阅读说明：本技术 一种可用于全波形反演技术的提取子波的方法 (Wavelet extraction method for full waveform inversion technology ) 是由 张金淼 丁继才 朱振宇 孙文博 姜秀娣 翁斌 张益明 王艳冬 糜芳 王清振 于 2019-02-28 设计创作，主要内容包括：本发明公开了一种可用于全波形反演技术的提取子波的方法,该提取子波的方法包括以下步骤：抽取数据；正演；求取平均数字滤波器；通过褶积运算得到初步提取子波；提取最终子波。本发明公开的可用于全波形反演技术的提取子波的方法利用观测数据和正演数据提取子波,契合全波形反演技术,符合全波形反演需求,该提取子波过程简单,算法稳定,能降低全波形反演在运用过程中发生波形错位的风险。(The invention discloses a method for extracting wavelets for a full waveform inversion technology, which comprises the following steps: extracting data; forward modeling; calculating an average digital filter; obtaining a primary extracted wavelet through convolution operation; and extracting the final wavelet. The wavelet extracting method for the full waveform inversion technology, disclosed by the invention, extracts wavelets by utilizing observation data and forward data, is matched with the full waveform inversion technology, meets the full waveform inversion requirement, has a simple wavelet extracting process and a stable algorithm, and can reduce the risk of waveform dislocation in the application process of the full waveform inversion.)

1. A method for extracting wavelets for use in full waveform inversion techniques, said method comprising the steps of:

extracting data: extracting a shot set from data to be extracted, and extracting data in a rectangular window from a small-to-medium offset distance part of the shot set to enable the rectangular window to contain direct waves to obtain an observation data subset d;

forward modeling: by means of Ricker wavelets with lower dominant frequencies, forward modeling is carried out by a forward modeling module of full waveform inversion software to obtain forward modeling data, the forward modeling data are consistent with observation data in size and have the same structure, data extraction is carried out from the forward modeling data strictly according to the data extraction mode of the observation data, and a forward modeling data subset p which is consistent with the observation data subset in size and has the same structure is obtained;

calculating an average digital filter;

obtaining a primary extraction wavelet through convolution operation: performing convolution by using an average digital filter and the Ricker wavelet to obtain a primary extracted wavelet;

extracting final wavelets: and (3) carrying out low-pass filtering on the preliminary extracted wavelet by using a FILTER module in seismic processing software 'ParadigmTM-2011.3-EchosTM' to obtain a final extracted wavelet.

2. The method of extracting wavelets usable with full waveform inversion techniques according to claim 1 wherein said extracting data is preceded by the steps of:

step S1011: firstly, suppressing random noise by using an AMPSCAL amplitude balance noise suppression module in seismic processing software 'ParadigmTM-2011.3-EchosTM'; then, a SUPPRES band-limited noise attenuation module in seismic processing software 'ParadigmTM-2011.3-EchosTM' is used for suppressing nonlinear noise;

step S1012: suppressing low-frequency linear noise by using a LEAF linear noise suppression module in seismic processing software 'ParadigmTM-2011.3-EchosTM';

step S1013: firstly, realizing the band-pass filtering of seismic data by using a FILTER module in seismic processing software 'ParadigmTM-2011.3-EchosTM'; and then, the MUTE module in the seismic processing software 'ParadigmTM-2011.3-EchosTM' is used for realizing the cutting operation of the seismic data, and OFFSET and TIME parameters are used for cutting residual noise.

3. The method of extracting wavelets usable with full waveform inversion techniques of claim 1 wherein said evaluating a digital filter comprises the steps of:

step S1041: for each pair of single tracks d in the observation data subset d and the forward data subset p_iAnd p_iCalculating a digital filter w_iSuch that the convolution of the forward single pass data and the single pass digital filter equals the single pass observed data, i.e.

i is the track number of the track,

step S1042: repeating the step S1041, and calculating each digital filter in sequence until the digital filters corresponding to all the channels of the data subset are calculated;

step S1043: and performing arithmetic mean calculation on all the digital filters to obtain a mean digital filter.

Technical Field

The invention relates to the field of oil-gas exploration, in particular to a method for extracting wavelets for a full-waveform inversion technology.

