A method for eliminating strong reflection amplitude seismic waves

Abstract

The invention relates to a method for eliminating strong reflection amplitude seismic waves, belonging to the technical field of oil and gas field exploration. The method firstly establishes a target layer geological model based on well logging data, analyzes the formation cause of the strong reflection amplitude seismic wave and the composite law of the seismic wave, then uses a matching pursuit algorithm based on the derivative of an inner product as a target to perform a first wavelet decomposition on the seismic data, then uses the first wavelet decomposition result as an initial input value for a second wavelet decomposition, uses the composite law of the seismic wave based on the geological model as a constraint condition to perform a second wavelet decomposition on the seismic data, and finally removes the strong reflection amplitude atoms in the second wavelet decomposition result based on the formation cause of the strong reflection amplitude seismic wave, and adjusts the amplitude coefficients of adjacent wavelet atoms, thereby accurately eliminating the strong reflection amplitude seismic wave.

Description

Technical Field

[0001] The invention relates to the technical field of oil and gas field exploration, and in particular to a method for eliminating strong reflection amplitude seismic waves. Background Art

[0002] When there is an interface with a large impedance difference underground, seismic waves with strong amplitude will be reflected in seismic exploration. When there is a reservoir nearby, the reflection of the reservoir will be submerged by the strong reflection amplitude, making it difficult to effectively predict the reservoir. Therefore, in order to accurately predict the reservoir near the strong reflection, it is necessary to separate and eliminate the strong reflection amplitude signal in the seismic data.

[0003] In the development of seismic wave technology for eliminating strong reflection amplitude over the years, decomposition and reconstruction based on matching pursuit is the mainstream method. This method believes that seismic records can be decomposed into a series of linear combinations of wavelet atoms, and the strong reflection amplitude can be eliminated by removing or weakening the wavelet atoms corresponding to the strong reflection amplitude and recombining the wavelet atoms.

[0004] The similarities and differences between the existing technologies can be divided into three aspects according to the differences in their matching pursuit elements: (1) Matching wavelet mother function: Gabor wavelet was used in the early stage of matching pursuit (Mallat, 1993), and Ricker wavelet and Morlet wavelet were subsequently introduced into matching pursuit (Liu J, 2004; Liu J, 2005), which are the two mainstream mother functions used at this stage; (2) Wavelet library establishment method: Matching pursuit must establish a wavelet library containing the best wavelet. Generally, more accurate initial parameters are calculated through signal analysis, and a neighborhood of the initial parameters is used as the wavelet library. The difference lies in the calculation method of the initial parameters (Li Chuanhui, 2012; Zhang Fanchang, 2016; Yin Xingyao, 2018); (3) Matching pursuit algorithm: Although the matching pursuit algorithm has been developed for many years, the most stable and most commonly used algorithm is still the full-area optimization algorithm with inner product as the objective function.

[0005] Based on sufficient research and actual data application, the present invention believes that the existing matching pursuit method for eliminating strong reflection amplitude seismic waves still has problems in at least the following aspects:

[0006] 1. The optimal solution deviates from the true solution: Matching pursuit is based on the assumption that the seismic waveform is orthogonal to the best atom. However, when the seismic waveform is a composite result of multiple reflections, the assumption that the seismic record is orthogonal to the best atom no longer holds, and the optimal solution of the matching pursuit iteration will deviate from the true solution (see Appendix Figure 3 ). In practical problems, these two situations correspond to the reflection interface being too close (0-80m), the reflected waves interfering with each other, and the reflected waves being the composite result of multi-band components. These two situations are widely present in the problem of strong reflection, and there is no solution to this in the prior art;

[0007] 2. Matching pursuit algorithm: The existing matching pursuit algorithm is still based on the global optimization of the dynamic atomic library. This method is highly dependent on the calculation of the initial values ​​of the atoms. When the matching atoms are more complex, the dynamic range is larger, or the initial parameters are inaccurate, the decomposition iterative algorithm has low computational efficiency. Summary of the invention

[0008] In order to solve the problem that the existing technology cannot accurately decompose strong reflection atoms and has low computational efficiency, the present invention proposes a method for eliminating strong reflection amplitude seismic waves. The method first establishes a strong reflection model, then performs preliminary matching pursuit decomposition on the seismic data, and then further updates the decomposition results based on the strong reflection model. Finally, the strong reflection atoms and their adjacent atoms are removed or adjusted according to geological significance, thereby overcoming the above-mentioned defects and achieving good results. The specific technical solution is as follows:

[0009] S01: Establish a strong reflection model based on the well logging data in the study area, analyze the causes of strong reflection amplitude seismic waves, and clarify whether strong reflections will interfere with adjacent reflections and whether there are multi-band components. In other words, clarify whether the iterative optimal solution of strong reflections deviates from the true solution.

