Seismic data resolution improving method based on sparse constraint deconvolution

Abstract

The invention provides a seismic data resolution improving method based on sparse constraint deconvolution. The common problems that in actual seismic exploration, a post-stack section is low in signal-to-noise ratio, insufficient in resolution, insufficient in data precision and the like are solved. Comprising the following steps: S1, determining a target iteration solution termination condition; s2, solving a plurality of smooth wavelets by using the post-stack seismic section; s3, performing fast Fourier transform on the smooth wavelet to obtain an amplitude spectrum; s4, obtaining a smoothed amplitude spectrum; s5, performing fast inverse Fourier transform by using the amplitude spectrum after smoothing to obtain an initial wavelet; and S6, substituting the initial wavelet into an improved deconvolution iteration solution formula for solution, and outputting a result after deconvolution when the iteration process meets a termination condition. In actual data testing, sparse constraint deconvolution processing is carried out on seismic data, the effective frequency domain of signals is widened, the dominant frequency is improved, and it is verified that the method has the effect of improving the longitudinal resolution of the seismic data.

Description

Technical Field

[0001] This invention belongs to the field of seismic data processing and serves to improve the signal-to-noise ratio of seismic data and to predict reservoirs using seismic attenuation characteristics, thereby obtaining more accurate amplitude spectrum results and improving the quality of seismic exploration data. Background Technology

[0002] Seismic data deconvolution methods: Post-stack seismic data deconvolution is a key technique for improving resolution. The basic principle of deconvolution is to obtain the reflection coefficient by deconvolving the effective wave and wavelet of the seismic trace. Traditional deconvolution methods include the following categories:

[0003] (1) Predictive deconvolution is based on the information of regular interference such as primary and multiple waves in the original seismic data. The pure interference part is predicted and then subtracted from the original seismic record to obtain the primary wave after interference elimination, thereby improving the signal-to-noise ratio and resolution of the seismic record. The interference side lobes with opposite polarity generated in the middle of the multiple waves will have a noise-like effect, which will have a significant impact on the next step of deconvolution. Furthermore, for multiple waves with multiple periods and different periods, as well as for multiple waves that are very close to the effective wave or have overlapping distributions, predictive deconvolution is difficult to accurately identify and process.

[0004] (2) Surface consistent deconvolution is mainly used to eliminate the differences in wavelets between seismic records caused by factors such as excitation and reception. At the same time, it performs inter-channel amplitude equalization compensation, maintains the phase consistency of the data, broadens the spectrum, corrects the phase spectrum of the seismic signal, and outputs zero-phase wavelets. However, its effect may be limited when the surface conditions are complex or the quality of the seismic data is poor.

[0005] (3) Least square deconvolution is the most commonly used type of deconvolution in seismic exploration. It was first proposed by Wiener in 1947, hence it is also called Wiener filtering. Least square deconvolution solves for the reflection coefficient by minimizing the sum of squared errors, and has high accuracy. Conventional least square deconvolution methods are suitable for minimum phase wavelets and Gaussian white noise reflection sequences. However, in actual data, the original seismic wavelet is not a minimum phase wavelet, which limits its practical application.

[0006] (4) Pulse deconvolution aims to obtain a pulse-shaped reflection coefficient sequence, transforming the actual seismic record into a sharp pulse corresponding to the reflection interface, thereby reflecting the underground geological structure more clearly. However, it has problems such as poor low-frequency recovery and severe interference from random noise. Summary of the Invention

[0007] This invention addresses the problems existing in the background technology by proposing a seismic data resolution enhancement method based on sparse constraint theory. The method comprises three core components: constructing a sparse objective function with regularization constraints, obtaining the initial wavelet, and solving for the reflection coefficient sequence. Furthermore, a corresponding standard workflow for actual seismic data resolution enhancement is developed, the specific implementation of which is as follows:

[0008] A method for improving the resolution of seismic data based on sparse-constrained deconvolution includes the following steps:

[0009] S1. Analyze the existing post-stack seismic data to determine the termination condition for the target iterative solution;

[0010] S2. Obtain multi-channel smooth wavelet using post-stack seismic profiles;

[0011] S3. Perform a fast Fourier transform on the smooth wavelet to obtain the amplitude spectrum;

[0012] S4. Obtain the smoothed amplitude spectrum. ;

[0013] S5. Use the smoothed amplitude spectrum to perform a fast inverse Fourier transform to obtain the initial wavelet;

[0014] S6. Substitute the initial wavelet into the improved deconvolution iterative solution formula for solving. When the iteration process meets the termination condition, output the deconvolution result.

