CN-118033743-B - Elastic wave inversion method based on natural frequency division time-frequency domain energy spectrum

CN118033743BCN 118033743 BCN118033743 BCN 118033743BCN-118033743-B

Abstract

The invention relates to an elastic wave inversion method based on natural frequency division time-frequency domain energy spectrum, which utilizes S transformation to extract single-frequency energy spectrum information of seismic data and nest frequency point circulation to realize natural frequency division multi-scale inversion, establishes a more accurate longitudinal and transverse wave velocity initial model under the condition of low frequency loss, and carries out frequency division group iterative updating of elastic wave full waveform inversion by utilizing a multi-scale strategy on the basis to obtain an accurate longitudinal and transverse wave velocity model, thereby improving the accuracy and reliability of the result of elastic wave full waveform inversion under the condition of low frequency loss.

Inventors

LI FANG
LI XIAOZHANG
LIU SHIYOU
HE JIANWEI
LIU BING
WANG RUI

Assignees

中海石油(中国)有限公司海南分公司

Dates

Publication Date: 20260508
Application Date: 20240318

Claims (1)

1. An elastic wave inversion method based on natural frequency division time-frequency domain energy spectrum is characterized by comprising the following steps: S1, inputting actual measurement three-component seismic records, reading in a field observation system from the head of actual measurement data, reading in a seismic wavelet, inversion parameters and an initial longitudinal wave velocity model mvp1 (0) and a transverse wave layer velocity model mvs1 (0), and carrying out grid subdivision on a calculation space according to the inversion parameters; S2, performing forward modeling of an elastic wave equation by using the set parameters and the current model to obtain a simulated seismic record currecordx, currecordz of x and z components under the current speed model; S3, reading in the measured seismic records orirecordx, orirecordz of the x and z components, determining the frequency range of the time-frequency domain inversion [ , Setting inversion frequency interval Inversion frequency point number Performing single-frequency S transformation on the x and z component simulation seismic records to obtain curx and curz, performing single-frequency S transformation on the x and z component actual measurement seismic records to obtain orix and oriz, and respectively taking squares of the amplitudes of curx, curz, orix, oriz to obtain respective single-frequency energy spectrums; S4, performing multi-scale inversion by utilizing natural frequency-division time-frequency domain energy spectrum elastic wave full waveform inversion to obtain a longitudinal and transverse wave layer speed model containing large-scale information; S5, taking a velocity model containing large-scale information as an initial model to perform full waveform inversion of the mixed domain multi-scale elastic wave to obtain a final high-precision inversion result; in the step S4, the specific embodiment is as follows: assuming a current frequency point number f, wherein the maximum iteration number under the current frequency point is Niter < 1 >, and respectively obtaining a single-frequency energy spectrum of curx, curz, orix, oriz according to S3, wherein the initial longitudinal wave model mvp1 (j, f) and the initial transverse wave model mvs1 (j, f) of the jth iteration of the f frequency point are provided 、、、 Calculating the accompanied source adjx and adjz of the full waveform inversion as ; Wherein, the In order to find the real part of the signal, The recording time is indicated as a time of recording, Representing the units of an imaginary number, Representing a time delay; Using elastic wave equation to reversely transmit adjx, adjz to obtain an accompanying wave field, using the accompanying wave field and the simulated wave field to obtain gradient according to gradient type or Newton type method, determining the update direction directionvp1 of the longitudinal wave and the update direction directionvs1 of the transverse wave of the iteration, and determining the update step length a; the longitudinal wave model update formula mvp1 (j+1, f) =mvp1 (j, f) +a directionvp1; The transverse wave model update formula mvs1 (j+1, f) =mvs1 (j, f) +a directionvs1; updating the speed model, judging whether the current iteration number j of the frequency point f is equal to Niter1, if so, entering the next frequency point, and if not, carrying out the next iteration of the current frequency point If the frequency is equal to the frequency, outputting a final longitudinal wave velocity model and a transverse wave velocity model, and if the frequency is not equal to the frequency, f=f+1, j=0, and entering the iteration of the next frequency point.

