CN115130529A - Pre-stack seismic inversion method and device - Google Patents
Pre-stack seismic inversion method and device Download PDFInfo
- Publication number
- CN115130529A CN115130529A CN202110294942.5A CN202110294942A CN115130529A CN 115130529 A CN115130529 A CN 115130529A CN 202110294942 A CN202110294942 A CN 202110294942A CN 115130529 A CN115130529 A CN 115130529A
- Authority
- CN
- China
- Prior art keywords
- information
- layer
- neural network
- determining
- convolutional neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 99
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 93
- 239000011435 rock Substances 0.000 claims abstract description 40
- 238000004088 simulation Methods 0.000 claims abstract description 34
- 238000011160 research Methods 0.000 claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 22
- 230000006870 function Effects 0.000 claims description 23
- 238000004590 computer program Methods 0.000 claims description 16
- 238000011161 development Methods 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 12
- 238000005259 measurement Methods 0.000 claims description 12
- 230000008021 deposition Effects 0.000 claims description 10
- 238000009826 distribution Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000003860 storage Methods 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000007405 data analysis Methods 0.000 claims description 6
- 230000002349 favourable effect Effects 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 230000009286 beneficial effect Effects 0.000 claims description 5
- 238000000638 solvent extraction Methods 0.000 claims description 4
- 239000000203 mixture Substances 0.000 claims description 2
- 238000004062 sedimentation Methods 0.000 claims 2
- 238000010586 diagram Methods 0.000 description 20
- 238000005457 optimization Methods 0.000 description 17
- 230000018109 developmental process Effects 0.000 description 11
- 238000004422 calculation algorithm Methods 0.000 description 9
- 238000012545 processing Methods 0.000 description 8
- 230000008859 change Effects 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 208000035126 Facies Diseases 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 239000004576 sand Substances 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 239000004215 Carbon black (E152) Substances 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 229930195733 hydrocarbon Natural products 0.000 description 1
- 150000002430 hydrocarbons Chemical class 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000000704 physical effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 239000013049 sediment Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/282—Application of seismic models, synthetic seismograms
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/30—Analysis
- G01V1/306—Analysis for determining physical properties of the subsurface, e.g. impedance, porosity or attenuation profiles
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/40—Seismology; Seismic or acoustic prospecting or detecting specially adapted for well-logging
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Remote Sensing (AREA)
- General Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geophysics (AREA)
- Theoretical Computer Science (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention provides a prestack seismic inversion method and a prestack seismic inversion device, wherein the method comprises the following steps: according to the prior information, reservoir beds of the research area are divided, and an area classification result is determined; determining label information according to the rock physical information, the prior information, the region classification result and a random geological simulation method constrained by logging information; classifying and marking the label information by using the region classification result, and determining label data with classification marks; building a convolutional neural network; training a convolutional neural network by using label data with classification labels, and determining the trained convolutional neural network; and performing pre-stack seismic inversion according to the trained convolutional neural network, and determining an elastic parameter inversion result. By means of the strong nonlinear problem solving capability of the convolutional neural network, the precision of the pre-stack seismic inversion is improved, and a more accurate elastic parameter prediction result is obtained.
Description
Technical Field
The invention relates to the technical field of geophysical, in particular to a pre-stack seismic inversion method and a pre-stack seismic inversion device.
Background
This section is intended to provide a background or context to the embodiments of the invention that are recited in the claims. The description herein is not admitted to be prior art by inclusion in this section.
The seismic waves are reflected when encountering a reflection interface in the underground propagation process, the amplitude of the reflected waves changes along with the change of an incident angle when the seismic waves are not vertically incident to the reflection interface, and various elastic parameters reflecting the physical property information of the underground reservoir are obtained from pre-stack seismic data containing rich dynamic information by researching the change rule of the amplitude of the seismic waves along with the incident angle, so that the quantitative basis is provided for reservoir prediction and hydrocarbon detection. At present, the inversion algorithm researched and applied in the research and industrial fields is based on the Zoeppritz equation and its approximate formula, and the Zoeppritz equation describes the energy distribution relation of the elastic wave on the elastic interface.
However, the application of the theoretical equation is that the underground medium is assumed to be isotropic and horizontal layered medium, and two sides of the interface are semi-infinite spaces, however, the actual distribution situation of the underground medium is complex and greatly different from the assumed situation, and the theoretical equation cannot accurately represent the change situation of the amplitude of the reflected wave along with the incident angle, so that certain error exists in the elastic parameter prediction by applying the theoretical equation to the prestack AVO inversion method.
Therefore, the theoretical equation cannot accurately describe the propagation rule of the seismic waves under the complex geological condition and the strong nonlinearity of the inversion problem.
Therefore, how to provide a new solution, which can solve the above technical problems, is a technical problem to be solved in the art.
Disclosure of Invention
The embodiment of the invention provides a prestack seismic inversion method, which improves the precision of prestack seismic inversion by means of the strong nonlinear problem solving capability of a convolutional neural network and obtains a more accurate elastic parameter prediction result, and the method comprises the following steps:
according to the prior information, reservoir beds of the research area are divided, and an area classification result is determined;
determining label information according to the rock physical information, the prior information, the region classification result and a random geological simulation method constrained by logging information;
classifying and marking the label information by using the region classification result, and determining label data with classification marks;
building a convolutional neural network;
training a convolutional neural network by using label data with classification labels, and determining the trained convolutional neural network;
and performing pre-stack seismic inversion according to the trained convolutional neural network, and determining an elastic parameter inversion result.