[0010] The preliminary decomposition module S1 is used to obtain the preliminary solution of matching pursuit:

[0011] S11: Search and record the maximum amplitude a0 and its corresponding time t0 in the seismic data;

[0012] S12: Calculate the instantaneous frequency f0 of the seismic trace at time t0, and search for the smaller one of the in-phase extreme values ​​adjacent to a0, record it as a, and use it to calculate the matching wavelet scale factor b0;

[0013] S13: construct an initial matching wavelet w(a0, t0, f0, b0) according to the obtained parameters;

[0014] S14: Build a dynamic atom library D, use the gradient descent method to iteratively search for the best matching atom w1, and obtain the residual seismic trace; use the residual seismic trace as the input signal, and go to S11 to continue the loop iteration until the energy of the residual signal is less than the given threshold value.

[0015] Further explanation, the mother function of the matching wavelet involved in S12 and S13 is:

[0016]

[0017] Where, t is time; f is the main frequency of the wavelet; is the scale factor, A1 and A2 are the amplitude of the main peak of the sub-wave and the amplitude of the in-phase peak adjacent to the main peak respectively; is the initial phase of the wavelet (the processed seismic data is generally zero phase). In practical applications, the relationship between the wavelet mother function and the parameters involved in S12 and S13 is:

[0018]

[0019] Further explanation, the optimization function involved in S14 is:

[0020]

[0021] Among them, w 1n represents the best matching atom obtained by the nth decomposition in the initial decomposition; the dynamic atom library D is the atom library around w 1n A neighborhood established by the initial parameters (a0, t0, f0, b0):

[0022]

[0023] argmin represents the function value corresponding to the minimum value of the objective function; s n is the residual seismic trace obtained by the nth decomposition, | n ,s n >′| is s n The absolute value of the derivative of the inner product with itself.

[0024] The model decomposition module S2 is used to adjust the preliminary solution to solve the deviation of the optimal solution of the matching pursuit iteration from the true solution, so that the result is closer to the true solution:

[0025] S21: Based on the strong reflection model, the interfering atoms are merged into a new atom w2;

[0026] S22: Perform matching pursuit decomposition on the interference atom w2 again to obtain its best solution w3, and subtract the best matching result w3 from the seismic data to obtain the residual seismic trace;

[0027] S23: Split the w3 merged state to obtain the best matching result of each interfering atom in the merged state; take the residual seismic trace as the input signal, and go to S21 to continue the loop iteration until the energy of the residual signal is less than the given threshold value.

[0028] Further explanation, the merging and splitting of the matching atoms involved in S21 and S22 refers to: after the model forward modeling of the well logging model confirms the existence of interference between interfaces (generally speaking, the interface distance is The following influences exist, The following have obvious effects. For strong reflection problems, the interface distance is generally Below), merge the preliminary decomposed atoms that interact with each other, w2 = w 11 (a​11 ,t 11 ,f 11 ,b 11 )+w 12 (a 12 ,t 12 ,f 12 ,b 12 ). At this time w2 also has w 11 and w 12 The parameter information of w2 is taken into account and enters the matching pursuit iteration process as a whole. The reason for this is that w2 as a whole re-satisfies the assumption that the matching atoms are orthogonal to the seismic traces. At this time, the optimal solution of the matching pursuit iteration is consistent with the true solution.

[0029] The data reconstruction module S3 is used to adjust the strong reflection atoms and their adjacent atoms in the matching atom library w3 to achieve the purpose of removing the strong reflection amplitude:

[0030] S31: determining the strong reflection atom in the matching atom library w3 corresponding to the strong reflection amplitude according to the strong reflection model established in the model decomposition module S2;

[0031] S32: Remove the strong reflection atom and adjust the amplitude coefficient of its neighboring atoms so that the strong reflection atom does not cause shielding effect on the neighboring atoms;

[0032] S33: The linearly combined adjusted atomic library w3 completes the processing of eliminating strong reflection amplitude.

