# Oil and gas reservoir parameter tag data generation method and device

## A technology for oil and gas reservoirs and tag data, applied in the field of oil and gas geophysical exploration, can solve the problems of poor mobility of tag data, low efficiency of tag data generation, and inability to generate tag data of oil and gas reservoir parameters.

Pending Publication Date: 2020-12-29

PETROCHINA CO LTD

0 Cites 1 Cited by

## AI-Extracted Technical Summary

### Problems solved by technology

However, this method will make the generation of label data heavily dependent on the quantity and quality of well logging in the study area. Therefore, it is impossible to generate label data of oil and gas reservoir parameters for the study area with few well logs.

In addition, the transfera...

### Method used

[0080] Further, in a specific embodiment, by adjusting the number of curves generated by Markov chain Monte Carlo random simulation and sequential Gaussian simulation, the scalability of label data can be realized. That is to say, the number of curves actually generated is controllable. According to the artificial intelligence prediction of oil and gas reservoir parameters for samples containing label data, and the characteristics of oil and gas vertical and horizontal changes in the research area, the number of required curves can be expanded. For example, in the research area that changes drastically with depth, increase the number of curves generated by Markov chain Monte Carlo stochastic...

## Abstract

The invention provides an oil and gas reservoir parameter tag data generation method and device. The method comprises the steps that according to logging data of a research area, a prior distributionprobability of lithofacies is obtained by utilizing a Gaussian mixture distribution function; Markov chain Monte Carlo random simulation is performed to generate a plurality of lithofacies curves changing along with depth; sequential Gaussian simulation is performed to generate a plurality of lithofacies curves changing along with transverse direction; according to the plurality of lithofacies curves changing along with depth, the plurality of lithofacies curves changing along with transverse direction and prior characteristics of oil and gas reservoir parameter logging curve values, random filling of all the lithofacies is performed so that a physical property parameter curve of the research area is obtained; lithofacies constraint statistics rock physical modeling is performed, and an elastic parameter curve of the research area is determined; and convolution is performed by utilizing a Zoeppritz reflection equation and seismic wavelets so that oil and gas reservoir parameter tag data are generated. The number and the quality of logging of the research area do not need to be depended on so that the mobility of the generated tag data can be enhanced and thus the generation efficiency can be enhanced.

Application Domain

Design optimisation/simulationConstraint-based CAD +2

Technology Topic

Gaussian mixture distributionConvolution +6

## Image

## Examples

- Experimental program(1)

### Example Embodiment

[0038]The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0039]The embodiment of the present invention provides a method for generating tag data of oil and gas reservoir parameters, which is used to improve the mobility of generated tag data and the generation efficiency of tag data without relying on the quantity and quality of logging in the study area, such asfigure 1 As shown, the method includes:

[0040]Step 101: Using the Gaussian mixture distribution function according to the logging data in the study area, obtain the prior distribution probability of the lithofacies in the study area;

[0041]Step 102: According to the prior distribution probability of lithofacies in the study area, perform Markov chain Monte Carlo random simulation to generate multiple lithofacies curves that vary with depth;

[0042]Step 103: Perform sequential Gaussian simulation according to the generated multiple lithofacies curves that vary with depth to generate multiple lithofacies curves that vary with the lateral direction;

[0043]Step 104: According to the prior characteristics of multiple lithofacies curves varying with depth, multiple lithofacies curves varying with lateral changes and the value of the oil and gas reservoir parameter logging curves in the study area, randomly fill each lithofacies to obtain the study area The physical parameter curve;

[0044]Step 105: Perform lithofacies constrained statistical rock physics modeling according to the physical parameter curve of the study area, and determine the elastic parameter curve of the study area;

[0045]Step 106: According to the elastic parameter curve of the study area, use Zoeppritz reflection equation and seismic wavelet to perform convolution to generate oil and gas reservoir parameter tag data.

