CN-122020147-A - Artificial intelligence virtual well sample expansion method based on geostatistics

CN122020147ACN 122020147 ACN122020147 ACN 122020147ACN-122020147-A

Abstract

The invention provides an artificial intelligent virtual well sample expansion method based on geostatistics, which comprises the steps of carrying out statistical analysis according to logging data to obtain priori information, sequentially indicating a simulation algorithm to simulate lithology conforming to the priori information, checking statistical parameters of simulated lithofacies and dividing the lithofacies of a virtual well, carrying out sequential Gaussian simulation on the lithofacies, sequentially carrying out Gao Sixie simulation algorithm to simulate physical property and elastic parameters, combining each well to obtain lithofacies, physical property and elastic parameters of a complete well, and checking the physical property and the elastic parameters of the simulated lithofacies by utilizing statistics and a variation function. The accuracy of the simulation data is verified through statistical analysis, the consistency of the simulation data and the actual well data is ensured, and the credibility of the virtual well sample is enhanced.

Inventors

CHEN YUMAO
XIA ZHIWEI
YAN FEI
ZHANG YUXIAO
ZHANG JUN
GAO SHOUTAO
ZHANG WEI
MENG XIANXIA

Assignees

中国石油化工股份有限公司
中国石油化工股份有限公司胜利油田分公司

Dates

Publication Date: 20260512
Application Date: 20241112

Claims (11)

1. An artificial intelligence virtual well sample expansion method based on geostatistics, which is characterized by comprising the following steps: carrying out statistical analysis according to logging data to obtain priori information; Sequentially indicating a simulation algorithm to simulate lithology conforming to prior information; Checking statistical parameters of the simulated lithofacies and dividing the lithofacies of the virtual well; carrying out sequential Gaussian simulation and sequential Gao Sixie simulation algorithm on the rock-separating facies to simulate physical properties and elastic parameters; Combining each well to obtain lithology, physical properties and elasticity parameters of the whole well; Physical properties and elastic parameters of the simulated lithofacies are checked by using statistics and variation functions.
2. The geostatistical-based artificial intelligence virtual well sample expansion method according to claim 1, wherein the step of performing statistical analysis based on logging data to obtain prior information comprises: Dividing each piece of original well data according to the lithology type; statistical parameters required for the spatial model of each variable in each facies are analyzed.
3. The geostatistical based artificial intelligence virtual well sample expansion method according to claim 1, wherein the classifying each piece of raw well data according to lithology type specifically comprises: for the non-stationarity of the logging data, each piece of raw well data is partitioned by facies type such that statistical data within the same facies is considered stationary.
4. The method for expanding artificial intelligence virtual well samples according to claim 1, wherein the statistical parameters required by the spatial model of each variable in each lithology specifically comprise a histogram, a spatial continuity model and correlations among variables.
5. The geostatistical based artificial intelligence virtual well sample expansion method according to claim 1, wherein the statistical parameters required for analyzing the spatial model of each variable in each lithology specifically comprise: the variation function is used for evaluating the variation condition of the data under various distances and measuring the spatial correlation between the data; The formula of the variation function mentioned in the description is as follows: Gamma (h) is the variance function value for the lag distance h, N (h) is the number of data pairs at the distance h, z (mu α ),z(μ α +h), is the spatial sample value for the head and tail of the lag distance h, respectively.
6. The geostatistical based artificial intelligence virtual well sample expansion method of claim 1, wherein the sequential instruction simulation algorithm simulates lithology conforming to a priori information specifically comprises: calculating a probability mass function of the lithology; Calculating lithofacies probabilities based on an indicated kriging method; based on the lithofacies probabilities, sampling generates a lithofacies sequence.
7. The geostatistical based artificial intelligence virtual well sample expansion method of claim 6, wherein said calculating a probability mass function of a lithology specifically comprises: Calculating probability mass functions of each lithofacies according to prior information and statistical analysis; there are N facies, each facies being F 1 ,F 2 ,……,F N respectively; for a certain lithology F i , its probability mass function is expressed as:
8. The geostatistical based artificial intelligence virtual well sample expansion method of claim 6, wherein said calculating lithofacies probabilities based on the indicated kriging method specifically comprises: at the simulated position P, computing a lithofacies probability for the given position based on lithofacies values of previous simulated positions; Assuming that the previous simulated position around P is P 1 ,P 2 ,……,P m , the corresponding lithofacies value is The indication kriging method is used to calculate the probability of lithofacies F i at a given location P: Wherein, the Is an indication function when 1 When the time is equal to or 0 when the time is equal to or less than the time; Sampling to generate a lithofacies sequence based on the calculated lithofacies probability; At the simulation position P, the lithofacies with the highest probability are selected as simulation results.
9. The geostatistical based artificial intelligence virtual well sample expansion method of claim 1, wherein said examining statistical parameters of simulated lithofacies and partitioning the lithofacies of a virtual well specifically comprises: Carrying out statistical analysis on the simulated virtual well data, and checking whether the lithofacies distribution proportion of the actual well data is met; Calculating a virtual well transition matrix, the probability of transitioning from one given facies to another facies; And dividing the virtual lithofacies when the simulated virtual lithofacies conform to the distribution of the real lithofacies.
10. The geostatistical-based artificial intelligence virtual well sample expansion method according to claim 1, wherein the sequential gaussian simulation, sequential Gao Sixie simulation algorithm simulation physical properties and elastic parameters of the lithology-separated phases specifically comprise: Normal transformation physical parameters: Carrying out normal transformation on physical parameters to ensure that data obeys normal distribution; Assuming that the original physical parameter is X, obtaining a transformed parameter Y through normal transformation, and representing the transformed parameter Y as follows: Wherein μ X and σ X are the mean and standard deviation of the original physical parameters; Modeling the variation function of the physical distribution and the petrophysical properties: For each lithofacies F i , obtaining physical property distribution according to a corresponding logging curve; the physical property corresponding to the logging curve is assumed to be The physical properties were simulated using gaussian distribution, expressed as: In modeling the variation function of petrophysical properties, a similar gaussian function is used, taking into account the spatial correlation of geological properties: γY(h)=C Y ·e -aY·h Sequential gaussian simulated filled reservoir physical properties: Sequential gaussian modeling is used to fill the physical property values obtained by the modeling into each of the rock phases; For each simulation position P, randomly sampling from the physical distribution to obtain transformed physical values Obtaining physical property parameters at the simulation position P through inverse normal transformation Physical Properties of reservoir Filled into the corresponding lithofacies F i .
11. The geostatistical based artificial intelligence virtual well sample expansion method of claim 1, wherein the method for generating the elastic parameters comprises: generating elastic properties based on a petrophysical model according to reservoir properties, and generating geophysical constraint data for marker training; based on the statistical data, the elastic properties are directly simulated within a predefined lithofacies range.

