Seismic data source wavelet estimation method based on cross-correlation waveform inversion

Abstract

The present application relates to a kind of seismic data source wave estimation methods based on cross-correlation waveform inversion, this method is under the framework of near trace direct wave full waveform inversion, by using normalized cross-correlation objective function, realize high quality source wave estimation.In the framework of full waveform inversion, the kinematics and dynamics characteristics of wave field are fully considered, the accuracy and resolution of inversion result are guaranteed;Using near trace direct wave information, the nonlinearity of inversion is reduced, and the calculation efficiency of inversion algorithm is guaranteed;Using cross-correlation objective function waveform inversion method, the influence of seismic data amplitude error on source wave estimation is effectively avoided.The method of the present application can be used in the case where there is obvious error in seismic data amplitude, and the direct wave cross-correlation full waveform inversion is used to efficiently and accurately estimate the source wave.

Description

Technical Field

[0001] This invention belongs to the field of seismic exploration technology, specifically relating to a method for estimating source wavelet using cross-correlation waveform inversion. Background Technology

[0002] Source wavelet estimation is a crucial step in seismic data inversion, imaging, and interpretation. Conventional source wavelet estimation methods are based on convolution models, which treat the seismic record as a convolution of the source wavelet and the sequence of subsurface reflection coefficients. Source wavelet estimates can be obtained through linear inversion or statistical methods. However, source wavelet estimation based on convolution models is only applicable to post-stack seismic data and neglects complex reflection information such as multiples. Therefore, the obtained source wavelet is not entirely suitable for wave equation-based inversion and imaging.

[0003] In full-waveform inversion, the source wavelet is often inverted simultaneously with the velocity model. Current source wavelet inversion methods mainly construct the L2 norm of the residuals between simulated and observed data, and then use the adjoint state method to obtain the update gradient of the source wavelet. However, because the data residuals are used as the standard, current source wavelet inversion methods are easily affected by data amplitude errors. For land seismic exploration, acquisition conditions, surface wave interference, noise, and data processing can all affect data amplitude, thus affecting the accuracy of source wavelet inversion. The source wavelet is the initial condition for forward modeling of the wave equation, and its accuracy directly affects the final results of inversion and imaging. Therefore, it is essential to propose a robust source wavelet estimation method that is unaffected by amplitude errors. Summary of the Invention

[0004] The purpose of this invention is to provide a method for high-quality source wavelet estimation by performing cross-correlation waveform inversion using short-path direct wave data, thereby solving the problem of obtaining accurate phase information for source wavelet estimation even when data amplitude errors exist. Using normalized cross-correlation matching of the entire waveform of the short-path direct wave as the objective function, source wavelet inversion and updating are performed to achieve high-quality source wavelet estimation even with data amplitude errors.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] First, the seismic data undergoes necessary preprocessing to determine the initial velocity model used for inversion. Then, based on the initial velocity model, the near-path observation direct wave data is backpropagated to obtain the initial source wavelet. Forward modeling is performed using the current source wavelet and velocity model to obtain the simulated direct wave data for the current iteration and to calculate the objective function for the current iteration. The associated source and associated wavefield of the source wavelet inversion are calculated to obtain the gradient of the source wavelet inversion. An appropriate step size is selected to update the source wavelet. The stopping condition for source wavelet updating is determined; if it is met, the source wavelet estimation result is output; otherwise, the wavelet is updated continuously.

[0007] A method for estimating seismic source wavelet based on cross-correlation waveform inversion includes the following steps:

[0008] a. Perform regularization, denoising, and direct wave interception preprocessing on the raw seismic data;

[0009] b. Determine the initial velocity model to be used for the inversion;

[0010] c. The objective function shown in formula (1) is the source wavelet inversion objective function based on cross-correlation waveform inversion.

[0011]

[0012] Where σ is the objective function value, s represents the source function, and ns and nr are the total number of sources and receivers, respectively. and These represent the normalized simulated direct wave data and the normalized observed direct wave data, respectively. u and d represent simulated direct wave data and observed direct wave data from the near-channel, respectively, and ||·|| represents the L2 norm;

[0013] d. Based on the initial velocity model, backpropagate the direct wave data d observed from the nearest path and extract the waveform of the backpropagated wave field corresponding to the source location as the initial source wavelet.