[0046]byfigure 1 It can be seen from the process shown that in the embodiment of the present invention, the prior distribution probability of lithofacies in the study area is obtained by using the Gaussian mixture distribution function according to the logging data in the study area; according to the prior distribution of lithofacies in the study area Probability, perform Markov chain Monte Carlo random simulation to generate multiple lithofacies curves that vary with depth; according to the generated multiple lithofacies curves that vary with depth, perform sequential Gaussian simulations to generate multiple rocks that vary with lateral Facies curve: According to the prior characteristics of multiple lithofacies curves varying with depth, multiple lithofacies curves varying with lateral changes and the value of the oil and gas reservoir parameter logging curves in the study area, each lithofacies is randomly filled to obtain the study area According to the physical parameter curve of the study area, perform lithofacies constrained statistical rock physics modeling to determine the elastic parameter curve of the study area; according to the elastic parameter curve of the study area, use Zoeppritz reflection equation and seismic wavelet for convolution , To generate oil and gas reservoir parameter label data; through Markov chain Monte Carlo stochastic simulation to generate multiple lithofacies curves that vary with depth and sequential Gaussian simulation to generate multiple lithofacies curves that vary with the lateral direction, making it in the study area When the number of well logs is small or the quality is poor, sufficient tag data can be generated. The generation of oil and gas reservoir parameter tag data does not need to rely on the number and quality of logging in the study area; by introducing the lithofacies of the study area in the process of tag generation Prior distribution probability, seismic wavelet, and lithofacies constrained statistical petrophysical modeling can obtain oil and gas reservoir parameter tag data that are compatible with the characteristics of the study area; therefore, when changing the study area, you only need to The characteristics and adaptability adjustment can make the generated tag data adapt to different research areas without reacquiring a large amount of logging data and well side channel seismic data, which improves the mobility of the generated tag data, thereby improving the generation of tag data effectiveness.

[0047]In the specific implementation, firstly, according to the logging data in the study area, using the Gaussian mixture distribution function, the prior distribution probability of the lithofacies in the study area is obtained.figure 2 Shown, including:

[0048]Step 201: Use the expected maximum estimation method to estimate the multivariate Gaussian mixture distribution function of the physical parameter logging curve and elastic parameter logging curve in the study area, and obtain the proportional coefficient and elastic parameter corresponding to the Gaussian mixture distribution function corresponding to the physical parameter The proportional coefficient in the Gaussian mixture distribution function of;

[0049]Step 202: According to the proportional coefficient in the Gaussian mixture distribution function corresponding to the physical property parameter, determine the type of lithofacies and the prior distribution probability based on the physical property division in the study area;

[0050]Step 203: According to the proportional coefficient in the Gaussian mixture distribution function corresponding to the elastic parameter, determine the type and prior distribution probability of the lithofacies based on elastic division in the study area;

[0051]Step 204: Obtain the prior distribution probability of the lithofacies in the study area according to the type and prior distribution probability of the lithofacies based on physical property division in the study area and the type and prior distribution probability of the lithofacies based on elastic division in the study area.

[0052]In a specific embodiment, the physical parameters of the study area include parameters such as oil and gas saturation, porosity, and mineral content of the study area, and the elastic parameters include parameters such as longitudinal wave velocity, shear wave velocity, and density. The expected maximum estimation method is used to estimate the multivariate Gaussian mixture distribution function of the physical parameter logging curve and elastic parameter logging curve in the study area, and estimate the number of Gaussian components and the proportional coefficient in the Gaussian mixture distribution function corresponding to the physical parameter, and The number of Gaussian components and the scale coefficient in the Gaussian mixture distribution function corresponding to the elastic parameter.

[0053]According to the number of Gaussian components in the Gaussian mixture distribution function corresponding to the physical property parameters (ie the number of scale coefficients), determine the number of lithofacies in the study area based on physical property classification, and according to the scale coefficient in the Gaussian mixture distribution function corresponding to the physical property parameters The value of is the prior distribution probability of lithofacies based on physical property classification in the study area.

[0054]According to the number of Gaussian components in the Gaussian mixture distribution function corresponding to the elastic parameters (that is, the number of proportional coefficients), determine the number of types of lithofacies based on elastic division in the study area, and the proportional coefficient in the Gaussian mixture distribution function corresponding to the elastic parameters The value of is the prior distribution probability of the lithofacies based on elastic division of the study area.

[0055]The type and prior distribution probability of lithofacies based on physical property division in the comprehensive study area and the type and prior distribution probability of lithofacies based on elastic division in the study area. The prior distribution probability corresponding to the largest number of lithofacies is selected as The prior distribution probability of lithofacies in the study area. When the types of the two lithofacies are equal, the mean value of the prior distribution probability of the two lithofacies is taken as the prior distribution probability of the lithofacies in the study area.