Description

Elastic wave inversion method based on natural frequency division time-frequency domain energy spectrum Technical Field The invention relates to the technical field of seismic exploration inversion, in particular to an elastic wave inversion method based on natural frequency division time-frequency domain energy spectrum. Background Velocity modeling problems are one of the most prominent problems in seismic exploration. The accuracy of velocity modeling can affect the authenticity and reliability of the seismic imaging, interfering with interpretation of the seismic data. How to build a high-precision velocity model is the key to solving the problem. The velocity modeling method mainly comprises three types, namely travel time chromatography, wherein only travel time information of seismic records is utilized, only a velocity model with lower precision can be obtained, the second type is velocity analysis, the modeling method is influenced by subjective factors of operators, and the third type is a full waveform inversion method, and the method can utilize all information in the seismic records, so that a velocity model with high precision can be obtained. The elastic wave full waveform inversion is more nonlinear than the acoustic wave full waveform inversion. The difficulty of the full waveform inversion of the elastic wave is that the transverse wave velocity model inversion has stronger dependence on an initial model and huge calculated amount when the elastic wave inversion is carried out, the former influences the accuracy of the transverse wave velocity model in the full waveform inversion of the elastic wave, and the latter is directly related to the application of the related technology in production. Sirgue provides a mixed domain full waveform inversion method based on time domain forward modeling and frequency domain inversion, which combines the advantages of time domain forward modeling and frequency domain inversion, and a Fourier transform is added in the time domain forward modeling process to obtain a frequency domain wave field. Wangwei et al propose a multi-scale inversion method based on envelope, which utilizes envelope inversion to provide an initial model containing large-scale information for full waveform inversion of the mixed domain elastic wave, thereby greatly relieving the phenomenon of cycle skip of full waveform inversion of the mixed domain elastic wave. In the multi-scale inversion method based on the envelope, firstly, the envelope of the seismic record is obtained, the envelope is utilized to perform inversion to obtain a large-scale initial model, and then the obtained initial model is utilized to realize high-precision establishment of a speed model by applying a Sirgue mixed domain full-waveform inversion method. The envelope inversion is based on the full frequency band of the seismic record, and if the seismic record is required to be subjected to the frequency division envelope inversion, the arrangement of a low-pass filter in the process can also influence the effect of the envelope inversion. Therefore, although the scheme reduces the dependence of the elastic wave full waveform inversion on low-frequency information to a certain extent, the additionally introduced low-pass filter increases the workload, and the frequency division frame of the multi-scale inversion is unnatural. Disclosure of Invention The invention provides an elastic wave inversion method based on natural frequency division time-frequency domain energy spectrum, which reduces error generation, reduces probability of inversion falling into local extremum and improves accuracy of a speed model, and aims to solve the problem that frequency division inversion in the elastic wave full waveform inversion technology based on envelope in the prior art. The elastic wave inversion method based on natural frequency division time-frequency domain energy spectrum comprises the following steps: S1, inputting actual measurement three-component seismic records, reading in a field observation system from the head of actual measurement data, reading in a seismic wavelet, inversion parameters and an initial longitudinal wave velocity model mvp1 (0) and a transverse wave layer velocity model mvs1 (0), and carrying out grid subdivision on a calculation space according to the inversion parameters; S2, performing forward modeling of an elastic wave equation by using the set parameters and the current model to obtain a simulated seismic record currecordx, currecordz of x and z components under the current speed model; S3, reading in the measured seismic records orirecordx, orirecordz of the x and z components, determining the frequency range of the time-frequency domain inversion [ ,Setting inversion frequency intervalInversion frequency point numberPerforming single-frequency S transformation on the x and z component simulation seismic records to obtain curx and curz, performing single-frequency S transformation on