Neural network-based volcano channel identification method and device, medium and terminal equipment
By using a volcanic conduit-based method for volcanic conduit identification, a model trained with post-stack 3D seismic data is automatically identified. This method solves the problems of low identification efficiency and poor accuracy in existing technologies, and achieves efficient and accurate volcanic conduit identification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- PETROCHINA CO LTD
- Filing Date
- 2024-11-21
- Publication Date
- 2026-05-22
AI Technical Summary
Identifying volcanic passages is challenging, and existing technologies are inefficient, susceptible to human error, and have poor accuracy.
A convolutional neural network-based approach was adopted, using post-stack 3D seismic data to establish a sample database. The convolutional neural network model was trained using grayscale seismic profiles to automatically identify volcanic conduits.
It improves the accuracy and efficiency of volcanic conduit identification, reduces the impact of human factors, and lowers labor costs.
Smart Images
Figure CN122072998A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a volcanic passage identification technology, specifically a volcanic passage identification method, device, medium, and terminal equipment based on neural networks. Background Technology
[0002] Volcanic conduits are channels through which magma flows from the magma chamber through underground rock strata and out to the surface via the crater or vent. They are a key component of volcanic structures, and identifying volcanic conduits is fundamental to studying favorable facies zones and volcanic reservoirs. However, due to the great burial depth, complex stratigraphic contact relationships, and the influence of seismic data quality, volcanic conduit identification is quite challenging. With the continuous advancement of oil exploration and development, especially in medium-deep oilfields with igneous rock development, the accurate identification of volcanic conduits directly affects the deployment of subsequent development wells.
[0003] Currently, many geologists have studied the use of seismic data for volcanic conduit identification, typically relying on seismic properties as the primary method. However, because volcanic conduits are located deep underground and approximately perpendicular to the horizontal plane, conventional seismic geometric properties such as coherence and root-mean-square properties only reflect information within a small area, often containing numerous interference terms and exhibiting strong ambiguity, severely impacting the accuracy of volcanic conduit identification. Furthermore, in practice, volcanic conduits are usually identified through manual observation of seismic profiles. This involves setting certain trace spacing within a 3D seismic dataset and observing each profile to locate the distribution of volcanic conduits. This method is labor-intensive, inefficient, and susceptible to human error, further affecting the accuracy of volcanic conduit identification. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this invention aims to provide a volcano channel identification method, apparatus, medium, and terminal equipment based on neural networks, in order to reduce labor, improve work efficiency, and enhance identification accuracy.
[0005] To achieve the above objectives, the technical solution adopted by this invention is as follows: A volcano channel identification method based on neural networks, comprising the following steps: Step 1: Obtain multiple sets of seismic profiles based on post-stack 3D seismic data; Step 2: Perform grayscale processing on multiple sets of seismic profile images and use the grayscale processed sets of seismic profile images as a sample database for building a convolutional neural network model; Step 3: Build a convolutional neural network model using a sample database and evaluate it through testing to ensure the accuracy of the model. Step 4: Input the seismic profile of the volcanic conduit to be identified into the established convolutional neural network model, identify the volcanic conduit in the seismic profile, and output the volcanic conduit map.
[0006] As a limitation of the present invention, each set of seismic profiles is a seismic profile with 128×128 pixels.
[0007] As a further limitation of the present invention, each set of seismic profiles includes training data and label data, wherein the training data is an initial seismic profile obtained directly from post-stack 3D seismic data, and the label data is a volcanic conduit map in which geologists identify and delineate the volcanic conduit morphology on the initial seismic profile; the grayscale-processed training data is used as the input to build a convolutional neural network model, and the grayscale-processed label data is used as the output to build a convolutional neural network model.
[0008] The present invention also provides a volcanic channel identification device based on a neural network, which can realize the volcanic channel identification method based on a neural network described in any one of the above claims; the volcanic channel identification device includes: an acquisition module, which acquires a sample database for establishing a convolutional neural network based on post-stack 3D seismic data; Establish a recognition model module and build a convolutional neural network model that can accurately identify volcanic conduits based on the sample database; The transmission module is used to transmit the seismic profile of the volcanic conduit to be identified to the convolutional neural network model and output the volcanic conduit map that identifies and delineates the morphology of the volcanic conduit.
[0009] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the neural network-based volcano channel identification method described in any one of the preceding claims.
[0010] The present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, can implement the neural network-based volcano channel identification method described above.
[0011] By adopting the above-described technical solution, the beneficial effects achieved by this invention compared to the prior art are as follows: This invention identifies volcanic conduits by combining stacked 3D seismic profiles with convolutional neural networks (CNNs). The CNN method is relatively less dependent on the quality of seismic data, and can still accurately identify volcanic conduits even when the surrounding seismic data quality is poor. The CNN method can capture the complex reflection characteristics of the seismic data within the volcanic conduit, effectively improving the accuracy of this invention. Furthermore, this invention offers a high degree of automation in identifying volcanic conduits. A CNN model specifically designed for volcanic conduit identification simulates the thought process of geologists, exhibiting strong specificity. In actual volcanic conduit identification, only the seismic profile of the conduit to be identified needs to be input into the CNN model to identify the volcanic conduit and output a volcanic conduit image. This effectively reduces the influence of human factors, ensuring accuracy while also reducing labor costs and improving work efficiency. Attached Figure Description
[0012] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.