[0015] Preferably, in S2, the multichannel smooth wavelet is obtained by the following formula:

[0016] In the formula, Represents the seismic wavelet, N sunm Indicates the number of channels. This represents a smooth sub-wave.

[0017] Preferably, in S3, the amplitude spectrum .

[0018] Preferably, in S4, the amplitude spectrum is smoothed by performing three B-spline smoothing to obtain the smoothed amplitude spectrum.

[0019] The preferred and improved deconvolution iterative solution formula is as follows:

[0020]

[0021] In the formula, the upper equation is the objective function, and r represents the reflection coefficient to be solved; the lower equation represents the constraint conditions. Represents the smallest absolute value among all non-zero reflectance coefficients. This represents the maximum absolute value among all non-zero reflectance coefficients. It represents a proportionality constant.

[0022] Preferred, The value is between 0 and 1.

[0023] Preferred, The value is 0.5.

[0024] Beneficial effects of the present invention

[0025] The beneficial effects of this invention lie in its significant improvement of seismic data processing resolution through the introduction of sparse-constrained deconvolution technology. This enables sharper separation of closely adjacent reflection layers, greatly enhancing vertical resolution and resulting in a qualitative leap in the identification of thin interlayers. Simultaneously, this sparse constraint naturally suppresses noise, achieving stable and high-fidelity reflection coefficient sequences even under low signal-to-noise ratio conditions, effectively avoiding noise amplification and spurious phase axis generation problems common in traditional methods. Furthermore, this method reduces the stringent requirements on wavelet morphology and seismic signal stability, exhibiting greater adaptability and practicality, and providing more accurate foundational data for subsequent fine reservoir characterization and direct hydrocarbon detection. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the main implementation of the present invention.

[0027] Figure 2 This is the pre-stacked cross section before deconvolution.

[0028] Figure 3 This is a cross-section after conventional deconvolution.

[0029] Figure 4 This is the cross-section after deconvolution in this invention.

[0030] Figure 5 This is a spectrum comparison diagram from the examples. Detailed Implementation

[0031] The present invention will be further described below with reference to embodiments, but the scope of protection of the present invention is not limited thereto:

[0032] This invention proposes a seismic data resolution enhancement method based on sparse constraint theory, comprising three core components: construction of a sparse objective function with regularization constraints, initial wavelet acquisition, and solution of the reflection coefficient sequence. Furthermore, a corresponding standard procedure for practical seismic data resolution enhancement is developed based on this method. This application employs a sparse-constrained deconvolution method under the steady-state wavelet assumption. By introducing sparse constraint terms into the deconvolution process, the deconvolution is transformed into a nonlinear objective functional solution process, thus optimizing the deconvolution results. Its advantages over other traditional methods are:

[0033] (1) Improved resolution: Under sparse constraints, deconvolution processing can more accurately separate seismic wavelets and reflection coefficients, significantly recover high-frequency components outside the effective frequency band of seismic records, making the reflection coefficient sequence sparser and clearer, thereby improving the resolution of seismic records;

[0034] (2) It has greater adaptability and flexibility. Traditional deconvolution methods, such as least squares deconvolution, usually assume that the seismic data is a linear signal or that the wavelet is in minimum phase, and that the reflection coefficient is white noise. However, actual seismic data is often nonlinear, and the reflection coefficient does not always satisfy the white noise assumption. In contrast, sparse-constrained deconvolution does not require these assumptions, and is therefore more suitable for processing actual seismic data.

[0035] (3) Maintaining a high signal-to-noise ratio, under sparse constraints, deconvolution processing can more effectively suppress noise and interference waves, thus obtaining clearer and more accurate seismic records.

[0036] (4) Optimize algorithms and computational efficiency. Sparse constraint deconvolution is usually optimized using regularization methods and iterative solution algorithms. These algorithms can stabilize the solution process and obtain more sparse and accurate solutions.

[0037] Combination Figure 1 The specific implementation process is as follows:

[0038] (1) Construct a sparse constraint objective function containing regularization constraints.

[0039] According to the traditional seismic data convolution model, the seismic record d can be represented as:

[0040] (1)

[0041] In the formula, For seismic wavelets; It is a sequence of reflection coefficients; t represents noise; t represents time; and "*" represents the convolution operator.