The embodiment of the invention also provides a pre-stack seismic inversion device, which comprises:
the reservoir dividing module is used for dividing the reservoir of the research area according to the prior information and determining the area classification result;
the tag information determining module is used for determining tag information according to the rock physical information, the prior information, the region classification result and a random geological simulation method constrained by the logging information;
the classification marking module is used for performing classification marking on the label information by using the region classification result and determining label data with the classification marking;
the convolutional neural network building module is used for building a convolutional neural network;
the convolutional neural network training module is used for training a convolutional neural network by using the label data with the classification labels and determining the trained convolutional neural network;
and the pre-stack seismic inversion module is used for performing pre-stack seismic inversion according to the trained convolutional neural network and determining an elastic parameter inversion result.
The embodiment of the invention also provides computer equipment which comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein the processor executes the computer program to realize the pre-stack seismic inversion method.
An embodiment of the present invention further provides a computer-readable storage medium, where a computer program for executing the pre-stack seismic inversion method is stored in the computer-readable storage medium.
The embodiment of the invention provides a pre-stack seismic inversion method and a device, which comprises the following steps: firstly, reservoir beds of a research region are divided according to prior information, and a region classification result is determined; then, determining label information according to the rock physical information, the prior information, the region classification result and a random geological simulation method constrained by logging information; classifying and marking the label information by using the region classification result, and determining label data with classification marks; building a convolutional neural network; training a convolutional neural network by using the label data with the classification labels, and determining the trained convolutional neural network; and finally, performing pre-stack seismic inversion according to the trained convolutional neural network, and determining an elastic parameter inversion result. The method has the advantages that the priori information is fully utilized to constrain the prestack AVO three-parameter inversion based on the convolutional neural network, the convolutional neural network is not simply applied to construct the nonlinear relation between the prestack common reflection point trace set and the elastic parameters to be inverted, the solving space of the optimization problem can be reduced through the constraint of the priori information, and the stability and the accuracy of the optimization problem can be improved. In addition, label data obtained by a random geological simulation method provides a large amount of label data for network training, so that optimization of a network model is facilitated, and generalization capability of the network is improved, so that final prediction capability of the network is improved. The embodiment of the invention improves the precision of the pre-stack seismic inversion by means of the powerful capability of solving the nonlinear problem of the convolutional neural network, and obtains a more accurate elastic parameter prediction result.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts. In the drawings:
fig. 1 is a schematic diagram of a pre-stack seismic inversion method according to an embodiment of the present invention.
FIG. 2 is a flow chart of a pre-stack seismic inversion method according to an embodiment of the invention.
Fig. 3 is a schematic diagram of a tag data generation process of a pre-stack seismic inversion method according to an embodiment of the present invention.
Fig. 4 is a schematic diagram of a convolutional neural network structure of a pre-stack seismic inversion method according to an embodiment of the present invention.
FIG. 5 is a schematic diagram illustrating the change of an error with the increase of iteration times in a network training process of a pre-stack seismic inversion method according to an embodiment of the present invention.
FIG. 6 is a schematic diagram of inversion results of a pre-stack seismic inversion method according to an embodiment of the invention.
FIG. 7 is a schematic diagram of a computer apparatus for operating a method for prestack seismic inversion in accordance with embodiments of the present invention.
FIG. 8 is a schematic diagram of a pre-stack seismic inversion apparatus according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the embodiments of the present invention are further described in detail below with reference to the accompanying drawings. The exemplary embodiments and descriptions of the present invention are provided to explain the present invention, but not to limit the present invention.
Fig. 1 is a schematic diagram of a pre-stack seismic inversion method according to an embodiment of the present invention, and as shown in fig. 1, an embodiment of the present invention provides a pre-stack seismic inversion method, which improves accuracy of pre-stack seismic inversion by means of strong capability of solving a nonlinear problem in a convolutional neural network, and obtains a more accurate elastic parameter prediction result, where the method includes:
step S01: according to the prior information, reservoir beds of the research area are divided, and an area classification result is determined;
step S02: determining label information according to the rock physical information, the prior information, the region classification result and a random geological simulation method constrained by logging information;
step S03: classifying and marking the label information by using the region classification result, and determining label data with classification marks;
step S04: building a convolutional neural network;
step S05: training a convolutional neural network by using label data with classification labels, and determining the trained convolutional neural network;
step S06: and performing pre-stack seismic inversion according to the trained convolutional neural network, and determining an elastic parameter inversion result.