[0056]After obtaining the prior distribution probability of the lithofacies in the study area, according to the prior distribution probability of the lithofacies in the study area, a Markov chain Monte Carlo stochastic simulation is performed to generate multiple lithofacies curves varying with depth. In specific implementation, it is assumed that the probability of belonging to a certain type of lithofacies at time t is only related to the probability of belonging to the lithofacies type at the previous time t-1, and changes in the depth direction are regarded as transfers between different lithofacies types. Define the Markov transition probability matrix of lithofacies downward as:

[0057]

[0058]Among them, the matrix element pi,j Represents the conditional probability of changing from lithofacies category i to lithofacies category j:

[0059]pi,j =P(Fj|Fi)

[0060]=P(Fi|Fj)P(Fj)/P(Fi)

[0061]In a specific embodiment, the above-mentioned Markov transition probability matrix can be obtained by statistics of logging samples in the study area, and the prior distribution probability P(Ft), using the Markov transition probability matrix obtained, we can get the prior distribution probability P(Ft+1):

[0062]P(Ft+1)=P(Ft)PT

[0063]The prior distribution probability of the lithofacies at each moment is obtained, and by using the Monte Carlo stochastic simulation method, multiple lithofacies curves varying with depth can be generated.

[0064]After generating multiple lithofacies curves that vary with depth, perform sequential Gaussian simulations based on the multiple lithofacies curves that vary with depth, and generate multiple lithofacies curves that vary with the lateral direction. In specific implementation, each lithofacies curve that varies with depth is used as a seed sample, and sequential Gaussian simulations are performed using the variogram to generate multiple lithofacies that vary with the lateral direction; among them, the variogram is based on the study area The lithofacies samples at the log are obtained by elliptic function fitting.

[0065]After generating multiple lithofacies curves that vary with the lateral direction, according to the multiple lithofacies curves that vary with depth, multiple lithofacies curves that vary with the lateral direction, and the prior characteristics of the value of the oil and gas reservoir parameter logging curves in the study area, Randomly fill in each lithofacies to obtain the physical parameter curve of the study area. The specific implementation process, such asimage 3 Shown, including:

[0066]Step 301: Construct a Gaussian mixture distribution function according to the prior characteristics of the oil and gas reservoir parameter logging curves in the study area, and randomly generate multiple reservoir parameters that meet the prior distribution characteristics of the oil and gas reservoir parameters of each lithofacies interval ;

[0067]Step 302: According to the above-mentioned multiple reservoir parameters, multiple lithofacies curves varying with depth, and multiple lithofacies curves varying with lateral direction, obtain physical parameter curves of the study area.

[0068]Among them, the prior features of the logging curve of oil and gas reservoir parameters in the study area include the maximum, minimum, mean and variance characteristics of the oil and gas reservoir parameters in each lithofacies in the study area. The physical property parameter curve may be, for example, a porosity curve, a water saturation curve, and a mud content curve.

[0069]After obtaining the physical parameter curve of the study area, according to the physical parameter curve of the study area, perform lithofacies constraint statistical petrophysical modeling to determine the elastic parameter curve of the study area. In the specific implementation, the relationship between the physical parameters and the elastic parameters in different lithofacies is different. The lithofacies constrained statistical rock physics modeling uses the following formula to determine the elastic parameter curve in the study area:

[0070]E=f(R,F)+χ

[0071]Among them, E represents elastic parameters, R represents physical property parameters, and F represents lithofacies;

[0072]f(·) represents the petrophysical model, which is obtained from petrophysical experiments or historical experience in the study area;

[0073]χ represents the statistical error between the petrophysical model and the actual value, which obeys the Gaussian cutoff distribution, and is statistically obtained from the error distribution characteristics between the actual logging curve and the simulated logging curve.

[0074]In a specific embodiment, the elastic parameter curve may be, for example, a longitudinal wave velocity curve, a shear wave velocity curve, and a density curve.

[0075]After determining the elastic parameter curve of the study area, according to the elastic parameter curve of the study area, the Zoeppritz reflection equation and seismic wavelet are used for convolution to generate oil and gas reservoir parameter label data. The specific implementation process, such asFigure 4 Shown, including:

[0076]Step 401: According to the elastic parameter curve of the study area, use Zoeppritz reflection equation and seismic wavelet to convolve, introduce random noise of different intensities, and synthesize pre-stack seismic angle gather; among them, pre-stack seismic angle gather contains multiple different Samples of pre-stack gathers whose amplitude varies with incident angle under noise intensity;

[0077]Step 402: Extract the lithofacies, physical property parameters, and elastic parameters of the above-mentioned samples as oil and gas reservoir parameter tag data.

