Low-speed mudstone intelligent prediction method, system, equipment and medium

Abstract

The invention discloses a low-speed mudstone intelligent prediction method, system, equipment and medium, and the method comprises the steps: carrying out the layer section calibration of a target layer through combining logging data, and dividing the layer section into background mudstone, low-speed mudstone, gas-bearing sandstone and dry layer sandstone; and based on the division result, generating a large number of data sets for neural network training through geologic body evolution. And constructing a convolutional neural network (CNN), and performing normalization processing on the input data to prevent gradient explosion. And training the model to obtain an optimized neural network model. The AVO pre-stack inversion technology is utilized to extract longitudinal and transverse wave velocity and density parameters from pre-stack seismic data to assist in low-speed mudstone recognition. And inputting the processed data into the trained model to realize intelligent prediction of the low-speed mudstone. According to the method, the identification accuracy of the middle-deep reservoir can be effectively improved, misjudgment caused by a'bright spot trap 'of low-speed mudstone is avoided, and reliable technical support is provided for oil-gas exploration.

Description

Technical Field

[0001] The present invention relates to the technical field of seismic data processing and interpretation in oil and gas exploration, and in particular to a low-speed mudstone intelligent prediction method, system, equipment and medium based on AVO pre-stack inversion and geological body evolution. Background Art

[0002] The Yinggehai Basin in the South China Sea is a Cenozoic sedimentary basin. After years of exploration and development, many large and medium-sized gas fields have been discovered, making it an important offshore natural gas production area in my country. However, exploration and development practices have shown that although the bright spot recognition technology has achieved a high success rate in identifying shallow reservoirs in the Yinggehai Basin, it has repeatedly failed in identifying medium and deep reservoirs. After in-depth exploration and analysis, it was found that the low-velocity mudstone widely developed from the Huangliu Formation to the Meishan Formation in the Yinggehai Basin is the main cause of this problem.

[0003] Low-velocity mudstone is a lower-velocity mudstone layer developed under the background mudstone. It appears as a "bright spot" on the seismic profile, which is very similar to the seismic response of gas-bearing sandstone and easily forms a "bright spot trap" of low-velocity mudstone. The existence of this low-velocity mudstone interferes with the accurate identification of gas-bearing reservoirs by traditional bright spot identification technology, making it difficult to distinguish between gas-bearing sandstone and low-velocity mudstone on the seismic profile, thereby increasing the difficulty and uncertainty of exploration.

[0004] In order to overcome the interference of low-velocity mudstone on the identification of gas-bearing sandstone, a low-velocity mudstone intelligent prediction method and system based on AVO pre-stack inversion and geological body evolution is proposed. By combining seismic data analysis, AVO inversion technology and geological body evolution simulation, the development interval of low-velocity mudstone can be effectively identified, thereby improving the identification accuracy of gas-bearing sandstone, providing important technical support for exploration in the Yinggehai Basin and other similar geological conditions. Summary of the invention

[0005] In view of the defects of the prior art, the present invention provides a low-speed mudstone intelligent prediction method, system, equipment and medium.

[0006] In order to achieve the above invention object, the technical solution adopted by the present invention is as follows:

[0007] A low-speed mudstone intelligent prediction method comprises the following steps:

[0008] Step 1: Combine logging data analysis and logging interpretation results to divide the target layer into sections. The division results are used as the output Label of the neural network, and the division types include background mudstone, low-velocity mudstone, gas-bearing sandstone, and dry layer sandstone.

[0009] Step 2: According to the results of the layer division, the geological body is evolved horizontally and vertically to generate a large amount of data for neural network training. During the evolution of the geological body, the logging data is used for wavelet analysis to generate prestack angle gathers, including TWT, prestack angle gathers, Vp, Vs, Den and Label.

[0010] Step 3: Select the ReLU function as the activation function and construct a CNN neural network, which includes four convolution modules and three fully connected layers. The first three convolution modules include a convolution layer and a pooling layer, and the last convolution module only includes a convolution layer. The convolution kernel size and step size of each module are set separately, and the maximum pooling process is performed.

[0011] Step 4: Normalize the sample set to prevent gradient explosion during model training.