[0014] e. Based on the current velocity model, use the current source wavelet to calculate the simulated direct wave data, and calculate the objective function of the current iteration according to formula (1);

[0015] f. Calculate the associated source of the source wavelet inversion according to formula (2), and calculate the gradient of the source wavelet inversion according to formula (3);

[0016]

[0017]

[0018] Where A represents the adjoint source, F represents the source matrix, the superscript T represents the transpose, L represents the forward modeling operator of the wave equation, and its inverse L...-1 This represents the backpropagation operator of the wave equation;

[0019] g. Select an appropriate update step size to update the source wavelet;

[0020] h. Determine the stopping condition for source wavelet update. If the condition is met, output the source wavelet estimation result; otherwise, execute step e.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0022] This invention realizes source wavelet inversion within the framework of direct wave cross-correlation waveform inversion, solving the problem of high-quality source wavelet estimation under data amplitude error. Direct wave waveform inversion fully utilizes the kinematic and dynamic information of the seismic wave field, ensuring high-resolution source wavelet estimation. The waveform inversion method using cross-correlation objective function effectively overcomes the influence of data amplitude error on the wavelet inversion process.

[0023] This invention has the following characteristics:

[0024] 1. This method only uses near-path direct wave seismic data. On the one hand, the required inversion recording time is short and the computational efficiency is high; on the other hand, the required velocity model depth is small and less affected by underground velocity.

[0025] 2. This method performs source wavelet estimation within the framework of full waveform inversion, making full use of the kinematic and dynamic information of the wave field and ensuring the accuracy and resolution of the inversion results;

[0026] 3. This method uses waveform inversion with cross-correlation objective function, which can obtain high-quality source wavelet estimation even when there are significant errors in the data amplitude. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A flowchart of the seismic source wavelet estimation method based on cross-correlation waveform inversion described in this invention;

[0029] Figure 2a Realistic speed model;

[0030] Figure 2b Initial velocity model.

[0031] Figure 3aObserve earthquake data;

[0032] Figure 3b Observe the short-path direct wave of the seismic data;

[0033] Figure 4 Source wavelet estimation results;

[0034] Figure 5a Objective function curve of source wavelet inversion process using L2WI method;

[0035] Figure 5b Objective function curve of source wavelet inversion process using CNWI method;

[0036] Figure 6a Full waveform inversion results using back-propagated wavelet;

[0037] Figure 6b Full waveform inversion results were obtained by estimating the wavelet using L2WI;

[0038] Figure 6c The full waveform inversion results were obtained by estimating the wavelet using CNWI. Detailed Implementation

[0039] The present invention will be further described below with reference to embodiments:

[0040] The seismic source wavelet estimation method based on cross-correlation waveform inversion described in this invention is implemented using the MATLAB platform. The method includes the following steps:

[0041] 1. Install the MATLAB software platform on a Windows 7 or Linux system, requiring MATLAB R2016a or later. The Parallel Computing Toolbox must also be included.

[0042] 2. Perform data preprocessing. Perform preprocessing on the raw background noise data, including regularization, denoising, and direct wave interception.

[0043] 3. Obtain the initial velocity model required for inversion through velocity analysis or tomographic imaging methods.

[0044] 4. The objective function shown in formula (1) is the source wavelet inversion objective function based on cross-correlation waveform inversion.

[0045]

[0046] Where σ is the objective function value, s represents the source function, and ns and nr are the total number of sources and receivers, respectively. and These represent the normalized simulated direct wave data and the normalized observed direct wave data, respectively. u and d represent the simulated direct wave data and the observed direct wave data from the near-path, respectively, and ||·|| represents the L2 norm.

[0047] 5. Based on the initial velocity model, the direct wave data d observed from the near-path observation is back-propagated, and the waveform of the back-propagated wave field corresponding to the source location is extracted as the initial source wavelet.

[0048] 6. Based on the current velocity model, use the current source wavelet to calculate the simulated direct wave data, and calculate the objective function of the current iteration according to formula (1).

[0049] 7. Calculate the associated source of the source wavelet inversion according to formula (2), and calculate the gradient of the source wavelet inversion according to formula (3);

[0050]

[0051]

[0052] Where A represents the adjoint source, s represents the source function, F represents the source matrix, the superscript T represents the transpose, L represents the forward modeling operator of the wave equation, and its inverse L -1 This represents the backpropagation operator of the wave equation.

[0053] 8. Select an appropriate update step size to update the source wavelet.

[0054] 9. Determine the stopping condition for source wavelet update. If the condition is met, output the source wavelet estimation result; otherwise, proceed to step 6.

[0055] Example 1

[0056] The overall process of this invention is as follows: Figure 1 As shown.