[0078]Among them, the main frequency and length of the seismic wavelet used need to be determined according to the main frequency of the seismic data in the study area and the thickness of the target layer. Random noise is Gaussian noise, and it is necessary to use Gaussian function to generate multiple random noises of different intensities to simulate seismic data with different signal-to-noise ratios. The pre-stack seismic angle gather is a gather whose amplitude varies with angle, and the number of angles needs to be determined according to the angle range of the pre-stack seismic data in the study area.

[0079]The synthesized pre-stack seismic angle gather contains samples of pre-stack gathers whose amplitude varies with the angle of incidence under different noise intensities. The lithofacies, physical parameters, and elastic parameters corresponding to each sample can be used as the oil and gas reservoir parameter labels in the study area. data.

[0080]Further, in a specific embodiment, by adjusting the number of curves generated by the Markov chain Monte Carlo stochastic simulation and the sequential Gaussian simulation, the scalability of the label data can be achieved. That is, the number of curves actually generated is controllable. According to the artificial intelligence prediction of oil and gas reservoir parameters, the demand for samples containing label data, and the characteristics of vertical and horizontal changes in oil and gas in the study area, the number of curves required can be expanded. For example, in the study area where the depth changes drastically, increase the number of curves generated by the Markov chain Monte Carlo random simulation; in the study area where the horizontal change drastically changes, increase the number of curves generated by the sequential Gaussian simulation.

[0081]In a specific embodiment, the value characteristics of the logging curve of the oil and gas reservoir parameters, the characteristics of the seismic wavelet, the lithofacies constraint statistical petrophysical modeling, and the angle of the synthesized pre-stack seismic gather are adjusted according to the characteristics of different research areas. The number can enable the generated tag data to be migrated to adapt to different research areas and realize the migration of tag data. Specifically, the value characteristics of the logging curve of oil and gas reservoir parameters specifically refer to the estimation of the multivariate Gaussian mixture distribution function of the physical parameter logging curve and the elastic parameter logging curve of the actual study area, and the estimated Gaussian mixture distribution function Gaussian part number, mean value, variance, scale factor and other data. The characteristics of the seismic wavelet include the main frequency and length of the seismic wavelet, which are determined according to the main frequency of the seismic data in the actual study area and the thickness of the target layer. Lithofacies constrained statistical petrophysical modeling specifically refers to a petrophysical model determined by constrained statistics based on the lithofacies in the actual study area. The number of angles of the synthesized pre-stack seismic gather is determined according to the angle range of the pre-stack seismic data in the actual study area.

[0082]A specific example is given below to illustrate how the embodiment of the present invention generates oil and gas reservoir parameter tag data. This example applies to a specific research area.

[0083]This specific example provides a device for making oil and gas reservoir parameter label data, the structure is asFigure 5 Shown, including:

[0084]The prior distribution calculation module 501 is used to obtain the prior distribution probability of the lithofacies in the study area by using the Gaussian mixture distribution function;

[0085]Monte Carlo simulation module 502 is used to generate a large number of lithofacies curves varying with depth by using Markov chain Monte Carlo stochastic simulation based on the prior distribution probability of lithofacies in the study area;

[0086]Sequential Gaussian simulation module 503 is used to generate multiple laterally varying lithofacies curves using sequential Gaussian simulation according to the lithofacies curves varying with depth;

[0087]The physical filling module 504 is used to randomly fill each lithofacies to obtain the physical parameter curve of the study area according to the lithofacies curve that varies with depth and the lithofacies curve with lateral variation, combined with the prior characteristics of reservoir parameter logging curves. ;

[0088]The statistical petrophysical modeling module 505 is used to use lithofacies constrained statistical petrophysical modeling according to the physical parameter curve of the study area to convert the physical parameter curve into the elastic parameter curve of the study area;

[0089]The pre-stack seismic angle gathers are assembled into a module 506, which is used to convolve the Zoeppritz reflection equation with seismic wavelets according to the elastic parameter curve of the study area, introduce random noise of different intensities, and synthesize the corresponding pre-stack seismic angle gathers;

[0090]The training sample generation module 507 is used to use the synthetic pre-stack seismic angle gather as the observation sample set for artificial intelligence prediction of oil and gas reservoirs, and the lithofacies, physical parameters, and elastic parameters corresponding to each sample in the sample set are used as label data;

[0091]The tag data expansion module 508 is used to expand the number of lithofacies curves that vary with depth and the number of lithofacies curves that vary with lateral changes;

[0092]The tag data migration module 509 is used to change the prior characteristics of reservoir parameter logging curve values, the characteristics of seismic wavelets, and to select appropriate petrophysical models based on lithofacies constraints, and to migrate tag data to adapt to different research areas .