[0033] Further explanation, the adjustment of the matching atoms involved in S32 refers to: according to the seismic convolution theory,

[0034]

[0035] The seismic record S(t) can be expressed as the reflection coefficient r at any point n (t n ) and the corresponding time-shifted wavelet W(tt n ) plus the noise N(t). When the matching pursuit is close to the true solution, the amplitude coefficient a of the matching atom w(a,t,f,b) obtained by matching pursuit can be regarded as the unnormalized reflection coefficient. Therefore, according to the principle of seismic reflection, the reflection coefficient corresponding to the strong axis can be reduced or eliminated, and the adjacent atoms can be adjusted. Taking a three-layer structure (Z1 / Z2 / Z3) to a two-layer structure (Z1 / Z3) as an example, the change of its reflection coefficient before and after is (see Appendix Figure 3 ):

[0036]

[0037] Compared with the prior art, the present invention can achieve the following positive technical effects:

[0038] 1) Computational efficiency of the matching pursuit algorithm: ① A matching wavelet mother function with clear physical meaning is used. Compared with other wavelets, its waveform conforms to the characteristics of seismic waves and its parameters are easier to obtain, which has more advantages in describing the scale of seismic waves; ② An optimization function of the derivative of the inner product is constructed, and the gradient descent method is used as the optimization direction function. Compared with the dynamic interval global search algorithm in the prior art, the computational efficiency of the algorithm is significantly improved;

[0039] 2) Accuracy of the matching pursuit algorithm: By adopting a secondary decomposition strategy and modifying the matching atoms based on the strong reflection model, the problem of the deviation of the optimal solution of matching pursuit from the true solution in the case of multiple reflection waves is solved, which leads to incomplete and incorrect elimination of strong reflection amplitude seismic waves. In fact, it is precisely because the strong impedance difference interface is close to the reservoir that the strong amplitude shielding problem exists, and the discussion and solution to this problem are rarely seen in the existing technology;

[0040] 3) Reasonable elimination of strong reflection atoms: Based on the principle of seismic reflection, the amplitude coefficient of strong reflection atoms is eliminated, so that the adjustment result has rigorous geological significance, and the reconstructed profile still conforms to the principle of seismic reflection. The existing technology usually adopts a simple excision or a certain proportion of scaling method to solve this problem. The result obtained has destroyed the seismic reflection and cannot represent the underground reflection result after removing the strong reflection interface. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is the principle diagram and flow chart of the present invention;

[0042] Figure 2 It is a schematic diagram of removing the strong axis and a principle diagram of the method of adjusting the strong axis atoms of the present invention;

[0043] Figure 3 Numerical simulation diagram to help illustrate the deviation of the optimal solution of matching pursuit iteration from the true solution in the case of atomic interference;

[0044] Figure 4 This is a bar chart of the comprehensive interpretation conclusion of well logging of a completed well in the study area of ​​the embodiment;

[0045] Figure 5 To eliminate the seismic profile of strong reflection amplitude seismic wavefront;

[0046] Figure 6 The seismic profile after eliminating the strong reflected amplitude seismic waves. DETAILED DESCRIPTION

[0047] The following is further described in conjunction with the accompanying drawings and embodiments. The specific steps of the present invention are as follows:

[0048] Step 1: Based on the seismic horizon, obtain a seismic record S(t) within the time window to be decomposed, where t is the time corresponding to the sampling point of the seismic trace;

[0049] Step 2: Search for the amplitude a0 (not the absolute value) corresponding to the maximum amplitude of the seismic trace S(t) in absolute value and its corresponding time t0, and calculate the instantaneous frequency f0 at time t0; and search for the smaller extreme value of a0 with the same sign above and below it as a′, and calculate the matching wavelet scale factor b0,

[0050] Step 3: Construct the initial matching wavelet w0(a0, t0, f0, b0) based on the obtained parameters, and establish a neighborhood ε as the dynamic atom library D with the initial parameters, where w0 and D are

[0051]

[0052] a∈[a0-ε,a0+ε],t d ∈[t0-ε,t0+ε],f∈[f0-ε,f0+ε],b∈[b0-ε,b0+ε];

[0053] Where a0, t0, t, f0, b0 are the same as those in step 2. d ,f,b represent the amplitude coefficient, center time, main frequency and scale factor of the wavelet respectively, a,t d ,The value range of f,b is a neighborhood centered on a0,t0,f0,b0;

[0054] Step 4: Use the matching pursuit algorithm based on gradient descent to optimize w0 in the dynamic atom library D to obtain the best matching atom w1, and subtract w1 from S(t) to obtain the residual seismic trace S1(t).