[0042] The convolution operation model in equation (1) can be rewritten using the wavelet matrix as follows:

[0043] (2)

[0044] In the formula This is the sum of the number of seismic data traces. To select the first earthquake data road, For the first Reflectance coefficient, For the first Noise from the seismic tunnel. The matrix is ​​the seismic wavelet in equation (1). The translation along the diagonal forms a Toeplitz matrix, which can be expressed in the following form:

[0045] (3)

[0046] As can be seen from equation (3), It has a distinct Toeplitz structure, therefore The solution process can be accelerated by introducing forward and inverse Fast Fourier Transforms.

[0047] When earthquake data It is known that the process of solving for the unknown wavelet and the reflection coefficient sequence can be expressed in the following form:

[0048] (4)

[0049] Equation (4) is a classic ill-posed problem, and it is difficult to converge to the optimal solution during the solution process. Currently, a common method to improve well-posedness is to add regularization constraints. This invention introduces an improved method... The regularization term rewrites equation (4) in the following form:

[0050] (5)

[0051] In the formula This is a hyperparameter, and its value varies depending on the type of seismic data.

[0052] (5) Initial wavelet acquisition

[0053] At the beginning of the calculation, multiple average seismic records were obtained. :

[0054] (6)

[0055] And its amplitude spectrum is obtained by fast Fourier transform:

[0056] (7)

[0057] The smoothed amplitude spectrum can be obtained by smoothing with multiple cubic B-spline polynomials. Then, the smoothed amplitude spectrum The initial wavelet can be obtained by performing an inverse Fourier transform. .

[0058] (6) Solving the reflection coefficient sequence

[0059] When the wavelet is fixed, since equation (5) is calculated on a separate seismic trace, it can be solved iteratively by ADMM (alternating direction multiplier method).

[0060] The ADMM algorithm combines the decomposability of dual ascent with the convergence of the multiplier method. This algorithm can be solved in the following way:

[0061] (8)

[0062] Among them, variables and variables , at the same time and Hypothesis function and The main difference between x and the conventional linear equation Ax = b constraint problem is that x is a convex function. In equation (8), the variables are x and z. Since the objective function can be separated, equation (8) can be expressed as:

[0063] (9)

[0064] Its Lagrange augmentation formula can be written as:

[0065] (10)

[0066] The ADMM algorithm iterative process mainly consists of three steps: x minimization, z minimization, and dual variable update.

[0067] (11)

[0068] As shown in the above equation, during the iterative calculation of the ADMM algorithm, the updates of the independent variables x and z are performed alternately, hence the term "alternating direction." Using this method, the reflection coefficient sequence can be obtained relatively well. .

[0069] To test the effectiveness of the method of this invention on actual seismic data, a test was conducted using measured data from the northern Jiangsu basin. Figure 2 This is the original cross-section. Figure 3 This is the cross-section extended by the conventional deconvolution method. Figure 4 This is a cross-section after bandwidth extension according to the present invention. Comparing the three, it can be seen that the cross-sectional resolution is significantly improved after deconvolution. From Figure 5 Spectral analysis shows that the profile processed by this invention has a wider bandwidth and higher frequency.

[0070] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for improving the resolution of seismic data based on sparse-constrained deconvolution, characterized in that... It includes the following steps: S1. Analyze the existing post-stack seismic data to determine the termination condition for the target iterative solution; S2. Obtain multi-channel smooth wavelet using post-stack seismic profiles; S3. Perform a fast Fourier transform on the smooth wavelet to obtain the amplitude spectrum; S4. Obtain the smoothed amplitude spectrum. ; S5. Use the smoothed amplitude spectrum to perform a fast inverse Fourier transform to obtain the initial wavelet; S6. Substitute the initial wavelet into the improved deconvolution iterative solution formula for solution. When the iteration process meets the termination condition, output the deconvolution result.

2. The method according to claim 1, characterized in that... In S2, the multichannel smooth wavelet is obtained by the following formula: In the formula, Represents the seismic wavelet, N sunm Indicates the number of channels. This represents a smooth sub-wave.

3. The method according to claim 1, characterized in that... In S3, the amplitude spectrum .

4. The method according to claim 1, characterized in that... In S4, the amplitude spectrum is smoothed by performing a third B-spline smoothing to obtain the smoothed amplitude spectrum.

5. The method according to claim 1, characterized in that... The improved deconvolution iterative solution formula is as follows: In the above equation, the objective function is given, and r represents the reflection coefficient to be solved; the below equation represents the constraint conditions. Represents the smallest absolute value among all non-zero reflectance coefficients. This represents the maximum absolute value among all non-zero reflectance coefficients. It represents a proportionality constant.

6. The method according to claim 5, characterized in that... The value is between 0 and 1.

7. The method according to claim 5, characterized in that... The value is 0.5.

Images