[0093]According to the above-mentioned oil and gas reservoir parameter label data production device, the production of oil and gas reservoir parameter label data is carried out. The process flow of the production method specifically includes the following steps:

[0094]Step S1: Use the expected maximum estimation method to estimate the multivariate Gaussian mixture distribution function of the elastic parameter logging curve in the study area, and estimate the number of Gaussian components in the Gaussian mixture distribution function, the mean value, the variance, and the proportional coefficient. According to the estimated proportional coefficient The number and value of, determine the type and prior probability of the lithofacies divided by elastic properties.

[0095]Figure 6 Shown are the estimation results of Gaussian mixture distribution function using elastic parameters such as longitudinal wave velocity, transverse wave velocity, density, etc. in a specific example of the present invention. It can be seen that the distribution form of the above elastic parameters is decomposed into 3 Gaussian functions, that is, the number of parts in the Gaussian mixture distribution is 3, which represents that the lithofacies in this area is divided into 3 lithofacies, and 3 Gaussian functions (solid line, Dotted and dashed lines) are superimposed together to form the prior distribution of elastic parameters, and the superposition ratios are 0.46, 0.28, and 0.26, respectively, representing the prior probability of each lithofacies. Therefore, in the specific example of the present invention, combined with the actual lithology and fluid information in the study area, it can be determined that the three types of lithofacies are divided into mudstone facies (solid line), water-bearing sandstone facies (dotted line), and gas-bearing facies based on elastic parameters. For sandstone facies (dashed lines), the corresponding prior probabilities are 0.46, 0.28, and 0.26, respectively.

[0096]Step S2: Use the expected maximum estimation method to estimate the multivariate Gaussian mixture distribution function of the physical parameter logging curve in the study area, and estimate the number of Gaussian components in the Gaussian mixture distribution function, the mean value, the variance, and the proportional coefficient. According to the estimated proportional coefficient The number and value of, determine the type and prior probability of lithofacies based on physical property classification.

[0097]Figure 7 Shown are the estimation results of Gaussian mixture distribution function using physical parameters such as mud content, porosity, and gas saturation in the embodiment of the present invention. It can be seen that the distribution of the above physical parameters is decomposed into 3 Gaussian functions, that is, the number of components in the Gaussian mixture distribution is 3, which means that the lithofacies in this area is divided into 3 lithofacies, and 3 Gaussian functions (solid line, Dotted and dashed lines) are superimposed together to form the prior distribution of elastic parameters, and the superimposition ratios are 0.42, 0.29, and 0.29, respectively, representing the prior probability of each lithofacies. Therefore, in the embodiment of the present invention, combined with the actual lithology and fluid information in the study area, it can be determined that the three lithofacies are classified based on physical parameters as mudstone facies (solid line), water-bearing sandstone facies (dotted line), and gas-bearing facies. For sandstone facies (dashed lines), the corresponding prior probabilities are 0.42, 0.29, and 0.29, respectively.

[0098]Step S3: Synthesize the lithofacies types and prior probabilities based on physical properties and based on elasticity division, and select the division result corresponding to the largest number of lithofacies types as the final lithofacies division and lithofacies prior probability result. If the number of lithofacies divided by the two is equal, the mean value of the prior probability of the two lithofacies is taken as the final prior probability of lithofacies.

[0099]Since there are three types of lithofacies based on physical properties and based on elasticity classification, the lithofacies classification types in this specific example are three, which are mudstone facies, water-bearing sandstone facies, and gas-bearing sandstone facies. The corresponding prior probability is 0.44. , 0.285, 0.275.

[0100]Step S4: According to the prior distribution probability of the lithofacies in the study area in step S3, use Markov chain Monte Carlo stochastic simulation to generate a large number of lithofacies curves varying with depth, where the lithofacies in the Markov chain transfer with depth The probability matrix is obtained by statistics of the existing logging comprehensive interpretation curve samples.

[0101]Such asFigure 8 Shown are five adjacent lithofacies curves varying with depth in a specific example of the present invention.

[0102]Step S5: Using each generated lithofacies curve varying with depth as a seed sample, use sequential Gaussian simulation technology to generate multiple lateral varying lithofacies curves. Among them, the variogram used in the sequential Gaussian simulation of lateral changes is determined by the research The lithofacies samples at known well points in the area are obtained by elliptic function fitting.