[0057] The actual velocity model used is as follows: Figure 2a As shown, this is the Marmousi velocity model. The initial velocity model used is a smoothed result of the true velocity model, and there is a difference between the initial velocity and the true velocity to approximate the situation encountered in source wavelet inversion of actual seismic data. The true source wavelet used is the 20Hz dominant frequency Ricker wavelet. When generating the observed seismic data, a total of 48 sources were uniformly distributed on the model surface, and each true source had the same source function. The observation data from one shot is shown below. Figure 3a As shown, the near-channel data contains significant errors to simulate the influence of factors such as excitation conditions, surface waves, and subsequent processing on the near-channel data waveform during actual acquisition. Because the source wavelet estimation method proposed in this invention uses near-channel direct wave data, it... Figure 3aThe earthquake record shown extracts the short-path direct wave, such as... Figure 3b As shown.

[0058] Adopting such Figure 2b The initial velocity model shown employs the reverse-time propagation source wavelet estimation method (RTP method), the L2 norm waveform inversion wavelet estimation method (L2WI method), and the cross-correlation waveform inversion wavelet estimation method (CNWI) to estimate the source wavelet from the observed data, corresponding to... Figure 3a The source wavelet estimation results of the data are as follows Figure 4 As shown in the figure. It can be seen that due to the impact of short-channel data quality, the RTP method and L2WI method can no longer obtain accurate source wavelet information, while the CNWI method proposed in this invention can still obtain relatively accurate source wavelet phase information. The objective function curves of the inversion process of the L2WI method and CNWI method are shown in the figure. Figure 5a and Figure 5b As shown, the CNWI method requires fewer iterations to converge than the L2WI method, indicating that the CNWI method has higher computational efficiency. Figure 2b The model shown is the initial velocity model. Full waveform inversion based on a cross-correlation objective function is performed using the estimated source wavelet to verify whether the accuracy of the source wavelet estimation meets the inversion requirements. The results of full waveform inversion using the RTP method, L2WI method, and the CNWI method proposed in this invention are as follows: Figure 6a , Figure 6b and Figure 6c As shown. It can be seen that, Figure 6a and Figure 6b The inversion results show obvious high-velocity anomalies in the shallow part, while Figure 6c The inversion results are the best, and the velocity structure is close to the true velocity model. Therefore, the seismic source wavelet estimation method based on cross-correlation waveform inversion proposed in this invention can obtain reliable source wavelet estimation results even when there are errors in the amplitude of seismic data.

[0059] This invention discloses a source wavelet estimation method based on cross-correlation waveform inversion of seismic data. The core step is to perform high-quality source wavelet estimation using cross-correlation waveform inversion of near-path direct wave data. Utilizing near-path direct wave data for full waveform inversion reduces the nonlinearity of the inversion and ensures the overall efficiency of the algorithm. Performing source wavelet inversion within the framework of full waveform inversion fully utilizes the kinematic and dynamic information of the wavefield, guaranteeing the accuracy and resolution of the inversion results. Employing a full waveform inversion method based on a cross-correlation objective function effectively avoids the influence of observational data amplitude errors on the source wavelet estimation results.

[0060] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A seismic data source wavelet estimation method based on cross-correlation waveform inversion, characterized in that, The method comprises the following steps: a. Regularization, denoising, and direct wave cutting preprocessing are performed on original seismic data; b. An initial velocity model for inversion is determined; c. The objective function shown in formula (1) is a source wavelet inversion objective function based on cross-correlation waveform inversion where σ is the objective function value, s represents the source function, ns and nr are the total number of sources and receivers, respectively, and and u and d represent the near-trace simulated and observed direct wave data, respectively, and ||·|| represents the L2 norm. d. On the initial velocity model, the near trace observation direct wave data d is back propagated, and the waveform of the corresponding back propagation wave field at the source position is extracted as the initial source wavelet; e. On the current velocity model, the current source wavelet is used to calculate the simulated direct wave data, and the objective function of the current iteration is calculated according to formula (1); f. The adjoint source of the source wavelet inversion is calculated according to formula (2), and the gradient of the source wavelet inversion is calculated according to formula (3); where A denotes the adjoint source, F denotes the source matrix, the upper index T denotes the transpose, L denotes the wave equation forward operator, its inverse L -1 denotes the wave equation backward operator; g. A suitable update step is selected to update the source wavelet; h. A stop condition for updating the source wavelet is judged, if the stop condition is met, the source wavelet estimation result is output, if the stop condition is not met, the step e is executed.

Images