[0012] Step 5: Input the normalized data into the CNN network, initialize the weight vector and bias, calculate the error and perform training until the error converges, and finally store the parameter weights.

[0013] Step 6: Use prestack seismic data to perform AVO prestack inversion to accurately invert the P- and S-wave velocities and density parameters above and below the interface to assist in the identification of low-velocity mudstone.

[0014] Step 7: Low-velocity mudstone prediction: Extract the P-wave velocity, S-wave velocity and density data obtained from the pre-stack seismic gathers and AVO inversion, intercept the data in the target layer, normalize them and input them into the trained model to obtain the prediction results of low-velocity mudstone.

[0015] Furthermore, the classification is based on:

[0016] Background mudstone: the mud content is greater than 0.9, and the P- and S-wave velocities are greater than or equal to the upper and lower sandstone intervals;

[0017] Low-velocity mudstone: the mud content is greater than 0.9, and the P- and S-wave velocities are less than those of the upper and lower gas-bearing sandstone layers;

[0018] Gas-bearing sandstone: Well logging interpretation results indicate gas-bearing sandstone;

[0019] Dry sandstone: The logging interpretation result is dry sandstone.

[0020] Furthermore, in the step of generating a data set by geological body evolution, when wavelet analysis is performed using well logging data, the Ricker wavelet used is a wavelet that conforms to the post-stack seismic data of the well bypass track.

[0021] Furthermore, the convolution module in the CNN neural network uses maximum pooling to process the pooling layer.

[0022] Furthermore, the formula for normalization is:

[0023] X ′ =X-μ

[0024] In the formula, X is the original data and μ is the mean.

[0025] Furthermore, the angle range of the pre-stack seismic gather is 0° to 30°, and one data line is calculated every 3°.

[0026] Furthermore, the extracted pre-stack seismic gather data and the P-wave, S-wave velocity and density data obtained by AVO inversion are unified into a length of 200 after interpolation.

[0027] The present invention also discloses a low-speed mudstone intelligent prediction system, which can be used to implement the above-mentioned low-speed mudstone intelligent prediction method, specifically including:

[0028] Layer calibration module: It is used to divide the target layer into layers by combining logging data analysis and logging interpretation results. This module can divide the background mudstone, low-velocity mudstone, gas-bearing sandstone and dry layer sandstone, and determine the type of each layer according to parameters such as mud content and P- and S-wave velocities.

[0029] Geological body evolution data generation module: It is used to perform lateral and vertical changes on the geological body based on the division results of the layer segment calibration module, and generate a large number of data sets that conform to the laws of geological evolution. This module includes functions such as wavelet analysis, reflection coefficient calculation, pre-stack angle gather generation, and time-depth relationship calculation.

[0030] CNN neural network building module: used to build convolutional neural networks (CNN), including setting up convolutional layers, pooling layers, and fully connected layers. This module is responsible for initializing the network structure and selecting activation functions (such as ReLU functions), and configuring the convolution kernel size, step size, and pooling method.

[0031] Data normalization module: used to perform zero mean processing on the input sample data to prevent gradient explosion during model training. This module normalizes the data according to a specific formula.

[0032] Model training module: used to input the normalized data into the CNN neural network, initialize the weight vector and bias, calculate the error and optimize the network parameters until the error converges. This module finally outputs the trained model and related weight parameters.

[0033] AVO prestack inversion module: used to perform AVO prestack inversion based on prestack seismic data, accurately invert the P- and S-wave velocities and density parameters above and below the interface, and provide auxiliary information for the identification of low-velocity mudstone.

[0034] Low-velocity mudstone prediction module: It is used to extract relevant data from pre-stack seismic data and AVO inversion results, process and normalize them, input them into the trained CNN model, and output the prediction results of low-velocity mudstone.

[0035] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the above-mentioned low-speed mudstone intelligent prediction method is implemented.

[0036] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the above-mentioned low-speed mudstone intelligent prediction method is implemented.

[0037] Compared with the prior art, the advantages of the present invention are:

[0038] 1. Improve identification accuracy: By combining AVO pre-stack inversion and geological body evolution technology, it is possible to effectively distinguish low-velocity mudstone and gas-bearing sandstone, significantly improving the identification accuracy of medium-deep reservoirs and avoiding misjudgment caused by the "bright spot trap" of low-velocity mudstone.

