering the subsequent production and construction of the Zi-215 well area. To solve this problem, the 3DU-Net convolutional neural network combined with ant tracking technology is used to improve the accuracy of shale fracture prediction, in order to find out the distribution law of shale fractures and provide strong guarantee for safe production and efficient gas production in the Zi-215 well area. Summary of the invention

[0005] The purpose of the present invention is to provide a shale fracture seismic identification method based on 3D U-Net convolutional neural network combined with ant tracking in view of the above-mentioned defects in the prior art. Considering the influence of seismic data resolution on fracture identification accuracy, the fracture seismic forward modeling is first used to clarify the dominant spectrum peak frequency band for fracture identification, and then the seismic data is spectrum decomposed based on the dominant frequency band of the forward modeling to obtain seismic data with the optimal resolution for shale fracture identification. Finally, the 3DU-Net convolutional neural network combined with ant tracking technology combined with microseismic monitoring verification is used to establish a shale fracture identification method based on 3D U-Net convolutional neural network combined with ant tracking. This shale fracture identification method has the advantages of low exploration cost, high identification accuracy, and wide identification range. It can effectively guide the well site deployment of shale layers, which is directly related to the problem of reducing shale gas extraction costs and increasing production capacity.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is:

[0007] The seismic identification method of shale fractures based on 3D U-Net convolutional neural network combined with ant tracking includes the following steps:

[0008] (1) Preparation of main data for fracture identification: Collect well logging data (including lithology, acoustic wave time difference and density) and 3D seismic data of a single well, and understand the peak characteristics of the spectral spectrum of 3D seismic data and the effective frequency band range;

[0009] (2) Using logging data to establish a shale fracture forward model, and based on actual seismic parameters, perform seismic forward modeling of different peak frequency bands to identify the dominant frequency band for shale fracture identification;

[0010] (3) Using spectrum decomposition technology to perform frequency division processing on seismic data, so that its frequency band is consistent with the dominant frequency band for identifying shale fractures;

[0011] (4) Perform 3D U-Net convolutional neural network calculation on the seismic data volume with dominant frequency band for shale fracture identification to obtain a 3D U-Net data volume;

[0012] (5) Perform ant tracking calculation on the 3D U-Net volume to obtain the final shale fracture identification data volume, and use microseismic to verify the accuracy of shale fracture prediction of the 3D U-Net ant data volume;

[0013] (6) By extracting the shale fracture identification data along the layer, the regional shale fracture prediction results can be obtained.

[0014] Compared with the prior art, the present invention has the following beneficial effects:

[0015] (1) Shale fractures can be predicted by combining single well data with seismic data, which significantly reduces exploration costs;

[0016] (2) Using the trinity method of seismic forward modeling, spectrum peak decomposition, and dominant frequency band data calculation to maximize the seismic resolution of shale fractures;

[0017] (3) The 3D U-Net convolutional neural network combined with ant tracking, combined with the verification of microseismic data profiles, has greatly improved the accuracy of fracture identification;

[0018] (4) By using the data volume to extract slices along the layer, the crack identification range is wider. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Attached Figure 1 It is a spectrum peak and effective frequency band analysis diagram of three-dimensional data from well area 215 in the example of the present invention;

[0023] Attached Figure 2 It is a fracture geological model diagram established based on the well data from Well 215 in the example of the present invention;

[0024] Attached Figure 3 It is the seismic forward modeling result of different peak frequency band signals of the shale fracture geological model based on the actual seismic main frequency wavelet in the example of the present invention;

[0025] Attached Figure 4 It is a seismic data profile of the dominant frequency band for identifying shale fractures in the 215 well area in the example of the present invention;

[0026] Attached Figure 5 It is a superposition diagram of the 3D U-Net data volume and the seismic volume from the 215 well area in the example of the present invention;

[0027] Attached Figure 6 It is a 3D U-Net ant volume section and microseismic superposition diagram from the 215 well area through the H59 and H3 wells in the example of the present invention;

[0028] Attached Figure 7 It is the variance volume, 3D U-Net ant volume and 3DU-Net ant volume and microseismic superposition map of the layer near the H1 well of the Longmaxi Formation in the 215 well area in the example of the present invention. DETAILED DESCRIPTION

[0029] In order to make the technical means and objectives adopted by the present invention easy to understand, the present invention is further described below in conjunction with specific implementation methods.

[0030] Unless otherwise specified, the methods used in the following examples are all conventional methods.

[0031] Combined with Figure 1-5 , the present invention is further described:

[0032] Step 1: Preparation of main data for earthquake prediction of shale fractures;

[0033] The present invention takes the Zi 215 well area in Zigong area in southern Sichuan as an example;

[0034] Collect 3D seismic data from Well 215, as well as lithology, acoustic time difference, density logging data, and microseismic and other dynamic and static data; understand the spectral characteristics of the block seismic volume, such as Figure 1 As shown, it is clear that the earthquake amplitude energy is greater than 0.3, that is, 10-50Hz is the effective frequency band of the seismic data. In the effective frequency band, the spectrum shows multiple peak characteristics, and 17Hz, 25Hz, 32Hz, and 43Hz are its spectrum peaks;

[0035] Step 2: If Figure 2 As shown in FIG. 1 , based on the logging data of lithology, acoustic time difference and density of Well 215, a fracture geological model of the shale layer in the Well 215 area is established. The fracture is regarded as a small fault throw fault, and a geological model with fault throws of 0.2m, 0.4m, 0.6m, 0.8m, 1.0m and 2.0m is established; Figure 3 As shown in the figure, the seismic forward modeling of different peak frequency band signals within the effective frequency band of the seismic data was carried out, and it was found that the fractures with a fault distance less than 2m could not be identified at all from the characteristics of the seismic reflection axis. However, when the main frequency was 25Hz, the bright and dark changes in the amplitude had a good correspondence with the location of the fracture development. Therefore, the main frequency of 25Hz is the dominant frequency band for identifying shale fractures in the 215 well area.