The pre-stack seismic inversion method provided by the embodiment of the invention comprises the following steps: firstly, reservoir beds of a research region are divided according to prior information, and a region classification result is determined; then, determining label information according to the rock physical information, the prior information, the region classification result and a random geological simulation method constrained by logging information; secondly, classifying and marking the label information by using the region classification result, and determining label data with classification marks; building a convolutional neural network; training a convolutional neural network by using the label data with the classification labels, and determining the trained convolutional neural network; and finally, performing prestack AVO (amplitude verses offset) three-parameter inversion in prestack seismic inversion according to the trained convolutional neural network, and determining an elastic parameter inversion result. The method has the advantages that the priori information is fully utilized to constrain the prestack AVO three-parameter inversion based on the convolutional neural network, the convolutional neural network is not simply applied to construct the nonlinear relation between the prestack common reflection point trace set and the elastic parameters to be inverted, the solving space of the optimization problem can be reduced through the constraint of the priori information, and the stability and the accuracy of the optimization problem can be improved. In addition, label data obtained by a random geological simulation method provides a large amount of label data for network training, so that optimization of a network model is facilitated, and generalization capability of the network is improved, so that final prediction capability of the network is improved. The embodiment of the invention improves the precision of three-parameter inversion of the pre-stack AVO (amplitude Versus offset) by means of the powerful capability of solving the nonlinear problem of the convolutional neural network, and obtains a more accurate elastic parameter prediction result.
With reference to the flow shown in fig. 2, the implementation of the method for prestack seismic inversion according to the embodiment of the present invention may include: according to the prior information, reservoir stratums of a research area are divided, and an area classification result is determined; determining label information according to the rock physical information, the prior information, the region classification result and a random geological simulation method constrained by logging information; classifying and marking the label information by using the region classification result, and determining label data with classification marks; building a convolutional neural network; training a convolutional neural network by using label data with classification labels, and determining the trained convolutional neural network; and performing pre-stack seismic inversion according to the trained convolutional neural network, and determining an elastic parameter inversion result.
Convolutional Neural Networks (CNN), one of the deep learning methods, can establish a nonlinear relationship between input and output by training to continuously optimize the network structure. The CNN can realize high-dimensional feature extraction and automatic pattern recognition, so that the problems of strong nonlinearity, image classification, image segmentation, target recognition, natural language processing and the like can be solved. In recent years, CNNs have also been used primarily in seismic exploration, such as fault recognition, seismic facies classification, and noise processing. In view of the fact that a theoretical equation cannot accurately describe the propagation rule of seismic waves under complex geological conditions and strong nonlinearity of a prestack AVO (amplitude Versus offset) three-parameter inversion problem, the method provides that a CNN algorithm is used for constructing a nonlinear relation between a prestack gather and elastic parameters to be inverted in a data-driven mode, and inversion errors caused by inaccuracy of the seismic waves propagated in an underground medium by applying a mathematical equation are reduced, so that high-precision prestack AVO three-parameter inversion is finally achieved. A prestack AVO (amplitude Versus offset) inversion method belongs to one of prestack seismic inversion methods, and can realize inversion prediction with more accurate elastic parameters.
The embodiment of the invention provides a technical method for performing prestack AVO (amplitude Versus offset) inversion prediction on longitudinal wave velocity, transverse wave velocity and density by applying a convolutional neural network algorithm (CNN) in a deep learning method. By utilizing the powerful capability of CNN for constructing the nonlinear mapping relation between input and output, the nonlinear mapping relation between the prestack seismic common reflection point gather and three parameters of longitudinal wave velocity, transverse wave velocity and density is established, and a data-driven prestack AVO (amplitude verse returns) three-parameter inversion method is formed.
When the pre-stack seismic inversion method provided by the embodiment of the present invention is specifically implemented, in an embodiment, the aforementioned partitioning the reservoir of the research region according to the prior information to determine the region classification result includes:
according to prior information consisting of geological structure information, deposition information, a logging data analysis result, a seismic attribute analysis result and single well lithofacies information, dividing a reservoir of a research area into a reservoir favorable area and a reservoir non-favorable area, and determining an area classification result; wherein, the region classification result comprises: the beneficial development area of the first type of reservoir, the medium development area of the second type of reservoir and the underdevelopment area of the third type of reservoir.
In the embodiment, the research target area is divided into a beneficial development area of a type I reservoir, a medium development area of a type II reservoir and a less development area of a type III reservoir according to the prior information of the regional structure, the deposition information, the logging data analysis and the seismic attribute analysis, and the prestack common reflection point gather data is classified and marked.
Fig. 3 is a schematic diagram of a tag data generation process of a prestack seismic inversion method according to an embodiment of the present invention, and as shown in fig. 3, when a prestack seismic inversion method provided by an embodiment of the present invention is specifically implemented, in an embodiment, the determining tag information according to the stochastic geologic simulation method constrained by petrophysical information, prior information, a region classification result, and logging information includes:
analyzing rock physical information and determining rock core measurement data;
analyzing the core measurement data, carrying out a rock physical model test by combining logging information, and establishing a rock physical model conforming to a research area;
acquiring geological structure information and deposition information in prior information, integrating geological and geophysical information, and acquiring different types of virtual well curves by using lithofacies probability distribution conditions of different classification areas and combining a random geological simulation method on the basis of region classification results;
modeling by utilizing a rock physical model according to the virtual well curve to determine a simulated elastic parameter;
performing forward modeling of a seismic wave field by using the simulated elastic parameters, and determining a prestack common reflection point gather;
and correspondingly combining the prestack common reflection point gather and the simulation elastic parameters to determine the label data.