[0039] 2. Strong adaptability: It can be applied to a variety of sedimentary basins with similar low-speed mudstone development characteristics and has wide applicability.

[0040] 3. Data-driven and intelligent: By building and training convolutional neural networks (CNNs), it is possible to process large amounts of complex seismic data and geological information, achieve intelligent prediction, and reduce errors caused by human subjective judgment.

[0041] 4. Reduce exploration risks: Accurately identify the boundaries of low-velocity mudstone and gas-bearing reservoirs, reduce uncertainty in exploration, reduce exploration risks caused by misjudgment of low-velocity mudstone, and improve the economic benefits of exploration.

[0042] 5. Automated process: The degree of automation is high. From data processing, model training to prediction result output, the entire process is completed automatically, saving a lot of time and human resources.

[0043] 6. Adapt to complex geological conditions: Through geological body evolution simulation and AVO inversion, it can adapt to complex geological conditions, such as areas with multi-layered mudstone development and complex sedimentary environment, and provide more accurate geological interpretation. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of the low-speed mudstone intelligent prediction method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and examples.

[0046] like Figure 1 As shown, the present invention provides a low-speed mudstone intelligent prediction method, comprising the following steps:

[0047] Step 1: Layer calibration. Combining the logging data analysis and logging interpretation results, the target layer can be divided into layers. The division result is the output of the neural network (Label). The division types include background mudstone, low-velocity mudstone, gas sandstone and dry layer sandstone. The division basis is as follows: ① Background mudstone: the mud content (0-1) is greater than 0.9, and the P- and S-wave velocities are greater than or equal to the upper and lower sandstone layers; ② Low-velocity mudstone: the mud content (0-1) is greater than 0.9, and the P- and S-wave velocities are less than the upper and lower gas sandstone layers; ③ Gas sandstone: the logging interpretation result is gas sandstone; ④ Dry layer sandstone: the logging interpretation result is dry layer.

[0048] Step 2: Evolution of geological bodies to generate data sets. The result of the layer division is the geological body. The geological body is changed horizontally and vertically within a certain probability range to generate a large amount of data to meet the data for neural network training. In the process of geological body evolution, the well logging data is first used for wavelet analysis to obtain the Ricker wavelet that conforms to the post-stack seismic data of the well side track; then the transformed geological body data is used to generate the logging data corresponding to the geological body, mainly including Vp, Vs and Den. Then the reflection coefficient corresponding to the geological body is calculated by combining the well logging data (Vp, Vs, Den) through the Zeopprits equation, and then the reflection coefficient matrix and the Ricker wavelet obtained by the wavelet analysis are convolved to obtain the pre-stack angle gather. The angle range of the angle gather is 0-30°, and one data is calculated every 3°. The final angle gather has 10 data. The TWT data of the "new" well is calculated using the time-depth relationship and Vp. The data length is set to 200, and the data finally obtained through geological body transformation includes TWT, prestack angle gathers, Vp, Vs, Den and Label.

[0049] Step 3: Construct a CNN neural network and select the ReLU function as the activation function. The CNN neural network includes convolutional layers and fully connected layers. There are four convolutional modules. The first three convolutional modules include a convolutional layer and a pooling layer. Here, we choose to use the maximum pooling (MaxPooling). The last convolutional module contains only one convolutional layer. The convolution kernel size of the convolutional layer of the first convolutional module is 3, the step size is 1, and no padding is performed. The convolution kernel size of the pooling layer is 3 and the step size is 2; the convolution kernel size of the convolutional layer of the second convolutional module is 2, the step size is 2, and no padding is performed. The convolution kernel size of the pooling layer is 3 and the step size is 2; the convolution kernel size of the convolutional layer of the third convolutional module is 2, the step size is 2, and the padding length is 1. The convolution kernel size of the pooling layer is 2 and the step size is 1; the convolution kernel size of the convolutional layer of the fourth convolutional module is 3, the step size is 2, and no padding is performed. The fully connected layer contains three linear layers.