[0013] Figure 1 This is a flowchart illustrating the volcanic passage identification method in Embodiment 1 of the present invention; Figure 2 This is an example diagram of the convolutional neural network model identifying volcanic channels in Embodiment 1 of the present invention; Figure 3 for Figure 2 The seismic profile image entered in the input; Figure 4 for Figure 2 The volcano passage map output from the source; Figure 5 This is a block diagram of a volcanic channel identification device based on a neural network according to Embodiment 2 of the present invention; In the diagram: 101, Acquisition module; 102, Recognition model establishment module; 103, Transmission module. Detailed Implementation
[0014] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.
[0015] Example 1: A method for identifying volcanic passages based on neural networks like Figure 1 As shown, this embodiment 1 includes the following steps: Step 1: Obtain multiple sets of seismic profiles based on post-stack 3D seismic data. Specifically, in this embodiment 1, a total of 200 sets of seismic profiles with a pixel size of 128×128 were obtained. Each set of seismic profiles includes training data and label data. The training data consists of the initial seismic profiles directly obtained from the post-stack 3D seismic data, while the label data consists of volcanic conduit maps that geologists identified and delineated on the initial seismic profiles to depict the morphology of volcanic conduits.
[0016] Step 2: The 200 sets of 128×128 pixel seismic profile images from Step 1 are processed into grayscale, and these 200 sets of grayscale-processed seismic profile images are used as the sample database for building the convolutional neural network model. The grayscale-processed training data is used as the input to the convolutional neural network model, and the grayscale-processed label data is used as the output. The convolutional neural network model referred to in Example 1 is the U-Net neural network model.
[0017] Step 3: Build a convolutional neural network model using a sample database and evaluate it through testing to ensure the accuracy of the model. When building the convolutional neural network model, the grayscale-processed training data is used as the input, and the grayscale-processed label data is used as the output.
[0018] Specifically, in this embodiment 1, the 128×128 pixel training data in the sample database is used as the input of the convolutional neural network, and pixel calculation is performed according to the convolutional neural network. First, a layer of 0 values is added to the outermost layer of the 128×128 pixel training data in the sample database to make it a 130×130 grayscale matrix. The first layer uses 64 3×3 convolutional kernels to obtain a 64×128×128 feature matrix. Then, a layer of 0 values is added to the outermost layer to make it a 64×130×130 matrix. Next, a second layer of 64 3×3 convolutional kernels is used to obtain a 64×128×128 feature matrix. Finally, a third layer of 2×2 maxpool downsampling is used to convert it into a 64×64×64 feature matrix.
[0019] The 64×64×64 feature matrix is transformed into a 64×66×66 matrix by adding a layer of 0 values to the outermost layer. The fourth layer has 128 3×3 convolutional kernels to obtain a 128×64×64 feature matrix. The outermost layer is then transformed into a 128×66×66 matrix by adding another layer of 0 values. The fifth layer has 128 3×3 convolutional kernels to obtain a 128×64×64 feature matrix. Finally, the sixth layer has 2×2 maxpool downsampling to transform it into a 128×32×32 feature matrix.
[0020] The 128×32×32 feature matrix is transformed into a 128×34×34 matrix by adding a layer of 0 values to the outermost layer. The seventh layer has 256 3×3 convolutional kernels to obtain a 256×32×32 feature matrix, which is then transformed into a 256×34×34 matrix by adding another layer of 0 values to the outermost layer. The eighth layer has 256 3×3 convolutional kernels to obtain a 256×32×32 feature matrix. Finally, the ninth layer has 2×2 maxpool downsampling to transform it into a 256×16×16 feature matrix.
[0021] The 256×16×16 feature matrix is transformed into a 256×18×18 matrix by adding a layer of 0 values to the outermost layer. The tenth layer uses 512 3×3 convolution kernels to obtain a 512×16×16 feature matrix. The outermost layer is then added with another layer of 0 values to transform it into a 512×18×18 matrix. The eleventh layer uses 512 3×3 convolution kernels to obtain a 512×16×16 feature matrix. Finally, the twelfth layer uses 2×2 deconvolution upsampling to transform it into a 512×32×32 feature matrix.
[0022] The 512×32×32 feature matrix and the 256×32×32 feature matrix of the eighth layer are concatenated to form a 768×32×32 fused feature matrix. An outermost layer with zero values is added to form a 768×34×34 matrix. The thirteenth layer has 256 3×3 convolution kernels to obtain a 256×32×32 feature matrix. Another outermost layer with zero values is added to form a 256×34×34 matrix. Then, the fourteenth layer has 256 3×3 convolution kernels to obtain a 256×32×32 feature matrix. Finally, the fifteenth layer has 2×2 deconvolution upsampling to transform it into a 256×64×64 feature matrix.
