Seismic quantitative prediction method for shale TOC based on sensitive parameter volumes

Patent metadata

Field	Value
Publication number	US-12487372-B2
Application number	US-202318341781-A
Country	US
Kind code	B2
Filing date	Jun 27, 2023
Priority date	Sep 19, 2022
Publication date	Dec 2, 2025
Grant date	Dec 2, 2025

Abstract

A seismic quantitative prediction method for shale total organic carbon (TOC) based on sensitive parameter volumes is as follows. A target stratum for a TOC content to be measured is determined, logging curves with high correlations with TOC contents are analyzed, the logging curves are found as sensitive parameters; sample data are constructed using the sensitive parameters; a radial basis function (RBF) neural network is trained with the sample data as an input and the TOC content at a depth corresponding to the sample data as an output to obtain a RBF neural network prediction model; sensitive parameter volumes are obtained by using the sensitive parameters and post stack three-dimension seismic data to invert; prediction samples are constructed using the sensitive parameter volumes; the predicted samples are input to the RBF neural network prediction model to calculate corresponding TOC values, thereby the TOC content of the target stratum is predicted.

First claim

What is claimed is:

1 . A seismic quantitative prediction method for shale total organic carbon (TOC) based on sensitive parameter volumes, wherein the seismic quantitative prediction method comprises:
step ( 1 ) determining a target stratum for a TOC content to be measured in a stratum; obtaining logging data of the target stratum and post stack three-dimensional (3D) seismic data; determining M numbers of depths at equal intervals on the target stratum and obtaining a TOC content of a core at each of the M numbers of depths; wherein the logging data comprise a plurality of logging curves; step ( 2 ) performing a correlation analysis on each of the plurality of logging curves and the TOC contents at the M numbers of depths to obtain a correlation coefficient between each logging curve and the TOC contents; setting a threshold and retaining the logging curves with the correlation coefficient greater than the threshold as sensitive parameters; and the number of the sensitive parameters being N, and the sensitive parameters being labeled as first to Nth sensitive parameters; step ( 3 ) constructing sample data:
constructing the sample data at each depth of the target stratum, wherein the sample data at a jth depth of the M numbers of depths is L j , L j ={L 1j , L 2j , . . . , L ij , . . . , L Nj }, where L ij represents a value of the ith sensitive parameter at the jth depth, i=1˜N, and j=1˜M;
step ( 4 ) establishing a radial basis function (RBF) neural network, and training the RBF neural network with the sample data as an input and the TOC content at the depth corresponding to the sample data as an output to obtain a RBF neural network prediction model; step ( 5 ) for the first to Nth sensitive parameters, using each sensitive parameter as a constraint, obtaining sensitive parameter volumes by performing inversion based on the post stack 3D seismic data; and the sensitive parameter volumes being labeled as first to Nth sensitive parameter volumes; step ( 6 ) constructing prediction samples, comprising steps ( 61 )˜( 65 );
step ( 61 ) forming a three-dimensional data volume of P×Q×H for each sensitive parameter volume, each sensitive parameter volume having a same size, and a line number, a trace number, and a sampling point of each sensitive parameter volume being P, Q, and H, respectively;
step ( 62 ) organizing the first sensitive parameter volume into a two-dimensional matrix of K×H, and converting the two-dimensional matrix into a one-dimensional array of 1×L wherein K=P×Q, and L=K×H;
step ( 63 ), processing the second to the Nth sensitive parameter volumes into one-dimensional arrays respectively; and in the one-dimensional arrays, elements at same positions correspond to same coordinate values;
step ( 64 ) taking the one-dimensional arrays corresponding to the first to the Nth sensitive parameter volumes as data from first to Nth rows of a matrix, respectively, to form a prediction matrix of N×L; and
step ( 65 ) extracting each column of the data from the prediction matrix to form a prediction sample, and the number of the prediction sample being L;
step ( 7 ) inputting the L numbers of prediction samples into the RBF neural network prediction model, outputting L numbers of TOC values, and for each TOC value, using coordinates of the prediction sample corresponding to the TOC value as coordinates of the TOC value to obtain a TOC value with coordinates, thereby obtaining a one-dimensional TOC array; the coordinates comprising a line number, a trace number, and a sampling point; step ( 8 ) transforming the one-dimensional TOC array to form a three-dimensional TOC data volume of P×Q×H according to the coordinates, thereby predicting a TOC content of the target stratum; and step ( 9 ) performing, based on the TOC content of the target stratum, shale gas exploration and development on the target stratum.