[0050] Step 4: Data normalization. In order to prevent the problem of gradient explosion during model training, the sample set is processed with zero mean. The calculation formula is:

[0051] X ′ =X-μ

[0052] In the formula, X is the original data and μ is the mean.

[0053] Step 5: Model training. Input the normalized training set data into the CNN network, initialize the weight vector w and bias b, calculate the error between the node output and the expected output, and when the error converges, the training ends and the weights of each parameter are stored.

[0054] Step 5: AVO pre-stack inversion. Using pre-stack seismic data, AVOID pre-stack inversion can accurately invert the P- and S-wave velocities and density parameters above and below the interface, assisting in the identification of low-velocity mudstone.

[0055] Step 6: Low-velocity mudstone prediction. First, extract the pre-stack seismic gathers, and extract the single-channel data at the same position in the P- and S-wave velocities and density volumes obtained by AVO inversion. Interpolate the data in the target layer to a length of 200 and then perform normalization. Input the normalized data into the trained network model to obtain the final low-velocity mudstone prediction results.

[0056] In another embodiment of the present invention, a low-speed mudstone intelligent prediction system is provided. The system can be used to implement the above-mentioned low-speed mudstone intelligent prediction method, specifically including:

[0057] Layer calibration module: It is used to divide the target layer into layers by combining logging data analysis and logging interpretation results. This module can divide the background mudstone, low-velocity mudstone, gas-bearing sandstone and dry layer sandstone, and determine the type of each layer according to parameters such as mud content and P- and S-wave velocities.

[0058] Geological body evolution data generation module: It is used to perform lateral and vertical changes on the geological body based on the division results of the layer segment calibration module, and generate a large number of data sets that conform to the laws of geological evolution. This module includes functions such as wavelet analysis, reflection coefficient calculation, pre-stack angle gather generation, and time-depth relationship calculation.

[0059] CNN neural network building module: used to build convolutional neural networks (CNN), including setting up convolutional layers, pooling layers, and fully connected layers. This module is responsible for initializing the network structure and selecting activation functions (such as ReLU functions), and configuring the convolution kernel size, step size, and pooling method.

[0060] Data normalization module: used to perform zero mean processing on the input sample data to prevent gradient explosion during model training. This module normalizes the data according to a specific formula.

[0061] Model training module: used to input the normalized data into the CNN neural network, initialize the weight vector and bias, calculate the error and optimize the network parameters until the error converges. This module finally outputs the trained model and related weight parameters.

[0062] AVO prestack inversion module: used to perform AVO prestack inversion based on prestack seismic data, accurately invert the P- and S-wave velocities and density parameters above and below the interface, and provide auxiliary information for the identification of low-velocity mudstone.

[0063] Low-velocity mudstone prediction module: It is used to extract relevant data from pre-stack seismic data and AVO inversion results, process and normalize them, input them into the trained CNN model, and output the prediction results of low-velocity mudstone.

[0064] In another embodiment of the present invention, a terminal device is provided, the terminal device includes a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding functions; the processor described in the embodiment of the present invention can be used for the operation of the low-speed mudstone intelligent prediction method.

[0065] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It is understandable that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as at least one disk storage.

[0066] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the low-speed mudstone intelligent prediction method in the above embodiment; one or more instructions in the computer-readable storage medium are loaded and executed by the processor.

[0067] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the implementation methods of the present invention, and should be understood that the protection scope of the present invention is not limited to such special statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A low-speed mudstone intelligent prediction method, characterized in that: The following steps are involved: Step 1: Combine logging data analysis and logging interpretation results to divide the target layer into sections; the division results are used as the output Label of the neural network, and the division types include background mudstone, low-velocity mudstone, gas-bearing sandstone and dry layer sandstone; Step 2: Based on the results of the layer segmentation, the geological body is evolved horizontally and vertically to generate a large amount of data for neural network training; During the evolution of geological bodies, wavelet analysis is performed using logging data to generate prestack angle gathers, including TWT, prestack angle gathers, Vp, Vs, Den, and Label. Step 3: Select the ReLU function as the activation function and construct a CNN neural network, which includes four convolution modules and three fully connected layers; the first three convolution modules include a convolution layer and a pooling layer, and the last convolution module only includes one convolution layer; the convolution kernel size and step size of each module are set separately, and the maximum pooling process is performed; Step 4: Normalize the sample set to prevent gradient explosion during model training; Step 5: Input the normalized data into the CNN network, initialize the weight vector and bias, calculate the error and train until the error converges, and finally store the parameter weights; Step 6: Use prestack seismic data to perform AVO prestack inversion to accurately invert the P- and S-wave velocities and density parameters above and below the interface to assist in the identification of low-velocity mudstone; Step 7: Low-velocity mudstone prediction: Extract the P-wave velocity, S-wave velocity and density data obtained from the pre-stack seismic gathers and AVO inversion, intercept the data in the target layer, normalize them and input them into the trained model to obtain the prediction results of low-velocity mudstone.