[0036] Step 3: If Figure 4 As shown in the figure, based on the seismic forward modeling results and the spectrum characteristics of the three-dimensional seismic data, the seismic data is spectrum decomposed to obtain the dominant frequency band data volume for shale fracture identification with a center frequency of 25 Hz;

[0037] Step 4: Figure 5 As shown, the 3D U-Net convolutional neural network is used to calculate the dominant frequency band data volume for shale fracture identification to obtain the 3DU-Net data volume;

[0038] Step 5: Figure 6 As shown, the 3DU-Net ant volume is obtained by ant tracking calculation on the 3DU-Net data volume, and the accuracy of the 3DU-Net ant volume is verified by using microseismic data;

[0039] Step 6: Figure 7As shown in the figure, by extracting the 3DU-Net ant body attributes along the layer, the regional fracture prediction results can be obtained. The fracture identification results around the H1 level in the 215 well area are analyzed: the conventional seismic attributes such as variance body are not well displayed, while the ant algorithm combined with the 3D U-Net convolutional neural network not only identifies the northwest-trending fractures of the H1 platform, but also identifies two northeast-trending large faults and a nearly east-west micro-fault and a series of northwest-trending fractures that run through the entire H1 platform. The microseismic signal is highly consistent with its results, proving that it has high accuracy in predicting shale fractures.

[0040] The beneficial effects of the present invention are:

[0041] (1) Shale fractures can be predicted by combining single well data with seismic data, which significantly reduces exploration costs;

[0042] (2) Using the trinity method of seismic forward modeling, spectrum peak decomposition, and dominant frequency band data calculation to maximize the seismic resolution of shale fractures;

[0043] (3) The 3D U-Net convolutional neural network combined with ant tracking, combined with the verification of microseismic data profiles, has greatly improved the accuracy of fracture identification;

[0044] (4) By using the data volume to extract slices along the layer, the crack identification range is wider.

Claims

1. A shale fracture seismic identification method based on 3D U-Net convolutional neural network combined with ant tracking, characterized by: Considering the influence of seismic data resolution on the accuracy of shale fracture identification, the geological model of underground shale fractures was first established using the lithology, acoustic time difference and density data of a single well. The dominant peak frequency band for shale fracture identification was clarified using seismic forward modeling. Then, the spectrum peak decomposition technology was used to obtain the seismic data with the optimal resolution for shale fracture identification. Finally, a shale fracture seismic identification method based on 3D U-Net convolutional neural network combined with ant tracking technology and microseismic monitoring verification was established, including the following steps: (1) Preparation of main data for shale fracture earthquake prediction. Taking the Zi 215 well area in Zigong, southern Sichuan as an example, the present invention collects 3D seismic data from the Zi 215 well as lithology, acoustic time difference and density logging data, and dynamic and static data such as microseismic data; understands the peak characteristics and effective frequency band of the block seismic volume spectrum; (2) establishing a fracture geological model of the shale layer in the 215 well area based on the logging data of lithology, acoustic wave time difference and density of the 215 well, and performing seismic forward modeling of signals of different frequency bands within the effective frequency band of the seismic data in step (1) to clarify the seismic response characteristics of the shale fractures and select the optimal frequency band for seismic identification of shale fractures; (3) Based on the seismic forward modeling shale fracture response characteristic results obtained in step (2) and the frequency spectrum characteristics of the three-dimensional seismic data in step (1), the seismic data is subjected to frequency spectrum peak decomposition to obtain a dominant frequency band data volume for shale fracture identification; (4) performing 3D U-Net convolutional neural network calculation on the dominant frequency band data volume for shale fracture identification obtained in step (3) to obtain a 3DU-Net data volume; (5) Perform ant tracking calculation on the 3DU-Net data volume in step (4) to obtain a 3DU-Net ant volume, and use microseismic data to verify the accuracy of the 3DU-Net ant volume; (6) By extracting the layer attributes of the 3DU-Net ant body in step (5), the regional shale fracture prediction results can be obtained, and intuitive fracture plane distribution results can be obtained.

2. The method for seismic identification of shale fractures based on 3D U-Net convolutional neural network combined with ant tracking according to claim 1, characterized in that: The spectrum peak decomposition refers to performing seismic forward modeling based on the spectrum peak characteristics and effective frequency bands of seismic data, optimizing the peak frequency band of the optimal seismic response of shale fractures, and decomposing the peak frequency band using spectrum decomposition to obtain advantageous data for shale fracture identification.

3. The method for seismic identification of shale fractures based on 3D U-Net convolutional neural network combined with ant tracking according to claim 1, characterized in that: The 3D U-Net convolutional neural network is combined with the ant tracking technology. The 3D U-Net convolutional neural network is used to perform 3D U-Net convolutional neural network calculation on the dominant frequency band data volume for shale fracture identification to obtain a 3D U-Net data volume, and then ant tracking calculation is performed on the data volume to obtain a 3D U-Net ant volume.

4. The method for seismic identification of shale fractures based on 3D U-Net convolutional neural network combined with ant tracking according to claim 1, characterized in that: The microseismic data verification verifies the accuracy of shale fracture prediction by superimposing the microseismic data of the profile and the plane with the 3D U-Net ant body.

Images