In the embodiment, the CNN algorithm is used as one of deep neural network algorithms, belongs to a supervised learning algorithm, needs a large number of data labels to train the network and optimize network model parameters, finally obtains an optimized network model, and then performs prediction work by using the network after training optimization is completed. In the field of oil exploration, well data obtained through well logging work can be used as tag data, but an exploration block has only a few wells in the early exploration stage, so that the obtained tag data quantity cannot meet the training requirement of the CNN network. The embodiment provides a random geological simulation method using rock physical information, geological structure and sediment information and logging information constraint to realize tag data generation.
Firstly, analyzing rock physical information and determining rock core measurement data;
and analyzing the core measurement data, and performing a rock physical model test by combining the well-drilled well logging information including information such as longitudinal wave velocity, transverse wave velocity, density, porosity, water saturation, shale content and the like, and preferably and constructing a rock physical model conforming to the research area.
Based on the obtained regional classification result, different types of virtual well curves including facies, porosity, shale content and saturation are obtained by using facies probability distribution conditions of different classification regions and combining a random geological simulation method. And finally, modeling by utilizing the rock physical model constructed and selected in the early stage to obtain corresponding simulated elastic parameters including longitudinal wave velocity, transverse wave velocity and density.
And finally, performing wave field forward modeling by using the elastic parameters obtained by simulation to obtain pre-stack common reflection point gathers, wherein one pre-stack common reflection point gather corresponds to a set of elastic parameter curves (a longitudinal wave velocity curve, a transverse wave velocity curve and a density curve), and the two gather form a group of label data. In addition, the label data are classified and marked according to the virtual well type adopted by the forward simulation record, and the generated label data are all clearly classified.
Fig. 4 is a schematic structural diagram of a convolutional neural network of a pre-stack seismic inversion method according to an embodiment of the present invention, and as shown in fig. 4, when the pre-stack seismic inversion method provided by the embodiment of the present invention is specifically implemented, in an embodiment, the building of the convolutional neural network includes:
constructing a first convolution layer, a second convolution layer and a full-connection layer; each convolutional layer comprises three parts of convolution operation, an activation function and batch standardization;
according to the functional connection sequence, an input layer, a first convolution layer, a second convolution layer, a full connection layer and an output layer are connected in sequence to build a convolution neural network; wherein, the function connection order includes: the input layer receives label data with classification marks; connecting the input layer to the input of the first convolution layer; taking the output of the first convolution layer as the input of the second convolution layer; connecting the output of the second convolution layer into a one-dimensional vector as the input of the full-connected layer; and connecting the output of the full connection layer with an output layer, and outputting an inversion result to be predicted.
When the pre-stack seismic inversion method provided by the embodiment of the invention is specifically implemented, in one embodiment, a convolutional layer can be constructed as follows:
ReLU＝max(0,x) (2)
wherein BN (-) is batch standardization processing; ReLU is an activation function, denoted by multiply, generally expressed as star multiply;
is the output result of the convolutional layer; w is a k Is a convolution operator; x is input data; b k Is a bias term; i is the row index of the matrix; j is the column index of the matrix.
The above expressions for constructing convolutional layers are exemplary, and those skilled in the art will understand that the above formulas may be modified in certain forms and other parameters or data may be added or other specific formulas may be provided according to the needs, and such modifications are within the scope of the present invention.
The input data of the convolutional neural network is a prestack common reflection point gather and a reservoir type, and the output data is corresponding longitudinal wave velocity, transverse wave velocity and density. The embodiment of the invention builds a three-layer convolutional neural network model, and the structure of the model is shown in figure 4, and the model comprises two convolutional layers and a full-connection layer. Each convolutional layer comprises three parts, namely convolution operation, an activation function and batch standardization, wherein the activation function adopts a ReLU function as shown in a formula (1), and is shown in a formula (2), and BN (·) is batch standardization processing. And connecting the output of the second layer of convolution layer into a one-dimensional vector as the input of the full-connection layer, and finally outputting the three parameters of the elasticity to be predicted from the full-connection layer.
The embodiment of the invention provides a technical method for predicting longitudinal wave velocity, transverse wave velocity and density by applying convolution neural network algorithm (CNN) in a deep learning method to perform prestack AVO inversion, and the technical method is characterized in that nonlinear mapping relations between a prestack earthquake CRP gather and three parameters of the longitudinal wave velocity, the transverse wave velocity and the density are established by utilizing the strong capability of constructing the nonlinear mapping relations between input and output of the CNN, so that a data-driven prestack AVO three-parameter inversion method is formed. In one example of the embodiment of the present invention, the method mainly includes:
and dividing the research target into a beneficial development area of a type I reservoir, a medium development area of a type II reservoir and an underdeveloped area of a type III reservoir according to the prior information of the regional structure, the deposition information, the logging data analysis and the seismic attribute analysis, and performing classification marking on the pre-stack CRP gather data.