[0023] The 256×64×64 feature matrix and the 128×64×64 feature matrix of the fifth layer are concatenated to form a 384×64×64 fused feature matrix. An outermost layer with zero values is added to form a 384×66×66 matrix. The sixteenth layer has 128 3×3 convolution kernels to obtain a 128×64×64 feature matrix. Another outermost layer with zero values is added to form a 128×66×66 matrix. Then, the seventeenth layer has 128 3×3 convolution kernels to obtain a 128×64×64 feature matrix. Finally, the eighteenth layer has 2×2 deconvolution upsampling to transform it into a 128×128×128 feature matrix.
[0024] The 128×128×128 feature matrix is concatenated with the 64×128×128 feature matrix of the second layer to form a 192×128×128 fused feature matrix. An outermost layer with zero values is added to form a 192×130×130 matrix. The nineteenth layer has 64 3×3 convolutional kernels to obtain a 64×128×128 feature matrix. Another outermost layer with zero values is added to form a 64×130×130 matrix. Then, the twentieth layer has 64 3×3 convolutional kernels to obtain a 64×128×128 feature matrix. Finally, the twenty-first fully connected layer is used to transform it into a 1×128×128 volcano channel prediction matrix.
[0025] The final output, a 1×128×128 image, contains the seismic profile label data of the volcanic conduit established by geologists. After building the initial convolutional neural network model, the training set data is input and the neural network weight parameters are fitted to establish the convolutional neural network model. The accuracy of the established convolutional neural network model is verified by testing the new grayscale seismic profile for identifying volcanic conduits. If the model passes the test, it is used as the final convolutional neural network model.
[0026] Step 4, as follows Figures 2 to 4 As shown, the seismic profile of the volcanic conduit to be identified is input into the established convolutional neural network model, which identifies the volcanic conduit in the seismic profile and outputs the volcanic conduit map.
[0027] Example 2: A volcanic passage identification device based on neural networks like Figure 5 As shown, this embodiment 2 includes an acquisition module 101, a model building module 102, and a transmission module 103. The acquisition module 101 acquires a sample database for building a convolutional neural network based on post-stack 3D seismic data. The model building module 102 builds a convolutional neural network model that can accurately identify volcanic conduits based on the sample database. The transmission module 103 transmits the seismic profile of the volcanic conduit to be identified to the convolutional neural network model and outputs a map of the identified and delineated volcanic conduit. It should be noted that the acquisition module 101, the model building module 102, and the transmission module 103 all implement their respective functions using the method described in embodiment 1.
[0028] Example 3: A computer-readable storage medium In this embodiment 3, the computer-readable storage medium stores a computer program that, when executed by a processor, can implement the aforementioned volcano channel recognition method based on a neural network. Specifically, the computer-readable storage medium stores non-transitory computer-readable instructions. When these non-transitory computer-readable instructions are executed by the processor, all or part of the steps of the method in the aforementioned embodiment 1 of this disclosure are performed. Computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or portable hard drive), media with built-in rewritable non-volatile memory (e.g., memory card), and media with built-in ROM (e.g., ROM cartridge).
[0029] Example 4: A terminal device This embodiment 4 includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the method in embodiment 1.
[0030] It should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still modify the technical solutions described in the above embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for identifying volcanic passages based on neural networks, characterized in that, Includes the following steps: Step 1: Obtain multiple sets of seismic profiles based on post-stack 3D seismic data; Step 2: Perform grayscale processing on multiple sets of seismic profile images and use the grayscale processed sets of seismic profile images as a sample database for building a convolutional neural network model; Step 3: Build a convolutional neural network model using a sample database and evaluate it through testing to ensure the accuracy of the model. Step 4: Input the seismic profile of the volcanic conduit to be identified into the established convolutional neural network model, identify the volcanic conduit in the seismic profile, and output the volcanic conduit map.
2. The volcanic conduit identification method based on neural networks according to claim 1, characterized in that, Each set of seismic profile images is a seismic profile image with 128×128 pixels.
3. The volcanic channel identification method based on neural networks according to claim 2, characterized in that, Each set of seismic profiles includes training data and label data. The training data is the initial seismic profile obtained directly from post-stack 3D seismic data, and the label data is a volcanic conduit map in which geologists identify and delineate the volcanic conduit morphology on the initial seismic profile. The grayscale-processed training data is used as the input to build a convolutional neural network model, and the grayscale-processed label data is used as the output to build a convolutional neural network model.
4. A volcanic passage identification device based on a neural network, characterized in that, The volcanic passage identification device can implement the neural network-based volcanic passage identification method described in any one of claims 1-3; the volcanic passage identification device includes: The acquisition module obtains a sample database for building a convolutional neural network based on post-stack 3D seismic data; Establish a recognition model module and build a convolutional neural network model that can accurately identify volcanic conduits based on the sample database; The transmission module is used to transmit the seismic profile of the volcanic conduit to be identified to the convolutional neural network model and output the volcanic conduit map that identifies and delineates the morphology of the volcanic conduit.
5. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of a neural network-based volcano channel identification method as described in any one of claims 1-3.
6. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a volcanic channel identification method based on a neural network as described in any one of claims 1-3.