Background

Wave propagation in a subsurface fluid porous medium is one of core research contents in the field of oil and gas exploration, and the wave propagation speed in a high-precision medium is an important parameter for constructing a low-frequency model and seismic data reservoir inversion, and can be used for explaining lithology and physical properties of the fluid porous medium, so that guidance is provided for indicating an oil and gas sweet spot. In recent years, in the process of forward modeling data to approach real data, full waveform inversion technology is realizing that an initial model increasingly approaches a real model. The full-waveform inversion technology simultaneously utilizes amplitude information and phase information of seismic data, and improves the precision of the propagation velocity parameter of the underground medium through inversion.

The application process of the full waveform inversion technology is a process of data matching of forward data and observation data, and wavelet extraction is the core in the process. The matching accuracy of forward data and observation data in the full waveform inversion process is determined by the wavelet extraction effect, waveform dislocation can easily occur in full waveform inversion due to poor wavelets, if the phase difference between the forward data and the observation data is more than half wavelength, inversion is directly caused to be locally minimum, and inversion is finished in failure.

In recent years, many scholars make a lot of research on the aspect of the traditional wavelet extraction method, but the existing wavelet extraction methods are not developed for a full waveform inversion technology and cannot be used for the full waveform inversion technology, and even if the full waveform inversion is applied, the full waveform inversion is easy to have higher waveform dislocation risk in the application process, so that the actual application effect is poor.

Disclosure of Invention

The invention aims to provide a wavelet extraction method for a full waveform inversion technology, which is used for reducing the risk of waveform dislocation in the application process of full waveform inversion.

The invention provides a method for extracting wavelets for a full waveform inversion technology, which comprises the following steps:

extracting data: extracting a shot set from data to be extracted, and extracting data in a rectangular window from a small-to-medium offset distance part of the shot set to enable the rectangular window to contain direct waves to obtain an observation data subset d;

forward modeling: by means of Ricker wavelets with lower dominant frequencies, forward modeling is carried out by a forward modeling module of full waveform inversion software to obtain forward modeling data, the forward modeling data are consistent with observation data in size and have the same structure, data extraction is carried out from the forward modeling data strictly according to the data extraction mode of the observation data, and a forward modeling data subset p which is consistent with the observation data subset in size and has the same structure is obtained;

calculating an average digital filter;

obtaining a primary extraction wavelet through convolution operation: performing convolution by using an average digital filter and the Ricker wavelet to obtain a primary extracted wavelet;

extracting final wavelets: and (3) carrying out low-pass filtering on the preliminary extracted wavelet by using a FILTER module in seismic processing software 'ParadigmTM-2011.3-EchosTM' to obtain a final extracted wavelet.

Further, the method for preprocessing the observation data before extracting the data comprises the following steps:

step S1011: firstly, suppressing random noise by using an AMPSCAL amplitude balance noise suppression module in seismic processing software 'ParadigmTM-2011.3-EchosTM'; then, a SUPPRES band-limited noise attenuation module in seismic processing software 'ParadigmTM-2011.3-EchosTM' is used for suppressing nonlinear noise;

step S1012: suppressing low-frequency linear noise by using a LEAF linear noise suppression module in seismic processing software 'ParadigmTM-2011.3-EchosTM';

step S1013: firstly, realizing the band-pass filtering of seismic data by using a FILTER module in seismic processing software 'ParadigmTM-2011.3-EchosTM'; and then, the MUTE module in the seismic processing software 'ParadigmTM-2011.3-EchosTM' is used for realizing the cutting operation of the seismic data, and the OFFSET and TIME parameters are used for cutting the residual noise to obtain the data to be extracted.

Further, the averaging digital filter comprises the following steps:

step S1041: for each pair of single tracks d in the observation data subset d and the forward data subset p_iAnd p_iCalculating a digital filter w_iSuch that the convolution of the forward single pass data and the single pass digital filter equals the single pass observed data, i.e.

i is the track number of the track,is a convolution symbol;

step S1042: repeating the step S1041, and calculating each digital filter in sequence until the digital filters corresponding to all the channels of the data subset are calculated;

step S1043: and performing arithmetic mean calculation on all the digital filters to obtain a mean digital filter.

The invention has the following advantages:

the wavelet extracting method for the full waveform inversion technology, disclosed by the invention, extracts wavelets by utilizing observation data and forward data, is matched with the full waveform inversion technology, meets the full waveform inversion requirement, has a simple wavelet extracting process and a stable algorithm, and can reduce the risk of waveform dislocation in the application process of the full waveform inversion.

Detailed Description