The label data generation is realized by utilizing a stochastic geological simulation method constrained by rock physical information, geological structure and deposition information and logging information: analyzing the core measurement data of the research area, combining the well-drilled well logging data including information such as longitudinal wave velocity, transverse wave velocity, density, porosity, water saturation, shale content and the like, and preferably and constructing a petrophysical model conforming to the research area. Based on the regional classification result, by using lithofacies probability distribution conditions of different classification regions and combining a random geological simulation method, different types of virtual well curves including lithofacies, porosity, shale content and saturation are obtained. And finally, modeling by utilizing the rock physical model determined and selected in the early stage to obtain the corresponding longitudinal wave velocity, transverse wave velocity and density. Performing wave field forward modeling by using the elastic parameters obtained by simulation to obtain pre-stack CRP gathers, wherein one pre-stack CRP gather corresponds to a set of elastic parameter curves (a longitudinal wave velocity curve, a transverse wave velocity curve and a density curve), and the two gather form a group of label data. In addition, the tag data is classified and marked according to the virtual well type adopted by the forward simulation record.
Building a CNN network: the input data of the network are pre-stack CRP gathers and reservoir types, and the output is corresponding longitudinal wave velocity, transverse wave velocity and density. And constructing a three-layer convolution neural network model which comprises two convolution layers and a full connection layer. Each convolution layer comprises three parts of convolution operation, an activation function and batch standardization, the activation function adopts a ReLU function, the output of the convolution layer of the second layer is connected into a one-dimensional vector which is used as the input of the full-connection layer, and finally the three parameters of the elasticity to be predicted are output from the full-connection layer.
According to the embodiment of the invention, geological and geophysical knowledge is fully utilized as prior information to constrain the prestack AVO (amplitude Versus offset) three-parameter inversion based on the convolutional neural network, the convolutional neural network is not simply applied to construct the nonlinear relation between the prestack common reflection point channel set and the elastic parameters to be inverted, the solving space of the optimization problem can be reduced through the constraint of the prior information, and the stability and the accuracy of the optimization problem can be improved. In addition, label data obtained by a random geological simulation method based on geological information and petrophysical information provides a large amount of label data for network training, so that optimization of a network model is facilitated, generalization capability of a network is improved, and final prediction capability of the network is improved.
In the implementation process of the prestack seismic inversion method, geological and geophysical information is used as constraints, a convolutional neural network algorithm is introduced into a prestack seismic inversion task, and by means of the powerful nonlinear problem solving capability of the convolutional neural network, the precision of three-parameter inversion of prestack AVO (amplitude returns) is improved, and a more accurate elastic parameter prediction result is obtained.
Fig. 2 is a flowchart of a pre-stack seismic inversion method according to an embodiment of the present invention, and as shown in fig. 2, an embodiment of the present invention further provides a flow of a pre-stack seismic inversion method, including:
step 101: the reservoir favorable area and the reservoir non-favorable area are divided according to the geological structure, geological deposition, seismic attributes and single-well lithofacies information of the area where the research area is located as shown in figure 3, and the sandstone reservoirs, namely the areas with higher sand area and lower sand area, are generally divided into three types, namely a type I reservoir favorable development area, a type II reservoir medium development area and a type III reservoir under-development area.
Step 102: and carrying out a rock physical modeling test according to the core measurement data and the logging data, and preferably selecting a theoretical rock physical model suitable for researching a target layer of the work area to carry out subsequent modeling. Then, the geological and geophysical information is integrated, and the virtual well data is generated by using a stochastic geologic simulation technique.
Step 103: and performing forward modeling on seismic waves according to the well data generated in the last step to obtain pre-stack seismic data, so as to construct a tag data set. The generated label data is of definite classification, belongs to one of three classes divided in step 101, and serves as data and the prestack common reflection point gather as input of the network.
Step 104: and building a CNN network, as shown in FIG. 4. The input data is a prestack common reflection point gather or an angle gather and partition information, and the output data is longitudinal wave velocity, transverse wave velocity and density. The optimization objective function of the model network gradually decreases as the number of iterations increases, as shown in fig. 5.
Step 105: utilizing the trained network to carry out prestack AVO (amplitude Versus offset) three-parameter inversion to obtain longitudinal wave velocity, transverse wave velocity and density, and obtaining an inversion result as shown in figure 6, wherein a predicted value is approximately consistent with a real value.
Fig. 7 is a schematic diagram of a computer device for executing a prestack seismic inversion method implemented by the present invention, and as shown in fig. 7, an embodiment of the present invention further provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor executes the computer program to implement the prestack seismic inversion method.
An embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program for implementing the pre-stack seismic inversion method.
The embodiment of the invention also provides a pre-stack seismic inversion device, which is described in the following embodiment. The principle of the device for solving the problems is similar to that of a prestack seismic inversion method, so the implementation of the device can be referred to that of the prestack seismic inversion method, and repeated parts are not described again.
Fig. 8 is a schematic diagram of a pre-stack seismic inversion apparatus according to an embodiment of the present invention, and as shown in fig. 8, an embodiment of the present invention further provides a pre-stack seismic inversion apparatus, including:
the reservoir dividing module 801 is used for dividing reservoirs of the research region according to the prior information and determining a region classification result;
the tag information determining module 802 is configured to determine tag information according to the petrophysical information, the prior information, the region classification result, and a random geological simulation method constrained by the logging information;
a classification marking module 803, configured to perform classification marking on the tag information according to the region classification result, and determine tag data with the classification marking;
a convolutional neural network building module 804, configured to build a convolutional neural network;
a convolutional neural network training module 805, configured to train a convolutional neural network using the tag data with the classification label, and determine the trained convolutional neural network;
the prestack seismic inversion module 806 is configured to perform prestack avo (amplitude verses offset) three-parameter inversion according to the trained convolutional neural network, and determine an elastic parameter inversion result.