[0103]Picture 9 Shown are five adjacent lithofacies curves that vary with the lateral direction in a specific example of the present invention.

[0104]Step S6: According to the lithofacies curves generated in steps S4 and S5, combined with the prior characteristics of the value of the reservoir parameter logging curve, randomly fill each lithofacies to obtain a physical property parameter curve. Among them, the prior characteristics of reservoir parameter logging values include the maximum, minimum, average, and variance characteristics of reservoir parameters in each lithofacies. According to the above characteristics, a Gaussian mixture distribution function is constructed to randomly generate a series of reservoir parameters that meet the prior distribution characteristics of the reservoir parameters of each lithofacies interval.

[0105]Picture 10 Shown is a lithofacies curve and a physical parameter curve after filling in a specific example of the present invention.

[0106]Step S7: According to the physical parameter curve generated in step S6, adopt lithofacies constraint statistical petrophysical modeling to convert the reservoir parameter curve into the corresponding elastic parameter curve; the following formula is used for modeling:

[0107]E=f(R,F)+χ

[0108]Among them, E represents elastic parameters, R represents physical property parameters, and F represents lithofacies;

[0109]f(·) represents the petrophysical model, which is obtained from petrophysical experiments or empirical relationships in the study area;

[0110]χ represents the statistical error between the petrophysical model and the actual value, which obeys the Gaussian cutoff distribution, and is statistically obtained from the error distribution characteristics between the actual logging curve and the simulated logging curve.

[0111]Picture 11 Shown are specific examples of the present invention, according toPicture 10 The physical parameters in the elastic parameter curve generated by the lithofacies constraint statistical rock physics modeling.

[0112]Step S8: Given the dominant frequency and length of the seismic wavelet, generate the seismic wavelet. In the specific example of the present invention, it is generated according to the Lake wavelet formula:

[0113]

[0114]Among them, w(t) represents the Lake wavelet; f0Indicates the main frequency, in hertz; t is the length of time, in seconds.

[0115]In this step, other types of seismic wavelets can also be selected, and the specific example of the present invention is not limited here.

[0116]Step S9: Use the elastic parameters generated in step S7 to calculate the reflection coefficient according to the Zoeppritz equation, and perform convolution operation with the seismic wavelet in step S8 to synthesize a pre-stack seismic angle gather.

[0117]Picture 12 It is the pre-stack seismic angle gather synthesized in the specific example of the present invention.

[0118]Step S10: Generate Gaussian noise of different intensities and add it to the pre-stack seismic angle gather synthesized in step S9 to simulate pre-stack seismic angle gathers with different signal-to-noise ratios.

[0119]Figure 13 Shown asPicture 12 The mid-pre-stack seismic angle gather is added to the noisy pre-stack seismic angle gather with signal-to-noise ratios of 10, 6, and 4 respectively.

[0120]Step S11: Use the noisy pre-stack seismic angle gathers synthesized in step S10 as training samples, the lithofacies curves generated in steps S4 and S5, the physical parameter curves generated in step S6, and the elastic parameter curves generated in step S7 as label data, Provide deep network training and learning data for intelligent prediction of reservoir parameters.

[0121]Step S12: Adjust the number of curves generated by Markov chain Monte Carlo stochastic simulation and sequential Gaussian simulation according to the requirements of step S11 for the number of training samples and label data sets, so as to realize the scalability of label data. In the study area where the depth changes drastically, increase the number of curves generated by the Markov chain Monte Carlo random simulation; in the study area where the horizontal change drastically changes, increase the number of curves generated by the sequential Gaussian simulation.

[0122]In the specific example of the present invention, 50 lithofacies curves are generated with the change of depth, and each lithofacies curve generates 50 curves with lateral changes, and a total of 2500 lithofacies curves are simulated.

[0123]Step S13: According to the requirements of the mobility of the training sample and the label data set in step S11, the elastic parameters and physical parameter logging curves in steps S1 and S2 are replaced, the petrophysical model in step S7 is changed, and the dominant frequency of the wavelet in step S8 is adjusted The time length and the angle range of the pre-stack seismic angle gather in step S9 make the generated training sample and label data set conform to the target research area.

[0124]In the specific example of the present invention, the petrophysical model used is the KT model, the main frequency of the seismic wavelet is 35 Hz, the duration is 0.064 seconds, and the angle range of the pre-stack seismic angle gather is 0-32°.