Description

Artificial intelligence virtual well sample expansion method based on geostatistics Technical Field The invention relates to the field of oil and gas exploration, in particular to an artificial intelligence virtual well sample expansion method based on geostatistics. Background In recent years, artificial intelligence has shown great potential in applications in the field of seismic exploration. Hampson et al in 2001 quantitatively predicted elastic and rock properties from seismic data using neural networks. The supervised learning technique derives statistical relationships between well position log data and seismic data. This relationship is then applied to the seismic data to estimate the logging properties at other locations in the seismic survey. However, a limiting factor in performing such analysis is that sufficient marking data (i.e., well control) is required to train and verify such relationships. However, in seismic exploration, well logging data are very small due to high well drilling cost, but work area data are very huge. Many different approaches have been taken by some students to increase the amount of training data. Wang et al (2019) perform a series of transformations on the tag data to increase the amount of data available. Although students have studied the idea of supplementing synthetic data to increase the training set size in the past (Wang et al, 2019), there is no guarantee that pseudo logs generated using these techniques follow the same petrophysical laws as the original logs. Disclosure of Invention In view of the above, the present invention has been made to provide a geostatistical-based artificial intelligence virtual well sample expansion method that overcomes or at least partially solves the above-described problems. According to one aspect of the present invention, there is provided a geostatistical-based artificial intelligence virtual well sample expansion method, the sample expansion method comprising: carrying out statistical analysis according to logging data to obtain priori information; Sequentially indicating a simulation algorithm to simulate lithology conforming to prior information; Checking statistical parameters of the simulated lithofacies and dividing the lithofacies of the virtual well; carrying out sequential Gaussian simulation and sequential Gao Sixie simulation algorithm on the rock-separating facies to simulate physical properties and elastic parameters; Combining each well to obtain lithology, physical properties and elasticity parameters of the whole well; Physical properties and elastic parameters of the simulated lithofacies are checked by using statistics and variation functions. Optionally, the obtaining prior information by performing statistical analysis according to logging data specifically includes: Dividing each piece of original well data according to the lithology type; statistical parameters required for the spatial model of each variable in each facies are analyzed. Optionally, the dividing each piece of original well data according to the lithology type specifically includes: for the non-stationarity of the logging data, each piece of raw well data is partitioned by facies type such that statistical data within the same facies is considered stationary. Optionally, the statistical parameters required by the spatial model of each variable in each lithology include, in particular, histograms, spatial continuity models and correlations between variables. Optionally, the statistical parameters required for analyzing the spatial model of each variable in each lithology specifically include: the variation function is used for evaluating the variation condition of the data under various distances and measuring the spatial correlation between the data; The formula of the variation function mentioned in the description is as follows: Gamma (h is the variance function value for the lag distance h, N (h) is the number of data pairs at the distance h, z (mu α),z(μα +h) is the spatial sample value for the head and tail of the lag distance h, respectively. Optionally, the sequential indication simulation algorithm simulates lithology conforming to the prior information specifically includes: calculating a probability mass function of the lithology; Calculating lithofacies probabilities based on an indicated kriging method; based on the lithofacies probabilities, sampling generates a lithofacies sequence. Optionally, the calculating the probability mass function of the lithology specifically includes: Calculating probability mass functions of each lithofacies according to prior information and statistical analysis; there are N facies, each facies being F 1,F2,……,FN respectively; for a certain lithology F i, its probability mass function is expressed as: Optionally, the calculating the lithofacies probability based on the indication kriging method specifically includes: At a simulated position v, computing a lithofacies probability for a given posi