When the pre-stack seismic inversion apparatus provided in the embodiment of the present invention is implemented specifically, in an embodiment, the reservoir partitioning module is specifically configured to:
according to prior information consisting of geological structure information, deposition information, a logging data analysis result, a seismic attribute analysis result and single well lithofacies information, dividing reservoirs of a research area, and determining an area classification result; wherein, the region classification result comprises: the beneficial development area of the first type of reservoir, the medium development area of the second type of reservoir and the underdevelopment area of the third type of reservoir.
In an embodiment of the invention, when the pre-stack seismic inversion apparatus provided in the embodiment of the present invention is implemented specifically, the tag information determination module is specifically configured to:
analyzing rock physical information and determining rock core measurement data;
analyzing the core measurement data, carrying out a rock physical model test by combining logging information, and establishing a rock physical model conforming to a research area;
acquiring geological structure information and deposition information in prior information, integrating geological and geophysical information, and acquiring virtual well curves of different types by using lithofacies probability distribution conditions of different classification areas and combining a random geological simulation method on the basis of region classification results;
modeling by utilizing a rock physical model according to the virtual well curve to determine a simulated elastic parameter;
performing forward modeling of a seismic wave field by using the simulated elastic parameters, and determining a prestack common reflection point gather;
and correspondingly combining the prestack common reflection point gather and the simulation elastic parameters to determine the label data.
When the pre-stack seismic inversion device provided by the embodiment of the invention is specifically implemented, in an embodiment, the convolutional neural network building module is specifically used for:
constructing a first convolution layer, a second convolution layer and a full-connection layer; each convolutional layer comprises three parts of convolution operation, an activation function and batch standardization;
according to the functional connection sequence, an input layer, a first convolution layer, a second convolution layer, a full connection layer and an output layer are connected in sequence to build a convolution neural network; wherein, the function connection order includes: the input layer receives label data with classification marks; connecting the input layer to the input of the first convolution layer; taking the output of the first convolution layer as the input of the second convolution layer; connecting the output of the second convolution layer into a one-dimensional vector as the input of the full-connected layer; and connecting the output of the full connection layer with an output layer, and outputting an inversion result to be predicted.
In a specific implementation of the pre-stack seismic inversion apparatus provided in the embodiment of the present invention, in an embodiment, the convolutional neural network building module is further configured to build a convolutional layer according to the following manner:
ReLU＝max(0,x)
wherein BN (-) is batch standardization processing; ReLU is the activation function, denoted multiply;
is the output result of the convolutional layer; w is a k Is a convolution operator; x is input data; b is a mixture of k Is a bias term; i is the row index of the matrix; j is the column index of the matrix.
To sum up, the pre-stack seismic inversion method and device provided by the embodiment of the invention comprise: firstly, dividing reservoirs in a research area according to prior information, and determining an area classification result; then, determining label information according to the rock physical information, the prior information, the region classification result and a random geological simulation method constrained by logging information; classifying and marking the label information by using the region classification result, and determining label data with classification marks; building a convolution neural network; training a convolutional neural network by using the label data with the classification labels, and determining the trained convolutional neural network; and finally, according to the trained convolutional neural network, performing three-parameter inversion of AVO (amplitude Versus offset) before stack, and determining an elastic parameter inversion result. The method has the advantages that the priori information is fully utilized to constrain the prestack AVO three-parameter inversion based on the convolutional neural network, the convolutional neural network is not simply applied to construct the nonlinear relation between the prestack common reflection point trace set and the elastic parameters to be inverted, the solving space of the optimization problem can be reduced through the constraint of the priori information, and the stability and the accuracy of the optimization problem can be improved. In addition, label data obtained by a random geological simulation method provides a large amount of label data for network training, so that optimization of a network model is facilitated, and generalization capability of the network is improved, so that final prediction capability of the network is improved. The embodiment of the invention improves the precision of three-parameter inversion of the pre-stack AVO (amplitude Versus offset) by means of the powerful capability of solving the nonlinear problem of the convolutional neural network, and obtains a more accurate elastic parameter prediction result.
According to the invention, geological and geophysical knowledge is fully utilized as prior information to constrain the prestack AVO (amplitude Versus offset) three-parameter inversion based on the convolutional neural network, the nonlinear relation between the prestack common reflection point channel set and the elastic parameters to be inverted is not established simply by applying the convolutional neural network, the solving space of the optimization problem can be reduced through the constraint of the prior information, and the stability and the accuracy of the optimization problem can be improved. In addition, label data obtained by a random geological simulation method based on geological information and petrophysical information provides a large amount of label data for network training, so that optimization of a network model is facilitated, generalization capability of a network is improved, and final prediction capability of the network is improved.