[0055]

[0056] In the formula, w 1n The suffix 1 indicates the result of the first wavelet decomposition, and n indicates w 1n is the nth matching pursuit in the first wavelet decomposition, argmin represents the function value corresponding to the minimum value of the solution objective function, N is the upper limit of the number of matching pursuit operations (set to 300), S n is the residual seismic trace obtained by the nth matching pursuit, | n ,S n >′| is S n The absolute value of the derivative of the inner product with itself;

[0057] ​Step 5: Use S1(t) as the input signal for the next matching pursuit operation, and go to step 2 to continue the iteration until the energy of the residual seismic trace is less than the given threshold value or the maximum matching pursuit operation number is reached. At the end of the iteration, the first wavelet decomposition result w1(a1, t1, f1, b1) of the seismic trace S(t) is obtained;

[0058] Step 6: Based on the well logging data of the study area (such as the attached Figure 5 The geological model is shown in Figure 1. The reasons for the formation of strong reflection amplitude seismic waves and the combination of seismic waves are analyzed. It is assumed that the seismic waves at interfaces 1 and 2 are interfering due to their close distances, and the corresponding sub-wave atoms are w 1m 、w 1n (The strong reflection model in this embodiment can be summarized as three strong reflection interfaces where low-impedance sandstone and high-impedance limestone contact each other. There is interference between them, and the strong impedance interface 3 has a shielding effect on the reservoir and is removed. See the attached figure. Figure 4 );

[0059] Step 7: Based on the seismic wave recombination reflected by the geological model in step 6, the matching atoms in the interference region are merged to obtain the composite atom w2=w 1m (a 1m ,t 1m ,f 1m ,b 1m )+w 1n (a 1n ,t 1n ,f 1n ,b 1n );

[0060] Step 8: Take S(t) as the input signal of the second wavelet decomposition, perform matching pursuit iteration on the composite atom w2, obtain the optimal solution of each composite term in w2 in the merged state, and subtract w2 from S(t) to obtain the residual seismic trace TS1(t);

[0061] Step 9: Go to step 7, take TS1(t) as the input signal, perform the same processing on other areas with interference, and directly use the result of the first wavelet decomposition on areas without interference, until all the results of the first wavelet decomposition w1 are updated to the matching atom library w3(a3, t3, f3, b3);

[0062] Step 10: According to the geological model established in step 7, determine the sub-wave atom and its upper neighboring atom corresponding to the strong reflection amplitude seismic wave to be eliminated in the matching atom library w3, assuming that they are w 3m 、w 3n ;

[0063] Step 11: Let the amplitude coefficient of the wavelet atom corresponding to the strong reflection amplitude seismic wave be a′ 3m = 0, and adjust the amplitude coefficient of the adjacent atom above it to

[0064] Step 12: Linearly combine the atoms adjusted by the matching atom library w3 and add the final residual seismic trace S in step 8 N (t), reconstruct the seismic trace S′(t), and complete the processing of S(t) to eliminate the strong reflection amplitude seismic wave;

[0065] Step 13: Go to step 1, obtain the next seismic trace data S(t) and decompose and reconstruct it until all seismic traces are processed;

[0066] Step 14: Extract the seismic profiles before and after the processing, compare the changes of the profiles and planes before and after the processing, and evaluate the effect of eliminating the strong reflection amplitude seismic wave processing. Figure 5 , 6 As shown in the figure, after eliminating the strong reflection amplitude seismic wave processing, the strong amplitude corresponding to the seismic layer 1 on the profile is eliminated, and the continuity and intensity of the peak indicating the reservoir development below are restored.