2. The low-speed mudstone intelligent prediction method according to claim 1 is characterized by: The classification is based on: Background mudstone: the mud content is greater than 0.9, and the P- and S-wave velocities are greater than or equal to the upper and lower sandstone intervals; Low-velocity mudstone: the mud content is greater than 0.9, and the P- and S-wave velocities are less than those of the upper and lower gas-bearing sandstone layers; Gas-bearing sandstone: Well logging interpretation results indicate gas-bearing sandstone; Dry sandstone: The logging interpretation result is dry sandstone.

3. The low-speed mudstone intelligent prediction method according to claim 1 is characterized by: In the step of generating a data set by geological body evolution, when wavelet analysis is performed using well logging data, the Ricker wavelet used is a wavelet that conforms to the post-stack seismic data of the well bypass track.

4. The low-speed mudstone intelligent prediction method according to claim 1 is characterized by: The convolution module in the CNN neural network uses the maximum pooling process pooling layer.

5. The low-speed mudstone intelligent prediction method according to claim 1 is characterized by: The normalization formula is: X ′ =X-μ In the formula, X is the original data and μ is the mean.

6. The low-speed mudstone intelligent prediction method according to claim 1 is characterized by: The angle range of the pre-stack seismic gather is from 0° to 30°, and a data track is calculated every 3°.

7. The low-speed mudstone intelligent prediction method according to claim 1 is characterized by: The extracted pre-stack seismic gather data and the P-wave, S-wave velocity and density data obtained by AVO inversion are unified into a length of 200 after interpolation.

8. A low-speed mudstone intelligent prediction system, which can be used to implement a low-speed mudstone intelligent prediction method according to any one of claims 1 to 7, specifically comprising: Layer calibration module: used to divide the target layer into layers by combining logging data analysis and logging interpretation results; this module can divide the background mudstone, low-velocity mudstone, gas-bearing sandstone and dry layer sandstone, and determine the type of each layer according to parameters such as mud content and P- and S-wave velocities; Geological body evolution data generation module: used to perform lateral and vertical changes on the geological body based on the division results of the layer segment calibration module, and generate a large number of data sets that conform to the laws of geological evolution; this module includes functions such as wavelet analysis, reflection coefficient calculation, pre-stack angle gather generation and time-depth relationship calculation; CNN neural network building module: used to build convolutional neural network (CNN), including setting convolution layer, pooling layer and fully connected layer; this module is responsible for initializing the network structure and selecting activation function (such as ReLU function), and configuring convolution kernel size, step size and pooling method; Data normalization module: used to perform zero mean processing on the input sample data to prevent gradient explosion during model training; this module normalizes the data according to a specific formula; Model training module: used to input normalized data into the CNN neural network, initialize weight vectors and biases, calculate errors and optimize network parameters until the error converges; This module finally outputs the trained model and related weight parameters; AVO pre-stack inversion module: used to perform AVO pre-stack inversion based on pre-stack seismic data, accurately invert the P- and S-wave velocities and density parameters above and below the interface, and provide auxiliary information for the identification of low-velocity mudstone; Low-velocity mudstone prediction module: It is used to extract relevant data from pre-stack seismic data and AVO inversion results, process and normalize them, input them into the trained CNN model, and output the prediction results of low-velocity mudstone.

9. A computer device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the low-speed mudstone intelligent prediction method according to one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the low-speed mudstone intelligent prediction method according to one of claims 1 to 7 is implemented.

Images