[0125]The above specific examples illustrate that the provided oil and gas reservoir parameter tag data production device and method have flexible scalability and mobility.

[0126]The implementation of the above specific application is only an example, and the rest of the implementation manners will not be repeated one by one.

[0127]Based on the same inventive concept, the embodiments of the present invention also provide an oil and gas reservoir parameter tag data generation device. Since the problem solved by the oil and gas reservoir parameter tag data generation device is similar to the oil and gas reservoir parameter tag data generation method, the oil and gas reservoir For the implementation of the layer parameter label data generation device, please refer to the implementation of the oil and gas reservoir parameter label data generation method. The repetition will not be repeated. The specific structure is asFigure 14 Shown:

[0128]The prior distribution calculation module 1401 is used to obtain the prior distribution probability of the lithofacies in the study area by using the Gaussian mixture distribution function according to the logging data in the study area;

[0129]The Monte Carlo simulation module 1402 is used to perform Markov chain Monte Carlo random simulation to generate multiple lithofacies curves varying with depth according to the prior distribution probability of the lithofacies in the study area;

[0130]Sequential Gaussian simulation module 1403, used to perform sequential Gaussian simulation based on the generated multiple lithofacies curves that vary with depth, and generate multiple lithofacies curves that vary with the lateral;

[0131]The physical property filling module 1404 is used to randomly fill each lithofacies according to the prior characteristics of multiple lithofacies varying with depth, multiple lithofacies varying with lateral and the logging curves of oil and gas reservoir parameters in the study area , Get the physical parameter curve of the study area;

[0132]The statistical petrophysical modeling module 1405 is used to perform lithofacies constrained statistical petrophysical modeling according to the physical parameter curve of the study area, and determine the elastic parameter curve of the study area;

[0133]The oil and gas reservoir parameter tag data generation module 1406 is used to generate oil and gas reservoir parameter tag data by convolution using the Zoeppritz reflection equation and seismic wavelet according to the elastic parameter curve of the study area.

[0134]In a specific embodiment, the prior distribution calculation module 1401, such asFigure 15 Shown, including:

[0135]The proportional coefficient determining unit 1501 is used to estimate the physical parameter log curve and elastic parameter log curve of the study area by using the expected maximum estimation method to estimate the multivariate Gaussian mixture distribution function to obtain the proportion in the Gaussian mixture distribution function corresponding to the physical parameter The coefficient and the proportional coefficient in the Gaussian mixture distribution function corresponding to the elastic parameter;

[0136]The physical property division unit 1502 is used to determine the type and prior distribution probability of lithofacies based on physical property division in the study area according to the proportional coefficient in the Gaussian mixture distribution function corresponding to the physical property parameter;

[0137]The elastic division unit 1503 is used to determine the type and prior distribution probability of the lithofacies based on elastic division in the study area according to the proportional coefficient in the Gaussian mixture distribution function corresponding to the elastic parameter;

[0138]The priori distribution determination unit 1504 is used to obtain the lithofacies of the study area according to the type and prior distribution probability of the lithofacies based on the physical property division of the study area and the type and prior distribution probability of the lithofacies based on the elastic division of the study area The probability of the prior distribution.

[0139]In a specific embodiment, the sequential Gaussian simulation module 1403 is specifically used for:

[0140]Take each lithofacies curve that varies with depth as a seed sample, and use the variogram to perform sequential Gaussian simulations to generate multiple lithofacies curves that vary with the lateral direction. Among them, the variogram is based on the logging location in the study area. The lithofacies sample is obtained by elliptic function fitting.

[0141]In specific implementation, the physical property filling module 1404, such asFigure 16 Shown, including:

[0142]The reservoir parameter determination unit 1601 is used to construct the Gaussian mixture distribution function according to the prior characteristics of the oil and gas reservoir parameter logging curves in the study area, and randomly generate the prior distribution characteristics of the oil and gas reservoir parameters in accordance with each lithofacies interval Of multiple reservoir parameters;

[0143]The physical parameter curve determining unit 1602 is used to obtain the physical parameter curve of the study area according to the above-mentioned multiple reservoir parameters, multiple lithofacies curves varying with depth, and multiple lithofacies curves varying with lateral direction.