Claims

1. A method for eliminating strong reflection amplitude seismic waves comprises the following specific steps: Step 1: Based on the seismic horizon, obtain a seismic record S(t) within the time window to be decomposed, where t is the time corresponding to the sampling point of the seismic trace; Step 2: Search for the amplitude a0 corresponding to the maximum absolute value of the amplitude of the seismic trace S(t) and its corresponding time t0, and calculate the instantaneous frequency f0 at time t0; and search for the smaller extreme value of a0 with the same sign above and below it as a′, and calculate the matching wavelet scale factor b0, Step 3: Construct the initial matching wavelet w0(a0, t0, f0, b0) based on the obtained parameters, and establish a neighborhood ε as the dynamic atom library D with the initial parameters, where w0 and D are a∈[a0-ε,a0+ε],t d ∈[t0-ε,t0+ε],f∈[f0-ε,f0+ε],b∈[b0-ε,b0+ε]; Where a0, t0, t, f0, b0 are the same as those in step 2. d ,f,b represent the amplitude coefficient, center time, main frequency and scale factor of the wavelet respectively, a,t d ,The value range of f,b is a neighborhood centered on a0,t0,f0,b0; Step 4: Use the matching pursuit algorithm based on gradient descent to optimize w0 in the dynamic atom library D to obtain the best matching atom w1, and subtract w1 from S(t) to obtain the residual seismic trace S1(t). S n =S n-1 -w 1n w 1n ∈D In the formula, w 1n The suffix 1 indicates the result of the first wavelet decomposition, and n indicates w 1n is the nth matching pursuit in the first wavelet decomposition, argmin represents the function value corresponding to the minimum value of the solution objective function, N is the upper limit of the number of matching pursuit operations (set to 300), S n is the residual seismic trace obtained by the nth matching pursuit, S n ,S n ′ is S n The absolute value of the derivative of the inner product with itself; Step 5: Use S1(t) as the input signal for the next matching pursuit operation, and go to step 2 to continue the iteration until the energy of the residual seismic trace is less than the given threshold value or the maximum matching pursuit operation number is reached. At the end of the iteration, the first wavelet decomposition result w1(a1, t1, f1, b1) of the seismic trace S(t) is obtained; Step 6: Establish a geological model based on the well logging data of the study area, analyze the causes of the strong reflection amplitude seismic waves and the seismic wave recombination, and assume that it reflects the seismic wave interference between interfaces 1 and 2 due to their close distances, and the corresponding sub-wave atoms are w 1m 、w 1n ; Step 7: Based on the seismic wave recombination reflected by the geological model in step 6, the matching atoms in the interference region are merged to obtain the composite atom w2=w 1m (a 1m ,t 1m ,f 1m ,b 1m )+w 1n (a 1n ,t 1n ,f 1n ,b 1n ); Step 8: Take S(t) as the input signal of the second wavelet decomposition, perform matching pursuit iteration on the composite atom w2, obtain the optimal solution of each composite term in w2 in the merged state, and subtract w2 from S(t) to obtain the residual seismic trace TS1(t); Step 9: Go to step 7, take TS1(t) as the input signal, perform the same processing on other areas with interference, and directly use the result of the first wavelet decomposition on areas without interference, until all the results of the first wavelet decomposition w1 are updated to the matching atom library w3(a3, t3, f3, b3); Step 10: According to the geological model established in step 7, determine the sub-wave atom and its upper neighboring atom corresponding to the strong reflection amplitude seismic wave to be eliminated in the matching atom library w3, assuming that they are w 3m 、w 3n ; Step 11: Let the amplitude coefficient of the wavelet atom corresponding to the strong reflection amplitude seismic wave be a3′ m = 0, and adjust the amplitude coefficient of the adjacent atom above it to Step 12: Linearly combine the atoms adjusted by the matching atom library w3 and add the final residual seismic trace S in step 8 N (t), reconstruct the seismic trace S′(t), and complete the processing of S(t) to eliminate the strong reflection amplitude seismic wave; Step 13: Go to step 1, obtain the next seismic trace data S(t) and decompose and reconstruct it until all seismic traces are processed; Step 14: Extract the seismic profiles before and after the processing, compare the changes of the profiles and planes before and after the processing, and evaluate the processing effect of eliminating the strong reflection amplitude seismic waves.

Images