The method takes geological and geophysical information as constraints, introduces a convolutional neural network algorithm into a pre-stack seismic inversion task, and improves the precision of the pre-stack AVO (amplitude verses offset) three-parameter inversion by means of the strong capability of solving the nonlinear problem of the convolutional neural network, thereby obtaining a more accurate elastic parameter prediction result.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention has been described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above-mentioned embodiments are provided to further explain the objects, technical solutions and advantages of the present invention in detail, and it should be understood that the above-mentioned embodiments are only examples of the present invention and should not be used to limit the scope of the present invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (12)
1. A method of prestack seismic inversion, comprising:
according to the prior information, reservoir beds of the research area are divided, and an area classification result is determined;
determining label information according to the rock physical information, the prior information, the region classification result and a random geological simulation method constrained by logging information;
classifying and marking the label information by using the region classification result, and determining label data with classification marks;
building a convolutional neural network;
training a convolutional neural network by using label data with classification labels, and determining the trained convolutional neural network;
and performing pre-stack seismic inversion according to the trained convolutional neural network, and determining an elastic parameter inversion result.
2. The method of claim 1, wherein the determining the region classification results by partitioning the reservoir of the study region according to the prior information comprises:
dividing reservoirs in a research area according to prior information consisting of geological structure information, sedimentation information, a logging data analysis result, a seismic attribute analysis result and single-well lithofacies information, and determining an area classification result; wherein, the region classification result comprises: the beneficial development area of the first type of reservoir, the medium development area of the second type of reservoir and the underdevelopment area of the third type of reservoir.
3. The method of claim 2, wherein determining label information from petrophysical information, prior information, region classification results, and a stochastic geologic modeling method constrained by well log information comprises:
analyzing rock physical information and determining rock core measurement data;
analyzing the core measurement data, carrying out a rock physical model test by combining logging information, and establishing a rock physical model conforming to a research area;
acquiring geological structure information and deposition information in prior information, integrating geological and geophysical information, and acquiring virtual well curves of different types by using lithofacies probability distribution conditions of different classification areas and combining a random geological simulation method on the basis of region classification results;
modeling by utilizing a rock physical model according to the virtual well curve to determine a simulated elastic parameter;
performing forward modeling of a seismic wave field by using the simulated elastic parameters, and determining a prestack common reflection point gather;
and correspondingly combining the prestack common reflection point gather and the simulation elastic parameters to determine the label data.
4. The method of claim 1, wherein building a convolutional neural network comprises:
constructing a first convolution layer, a second convolution layer and a full-connection layer; each convolution layer comprises three parts, namely convolution operation, an activation function and batch standardization;
according to the functional connection sequence, an input layer, a first convolution layer, a second convolution layer, a full connection layer and an output layer are connected in sequence to build a convolution neural network; wherein, the function connection order includes: the input layer receives label data with classification marks; connecting the input layer to the input of the first convolution layer; taking the output of the first convolution layer as the input of the second convolution layer; connecting the outputs of the second convolution layer into a one-dimensional vector as the input of the fully-connected layer; and connecting the output of the full connection layer with an output layer, and outputting an inversion result to be predicted.
5. The method of claim 4, wherein the convolutional layer is constructed as follows:
ReLU＝max(0,x)
wherein BN (-) is a batch standardization process; ReLU is the activation function, denoted multiply;
is the output result of the convolutional layer; w is a k Is a convolution operator; x is input data; b k Is a bias term; i is the row index of the matrix; j is the column index of the matrix.
6. A pre-stack seismic inversion apparatus, comprising:
the reservoir dividing module is used for dividing reservoirs of the research area according to the prior information and determining an area classification result;
the tag information determining module is used for determining tag information according to the rock physical information, the prior information, the region classification result and a random geological simulation method constrained by the logging information;
the classification marking module is used for performing classification marking on the label information by using the region classification result and determining label data with the classification marking;
the convolutional neural network building module is used for building a convolutional neural network;
the convolutional neural network training module is used for training a convolutional neural network by using the label data with the classification labels and determining the trained convolutional neural network;
and the pre-stack seismic inversion module is used for performing pre-stack seismic inversion according to the trained convolutional neural network and determining an elastic parameter inversion result.
7. The apparatus of claim 6, wherein the reservoir partitioning module is specifically configured to:
dividing reservoirs in a research area according to prior information consisting of geological structure information, sedimentation information, a logging data analysis result, a seismic attribute analysis result and single-well lithofacies information, and determining an area classification result; wherein, the region classification result comprises: the method comprises the following steps of a favorable development area of a first type of reservoir, an intermediate development area of a second type of reservoir and an underdeveloped area of a third type of reservoir.
8. The apparatus of claim 7, wherein the tag information determination module is specifically configured to:
analyzing rock physical information and determining rock core measurement data;
analyzing the core measurement data, performing a rock physical model test by combining logging information, and establishing a rock physical model conforming to a research area;
acquiring geological structure information and deposition information in prior information, integrating geological and geophysical information, and acquiring virtual well curves of different types by using lithofacies probability distribution conditions of different classification areas and combining a random geological simulation method on the basis of region classification results;
modeling by utilizing a rock physical model according to the virtual well curve to determine a simulated elastic parameter;
performing forward modeling of a seismic wave field by using the simulated elastic parameters, and determining a prestack common reflection point gather;
and correspondingly combining the prestack common reflection point gather and the simulation elastic parameters to determine the label data.