[0144]In specific embodiments, the statistical petrophysical modeling module 1405 is specifically used for:

[0145]According to the following formula, according to the physical parameter curve of the study area, perform lithofacies constrained statistical rock physics modeling to determine the elastic parameter curve of the study area:

[0146]E=f(R,F)+χ

[0147]Among them, E represents elastic parameters, R represents physical property parameters, and F represents lithofacies;

[0148]f(·) represents the petrophysical model, which is obtained from petrophysical experiments or empirical relationships in the study area;

[0149]χ represents the statistical error between the petrophysical model and the actual value, which obeys the Gaussian cutoff distribution, and is statistically obtained from the error distribution characteristics between the actual logging curve and the simulated logging curve.

[0150]In a specific embodiment, the oil and gas reservoir parameter tag data generation module 1406 is specifically used for:

[0151]According to the elastic parameter curve of the study area, the Zoeppritz reflection equation and seismic wavelet are used to convolve, and random noises of different intensities are introduced to synthesize the pre-stack seismic angle gather; among them, the pre-stack seismic angle gather contains multiple noise intensities. Samples of pre-stack gathers whose amplitude varies with incident angle;

[0152]Extract the lithofacies, physical parameters and elastic parameters of multiple samples as oil and gas reservoir parameter tag data.

[0153]An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the above-mentioned oil and gas reservoir parameter tag data when the computer program is executed. Generation method.

[0154]The embodiment of the present invention also provides a computer-readable storage medium storing a computer program for executing the above-mentioned method for generating oil and gas reservoir parameter tag data.

[0155]In summary, the method and device for generating oil and gas reservoir parameter tag data provided by the embodiments of the present invention have the following advantages:

[0156]According to the log data of the study area and the Gaussian mixture distribution function, the prior distribution probability of the lithofacies in the study area is obtained; according to the prior distribution probability of the lithofacies in the study area, a Markov chain Monte Carlo random simulation is performed. Multiple lithofacies curves varying with depth; according to the generated multiple lithofacies curves varying with depth, a sequential Gaussian simulation is performed to generate multiple lithofacies curves varying with the depth; according to multiple lithofacies curves varying with the depth , A number of laterally varying lithofacies curves and the priori characteristics of the value of the oil and gas reservoir parameter logging curves in the study area are randomly filled in each lithofacies to obtain the physical parameter curves of the study area; according to the physical parameter curves of the study area, Carry out lithofacies constrained statistical rock physics modeling to determine the elastic parameter curve of the study area; according to the elastic parameter curve of the study area, use Zoeppritz reflection equation and seismic wavelet to convolve to generate oil and gas reservoir parameter label data; through Markov Chain Monte Carlo stochastic simulation generates multiple lithofacies curves that vary with depth and sequential Gaussian simulation generates multiple lithofacies curves that vary laterally, so that even when the number of well logs in the study area is small or poor quality Sufficient label data, oil and gas reservoir parameter label data generation does not need to rely on the quantity and quality of logging in the study area; by introducing the prior distribution probability of lithofacies, seismic wavelets, and lithofacies constraints in the study area during the label generation process Statistical petrophysical modeling can obtain oil and gas reservoir parameter tag data that is compatible with the characteristics of the study area; therefore, when changing the study area, only need to adjust the characteristics of the study area to adapt the generated tag data In different research areas, there is no need to re-acquire a large amount of logging data and well-side channel seismic data, which improves the mobility of the generated tag data, thereby increasing the efficiency of tag data generation.

[0157]Those skilled in the art should understand that the embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may be in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0158]The present invention is described with reference to flowcharts and/or block diagrams of methods, devices and computer program products according to embodiments of the present invention. It should be understood that each process and/or block in the flowchart and/or block diagram, and the combination of processes and/or blocks in the flowchart and/or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing equipment to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing equipment can be generated In the processFigure oneProcess or multiple processes and/or boxesFigure oneA device with functions specified in a block or multiple blocks.

[0159]These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing equipment to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including the instruction device. The device is implemented in the processFigure oneProcess or multiple processes and/or boxesFigure oneFunctions specified in a box or multiple boxes.

[0160]These computer program instructions can also be loaded on a computer or other programmable data processing equipment, so that a series of operation steps are executed on the computer or other programmable equipment to produce computer-implemented processing, so as to execute on the computer or other programmable equipment. Instructions are provided to implement the processFigure oneProcess or multiple processes and/or boxesFigure oneSteps of functions specified in a box or multiple boxes.

[0161] The foregoing descriptions are only preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc., made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

## PUM

## Description & Claims & Application Information

We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.