2 . The seismic quantitative prediction method for the shale TOC based on the sensitive parameter volumes as claimed in claim 1 , wherein the logging curves comprise a density curve, an interval transit time curve, a porosity curve, a resistivity curve, a potassium content curve and a uranium content curve, and the TOC content of the core at each depth is obtained by a geochemical analysis of a TOC content of a shale core.

3 . The seismic quantitative prediction method for the shale TOC based on the sensitive parameter volumes as claimed in claim 1 , wherein the correlation analysis in step ( 2 ) comprises using a statistical product and service solutions (SPSS) statistical analysis software for the correlation analysis.

4 . The seismic quantitative prediction method for the shale TOC based on the sensitive parameter volumes as claimed in claim 1 , wherein the performing, based on the TOC content of the target stratum, oil and gas exploration of the target stratum comprises:
in response to the TOC content of the target stratum being greater than 2%, exploring shale gas in the target stratum.

5 . The seismic quantitative prediction method for the shale TOC based on the sensitive parameter volumes as claimed in claim 1 , wherein the TOC content of the target stratum is used to guide exploration and development of unconventional oil and gas reservoirs of the target stratum.

6 . A seismic quantitative prediction method for shale total organic carbon (TOC) based on sensitive parameter volumes, wherein the seismic quantitative prediction method comprises:
step ( 1 ) determining a target stratum for a TOC content to be measured in a stratum; obtaining logging data of the target stratum and post stack three-dimensional (3D) seismic data; determining M numbers of depths at equal intervals on the target stratum and obtaining a TOC content of a core at each of the M numbers of depths; wherein the logging data comprise a plurality of logging curves, where the target stratum comprises multiple sub-stratums; step ( 2 ) performing a correlation analysis on each of the plurality of logging curves and the TOC contents at the M numbers of depths to obtain a correlation coefficient between each logging curve and the TOC contents; setting a threshold and retaining the logging curves with the correlation coefficient greater than the threshold as sensitive parameters; and the number of the sensitive parameters being N, and the sensitive parameters being labeled as first to Nth sensitive parameters; step ( 3 ) constructing sample data:
constructing the sample data at each depth of the target stratum, wherein the sample data at a jth depth of the M numbers of depths is L j , L j ={L 1j , L 2j , . . . , L ij , . . . , L Nj }, where L ij represents a value of the ith sensitive parameter at the jth depth, i=1˜N, and j=1˜M;
step ( 4 ) establishing a radial basis function (RBF) neural network, and training the RBF neural network with the sample data as an input and the TOC content at the depth corresponding to the sample data as an output to obtain an RBF neural network prediction model; step ( 5 ) for the first to Nth sensitive parameters, using each sensitive parameter as a constraint, obtaining sensitive parameter volumes by performing inversion based on the post stack 3D seismic data; and the sensitive parameter volumes being labeled as first to Nth sensitive parameter volumes; step ( 6 ) constructing prediction samples, comprising steps ( 61 )˜( 65 );
step ( 61 ) forming a three-dimensional data volume of P×Q×H for each sensitive parameter volume, each sensitive parameter volume having a same size, and a line number, a trace number, and a sampling point of each sensitive parameter volume being P, Q, and H, respectively;
step ( 62 ) organizing the first sensitive parameter volume into a two-dimensional matrix of K×H, and converting the two-dimensional matrix into a one-dimensional array of

Assignees

• Univ Chengdu Technology

Inventors

• Wu Chaorong
• Liu Cheng
• Huang Kaixing
• Li Yong
• Li Yizhen
• Li Junxiang
• Hao Yuexiang

Classifications

• G01V2210/6169

  using well-logging · CPC title

• G01V1/306

  for determining physical properties of the subsurface, e.g. impedance, porosity or attenuation profiles · CPC title

• G01V1/301

  for determining seismic cross-sections or geostructures · CPC title

• G01V1/30 Primary

  Analysis (G01V1/50 takes precedence) · CPC title

• G01V1/282 Primary

  Application of seismic models, synthetic seismograms · CPC title