9. The apparatus of claim 6, wherein the convolutional neural network building module is specifically configured to:
constructing a first convolution layer, a second convolution layer and a full-connection layer; each convolution layer comprises three parts, namely convolution operation, an activation function and batch standardization;
according to the functional connection sequence, an input layer, a first convolution layer, a second convolution layer, a full connection layer and an output layer are connected in sequence to build a convolution neural network; wherein, the function connection order includes: the input layer receives label data with classification marks; connecting the input layer to the input of the first convolution layer; taking the output of the first convolution layer as the input of the second convolution layer; connecting the output of the second convolution layer into a one-dimensional vector as the input of the full-connected layer; and connecting the output of the full connection layer with an output layer, and outputting an inversion result to be predicted.
10. The apparatus of claim 9, wherein the convolutional neural network building block is further configured to build a convolutional layer as follows:
ReLU＝max(0,x)
wherein BN (-) is a batch standardization process; ReLU is the activation function, denoted multiply;
is the output result of the convolutional layer; w is a k Is a convolution operator; x is input data; b is a mixture of k Is a bias term; i is the row index of the matrix; j is the column index of the matrix.
11. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the method of any of claims 1 to 5 when executing the computer program.
12. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program for executing a method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110294942.5A CN115130529A (en) | 2021-03-19 | 2021-03-19 | Pre-stack seismic inversion method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110294942.5A CN115130529A (en) | 2021-03-19 | 2021-03-19 | Pre-stack seismic inversion method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115130529A true CN115130529A (en) | 2022-09-30 |
Family
ID=83374765
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110294942.5A Pending CN115130529A (en) | 2021-03-19 | 2021-03-19 | Pre-stack seismic inversion method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115130529A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117250658A (en) * | 2023-11-17 | 2023-12-19 | 核工业北京地质研究院 | Method for creating seismic dataset of investigation region |
| WO2024087827A1 (en) * | 2022-10-26 | 2024-05-02 | 中国石油天然气股份有限公司 | Reservoir physical property parameter prediction method and apparatus |
-
2021
- 2021-03-19 CN CN202110294942.5A patent/CN115130529A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2024087827A1 (en) * | 2022-10-26 | 2024-05-02 | 中国石油天然气股份有限公司 | Reservoir physical property parameter prediction method and apparatus |
| CN117250658A (en) * | 2023-11-17 | 2023-12-19 | 核工业北京地质研究院 | Method for creating seismic dataset of investigation region |
| CN117250658B (en) * | 2023-11-17 | 2024-02-09 | 核工业北京地质研究院 | Method for creating seismic dataset of investigation region |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA3122686C (en) | Automated reservoir modeling using deep generative networks | |
| AU2019338412B2 (en) | Machine learning-based analysis of seismic attributes | |
| CN109709603B (en) | Seismic horizon identification and tracking method and system | |
| RU2573746C2 (en) | Systems and methods for well performance forecasting | |
| AU2011283192B2 (en) | Methods and systems for machine-learning based simulation of flow | |
| CN110031896B (en) | Seismic random inversion method and device based on multi-point geostatistics prior information | |
| CN111596978A (en) | Web page display method, module and system for lithofacies classification by artificial intelligence | |
| AU2011283193A1 (en) | Methods and systems for machine-learning based simulation of flow | |
| CA2720117A1 (en) | Systems, methods, and computer program products for modeling dynamic systems by visualizing a parameter space and narrowing the parameter space | |
| AU2011283191A1 (en) | Methods and systems for machine-learning based simulation of flow | |
| US11397279B2 (en) | Comparison of wells using a dissimilarity matrix | |
| CN112965103B (en) | Multi-pore reservoir prestack seismic probability multi-channel inversion method | |
| CN115130529A (en) | Pre-stack seismic inversion method and device | |
| CN116047583A (en) | Adaptive wave impedance inversion method and system based on depth convolution neural network | |
| Han et al. | Multiple point geostatistical simulation with adaptive filter derived from neural network for sedimentary facies classification | |
| EP3948365A1 (en) | Automatic calibration of forward depositional models | |
| US20220245300A1 (en) | Method and system for constructing statistical emulators for reservoir simulation models with uncertain inputs | |
| RU2619803C2 (en) | Determining fluid motion path | |
| CN111751882B (en) | Method and device for generating sensitive rock physical parameters | |
| CN115880455A (en) | Three-dimensional intelligent interpolation method based on deep learning | |
| Marques et al. | Analysis of Deep Learning Neural Networks for Seismic Impedance Inversion: A Benchmark Study. Energies 2022, 15, 7452 | |
| WO2023219936A1 (en) | Adaptive multicolor reordering simulation system | |
| CN117724151A (en) | Construction method of river facies compact heterogeneous reservoir low-frequency model | |
| WO2023154610A1 (en) | Method and apparatus for seismic data inversion | |
| WO2023039367A1 (en) | Method and apparatus for performing wavefield predictions by using wavefront estimations |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |