Method for matching pursuit strong coal seam reflection elimination based on online dictionary learning

Abstract

The application provides a matching pursuit strong coal seam reflection elimination method based on online dictionary learning, comprising the following steps: step 1, constructing a single-phase wavelet library based on online dictionary learning according to the seismic waveform characteristics of a target layer; step 2, performing matching pursuit signal decomposition on the target layer reflection signal based on the single-phase wavelet library of online dictionary learning; step 3, subtracting the extracted coal seam reflection from the original data to obtain a reflection profile after coal removal; and step 4, extracting a prestack gather coal seam reflection characteristic wavelet according to the coal seam reflection travel time, then eliminating or weakening the coal seam reflection, and performing prestack time migration on the coal-removed prestack data after deconvolution. The matching pursuit strong coal seam reflection elimination method based on online dictionary learning can effectively eliminate the influence of strong interference, thereby verifying the effectiveness of the technical scheme for eliminating strong coal seam reflection and improving the effective signal energy and resolution.

Description

Technical Field

[0001] This invention relates to the field of oilfield development technology, and in particular to a matching pursuit method for eliminating strong coal seam reflections based on online dictionary learning. Background Technology

[0002] In the field of seismic data processing, the sparse representation of seismic records has become a hot topic of research in recent years. By constructing an overcomplete atomic library, seismic records are decomposed into a linear superposition of the best-matching atoms in the library, achieving an adaptive sparse representation. The matching pursuit algorithm, with its simplicity and flexibility, is widely used. It projects the signal onto the optimal atom through a greedy search of the overcomplete atomic library. To ensure sparsity, the atomic library is generally large, and all atoms participate in the inner product operation in each iteration, resulting in low computational efficiency. Furthermore, due to the overcomplete nature of the atomic library, the atoms obtained in each iteration cannot be guaranteed to be orthogonal to previously searched atoms, leading to the introduction of new residual components, reducing convergence speed, and easily causing over-matching, thus compromising the sparsity of the decomposition results. Therefore, how to select a suitable decomposition method to improve decomposition accuracy and speed is a problem that urgently needs to be solved in this field. Regarding improving decomposition accuracy, the traditional approach is to expand the size of the atomic library as much as possible. While this method is beneficial to the accuracy of the decomposition results, the increased size of the atomic library undoubtedly increases the burden of the decomposition process, expands the atom search range, and reduces work efficiency. In recent years, more and more researchers have focused on dictionary learning to improve computational efficiency. The goal of dictionary learning is to train a dictionary using a pre-given set of signals and then use this dictionary to achieve a sparse representation of any given signal. Research results show that learning methods based on overcomplete redundant dictionaries have significant advantages over traditional fixed dictionaries and have achieved many ideal applications in image processing. However, traditional overcomplete redundant dictionaries are based on mathematical formulas of analytical expressions and cannot optimally represent the local features of signals. Therefore, how to obtain a dictionary that can effectively represent the local features of signals has become a research hotspot. Against this backdrop, dictionary learning algorithms have emerged, but traditional dictionary learning methods share a common drawback—extremely high computational cost, making them unsuitable for large-scale sample training. In dictionary learning problems, the larger the training samples, the richer the learned features and the better the sparse representation capability of the dictionary. Therefore, to address the limitation of training sample size, the Online Dictionary Learning (ODL) algorithm was proposed (Mairal et al., 2010). This algorithm preprocesses the data to be processed, cuts it into a series of small samples, trains the dictionary on each small sample, and performs sparse encoding based on the dictionary generated in the previous iteration. This algorithm can significantly improve the convergence speed and can easily achieve dictionary learning on the scale of millions of samples.

[0003] Conventional matching pursuit algorithms require scanning the entire atom library in each match, but the atom library is overcomplete, resulting in a computationally intensive and slow scanning process. To reduce the scanning workload in matching, this invention first constructs a wavelet library using an online dictionary, and then employs the Complex Domain Fast Matching Pursuit Decomposition (CFMP) method to improve computational efficiency. The CFMP method requires obtaining complex seismic records from real seismic records, extracting prior information (such as the time, instantaneous frequency, and instantaneous phase at the maximum amplitude envelope) from the complex signal, and then scanning a small portion of the wavelet library under the constraints of the prior information to obtain the best match. After reconstructing the complex signal, the result is returned to the real domain, thereby reducing the computational workload in matching to a certain extent and improving the algorithm efficiency.

[0004] In seismic records, the locations of coal seams, unconformities, and carbonate rock formations typically exhibit strong reflection characteristics. This strong reflection easily masks the reflection information of adjacent sand bodies, making such reservoirs unidentifiable. Therefore, the recovery of effective signals under strong reflection backgrounds, as well as the evaluation of high-quality reservoirs and the detection of fluid-bearing properties, have become hot and challenging research topics for many scholars both domestically and internationally. For example, during the deposition of coal-bearing strata, coal seams and target reservoirs are often superimposed, causing the effective sand body reflection information to be submerged by the coal seam reflection; geological sedimentary discontinuities can cause strong reflections in unconformity fracture zones, thus shielding the effective sand body reflections near the unconformity.

[0005] Chinese patent application CN201410171953.4 discloses a method for removing strong reflection signals from coal seams, comprising the following steps: selecting an atomic library; reconstructing the original seismic signal using a matching pursuit algorithm; obtaining strong reflection signals from the coal seams using the matching pursuit algorithm; subtracting the obtained strong reflection signals from the original seismic signal, with the remaining signal being the desired de-reflected coal seam signal. This invention reconstructs the original seismic signal using a matching pursuit algorithm, resulting in a simple reconstruction process and more ideal reconstruction results. Simultaneously, the strong reflection signals from the coal seams are obtained through matching and calculation using the matching pursuit algorithm, a very simple process. The de-reflected coal seam signal obtained by the method described herein effectively highlights the weak reflection information of sandstone and mudstone, facilitating reservoir prediction research and providing more reliable and effective results, thus providing a basis for reservoir prediction and oil and gas detection.

[0006] Chinese patent application CN201510379767.4 discloses a method for eliminating strong reflection amplitudes of seismic marker layers based on empirical mode decomposition. This method utilizes empirical mode decomposition, combined with correlation function analysis and well logging data, to select the IMF component from the original seismic data that best reflects the strong amplitude thin-layer characteristics of coal seams and source rocks. For the selected main IMF component, its maximum energy point is located, and the dominant frequency of the IMF component signal is estimated. Based on this, the time thickness of the top and bottom thin layers of the strong amplitude thin layers of coal seams and source rocks reflected in the IMF component is determined. Strong amplitude suppression is applied to the IMF signal components within this time thickness range, while the remaining IMF components are retained unchanged. Finally, all processed IMF component signals are reconstructed to eliminate strong reflection amplitudes of seismic marker layers in the seismic trace signal, enhance subtle changes in formation signals under coal-bearing strata coexistence, and strengthen the weak hydrocarbon response in adjacent areas.

[0007] Chinese patent application CN201910675744.6 discloses a method for separating strong reflections in thin coal seams based on phase-shift wavelet, a computer storage medium, and a computer device. This method first studies the reflection modes of thin coal seams, determining the seismic reflection modes. Then, it selects a 90° phase-shift wavelet as the matching wavelet for strong reflection separation. By analyzing the relationship between strong reflection energy and total energy, it determines the strong reflection separation coefficient, thus enabling more reasonable and effective separation of strong reflections in thin coal seams. In practical data applications, this improved algorithm has achieved good results.

[0008] The above-mentioned existing technologies are all significantly different from the present invention and have failed to solve the technical problem to be solved. Therefore, a new method for eliminating strong coal seam reflections based on online dictionary learning has been invented. Summary of the Invention

[0009] The purpose of this invention is to provide a matching pursuit method for eliminating strong coal seam reflections based on online dictionary learning, which is developed according to the needs of practical applications and is designed for the characteristics of non-stationary signals.

[0010] The objective of this invention can be achieved through the following technical measures: a matching pursuit method for eliminating strong coal seam reflections based on online dictionary learning, which includes:

[0011] Step 1: Based on the seismic waveform characteristics of the target layer, construct a single-phase wavelet library using online dictionary learning;

[0012] Step 2: Based on the online dictionary learning of the single-phase wavelet library, perform matched tracking signal decomposition on the target layer reflection signal;

[0013] Step 3: Subtract the extracted coal seam reflection from the original data to obtain the reflection profile after coal removal;

[0014] Step 4: Extract the characteristic wavelet of coal seam reflection from the pre-stack gather based on the coal seam reflection travel time, then eliminate or weaken the coal seam reflection, and perform pre-stack time offset of the pre-stack data after reaction correction.

[0015] The objective of this invention can be achieved through the following technical measures:

[0016] In step 1, based on the seismic waveform characteristics of the target layer, online dictionary learning technology is used to extract wavelets that can reflect the local waveform characteristics of the target layer, thus forming a single-phase wavelet library.

[0017] In step 1, assume a finite training set of the signal X = [x1, ..., x2]. N ]exist In this context, the dictionary learning algorithm achieves its goal by optimizing the experience cost function:

[0018]

[0019] In the formula, D is A dictionary, where each column consists of basis vectors, l(x i D) is the loss function; if the dictionary D can well represent the signal X, then the loss function l(x, D) is the loss function. i D) should be very small; l(x) i D) is considered the optimal solution to the l1 sparse coding problem:

[0020]

[0021] Here, λ is the regularization parameter, and the above problem is also a solution of basis pursuit. The sparse solution of α is obtained through L1 penalty, but there is no analytical relationship between the value of λ and the corresponding effective sparsity ||α||0.

[0022] In step 2, a single-phase wavelet library is learned based on an online dictionary, and a fast matching pursuit decomposition algorithm in the complex domain is used to perform matching pursuit signal decomposition on the target layer reflection signal.

[0023] In step 2, the Fast Complex Domain Matching Pursuit (CFMP) method obtains complex seismic records based on real seismic records. It extracts prior information from the complex signal, including the time, instantaneous frequency, and instantaneous phase at the maximum amplitude envelope. Then, under the constraint of the prior information, it scans a small portion of the atom library to obtain the best match. After reconstructing the complex signal, the result is returned to the real domain, thereby reducing the amount of computation in the matching process and improving the algorithm efficiency to a certain extent.

[0024] In step 2, the formulas for calculating the amplitude envelope, instantaneous frequency, and instantaneous phase of the complex signal are as follows:

[0025]

[0026]

[0027]

[0028] Where f(t) is the real part of the signal, f * (t) represents the imaginary part of the signal; based on this prior information, the matching atom of the response can be quickly found in the dictionary.

[0029] In step 2, let H be the Hilbert space, and G = {g r |||g r ||=1,r∈M} are basic functions in H, and each g r All signals are standardized, and M is the total number of time-frequency atoms in the time-frequency atom library. Assuming a complex signal F ∈ H, the goal of the CFMP algorithm is to represent F as a linear combination of complex time-frequency atoms selected from G. The complex signal is obtained through the Hilbert transform.

[0030] F(t)=f(t)+i·H(f(t)) (6)

[0031] In the formula: f(t) is the real signal; the imaginary part H(f(t)) is the Hilbert transform of the real signal, which is essentially a 90° phase shift of the real part.

[0032] In step 2, assuming g(t) is a time-frequency atom and H(g(t)) represents the Hilbert transform of the time-frequency atom, the expression for the complex time-frequency atom is:

[0033] G(t)=g(t)+i·H(g(t)) (7)

[0034] Matching and tracking from the initial residual R 0 Starting with F = F, first calculate the instantaneous amplitude envelope, instantaneous frequency, and instantaneous phase of the residual. Then, assign the time, instantaneous frequency, and instantaneous phase at the maximum value of the amplitude envelope to t. j f j and Through n matching calculations, we obtain:

[0035]

[0036] R n F is a complex signal; where... It is a complex time-frequency atom, and R n F is the residual quantity. Let represent the inner product of the signal and the residual; therefore, we have:

[0037]

[0038] In order to make ||R n+1 If F|| is minimized, then the selected complex wavelet needs to be minimized. Satisfaction makes The optimal selection criterion for the largest, or best, atom is:

[0039]

[0040] The time-frequency atom selected by equation (10) can minimize the residual obtained in each iteration; the best matching wavelet atom can be calculated by equation (10), that is, the optimal amplitude, time shift, main frequency and phase parameters can be calculated.

[0041] In step 3, for the post-stack data, single-phase wavelets characterizing coal seam reflection are extracted along the layers based on the coal seam reflection characteristics. Multiple single-phase wavelets of the coal seam are superimposed to form a complete coal seam reflection characteristic. The extracted coal seam reflection is subtracted from the original data to obtain the reflection profile after coal removal.

[0042] In step 4, for the pre-stack gather data after dynamic correction, the characteristic wavelet of coal seam reflection in the pre-stack gather can be extracted based on the coal seam reflection travel time. Then, the coal seam reflection is eliminated or weakened. After dynamic correction, the pre-stack time shift of the pre-stack data is performed to effectively improve the reflection energy of the upper and lower reservoirs of the coal seam while eliminating the coal seam response.

[0043] The matching pursuit method for eliminating strong coal seam reflections based on online dictionary learning in this invention can process pre-stack and post-stack data separately. Based on an online dictionary learning algorithm, a wavelet library is established by learning from actual seismic data. Then, based on the learned wavelet library, a fast matching pursuit decomposition algorithm in the complex domain is used to decompose the seismic signal into independent wavelets that can express the local characteristics of the seismic signal. For post-stack strong coal seam reflections, the characteristic phase wavelet of the coal seam reflection can be extracted based on the coal seam reflection characteristics, along the interpreted coal seam horizon, using a single phase. Because it is difficult to completely extract the wavelet based on the complete wavelet waveform and it is easily affected by coal seam interlayers, single-phase horizon control extraction is chosen to increase extraction accuracy and improve applicability and stability. For pre-stack strong coal seam reflections, the pre-stack gathers are first subjected to dynamic correction processing. Then, with the interpreted coal seam horizon as the center time, the strong reflection wavelet is extracted and the influence of strong coal seam reflections is eliminated or weakened. Next, the gathers without coal seam reflections are subjected to reaction correction processing. Finally, pre-stack time migration processing is performed, and the resulting new migration profile is the new data volume with the strong coal seam response eliminated.

[0044] The wavelet library construction method based on online dictionary learning of this invention effectively reduces the redundancy of the wavelet library by learning wavelet characteristics from actual seismic data. The constructed wavelet library is closer to the actual data and can truly reflect the local characteristics of the actual data. Based on this, fast complex domain matching pursuit is used to effectively improve the accuracy and speed of wavelet decomposition. Then, along the coal seam reflection horizon, the corresponding single-phase energy of the coal seam is selectively extracted, and all independent single-phase energies are superimposed to form a multi-layered coal seam extraction technology that is efficient, accurate, and stable. Compared with conventional methods, this invention, utilizing the above technical solution, has the following advantages:

[0045] (1) This technical solution provides an efficient and reliable online dictionary learning algorithm, which learns the local features of seismic signals based on the target layer and constructs a wavelet library. Compared with the traditional redundant wavelet library constructed using analytical expressions, the wavelets extracted by online dictionary learning have higher accuracy and clearer geological meaning.

[0046] (2) This technical solution provides a wavelet library based on online dictionary learning combined with a fast complex domain matching and tracking algorithm, which lays the foundation for efficient and rapid seismic wavelet decomposition and accurate feature extraction.

[0047] (3) The post-stack data provided by this technical solution can selectively extract coal seam reflection along the layer using single phase extraction, which can target multiple complex coal seam reflections and effectively avoid the influence of interlayers.

[0048] (4) When processing pre-stack gather data, this technical solution first performs dynamic correction processing on the pre-stack gather data, then extracts the coal seam reflection characteristics based on the coal seam reflection travel time, eliminates or weakens the coal seam reflection, and performs pre-stack time offset of the coal seam pre-stack data after dynamic correction. Under the premise of eliminating coal seam response, it effectively improves the reflection energy of the upper and lower reservoirs of the coal seam. Attached Figure Description

[0049] Figure 1 This is a flowchart of a specific embodiment of the matching and tracking method for eliminating strong coal seam reflections based on online dictionary learning according to the present invention;

[0050] Figure 2 This is a schematic diagram illustrating the reconstruction of data based on reflected energy and frequency in a specific embodiment of the present invention;

[0051] Figure 3 This is a schematic diagram of the peak portion of the coal seam response extracted by wavelet modulation and extracting the in-phase axis separately according to the peak and valley energy in a specific embodiment of the present invention.

[0052] Figure 4 This is a schematic diagram of the valley portion of the coal seam response extracted by wavelet modulation and extracting the in-phase axis separately according to the peak and valley energy in a specific embodiment of the present invention.

[0053] Figure 5 This is a schematic diagram of the peak portion of the coal seam response extracted by wavelet modulation and extracting the in-phase axis separately according to the peak and valley energy in a specific embodiment of the present invention.

[0054] Figure 6 This is a schematic diagram illustrating the combination of the three elements and the extraction of the peak-valley combination of the coal seam response in a specific embodiment of the present invention.

[0055] Figure 7 In a specific embodiment of the present invention, the upper figure is the original seismic profile, and the lower figure is a schematic diagram of the seismic profile after the coal seam energy has been attenuated by 70%.

[0056] Figure 8 This is a schematic diagram of the pre-stack gather before eliminating the influence of the coal seam in a specific embodiment of the present invention;

[0057] Figure 9 This is a schematic diagram of the pre-stack gather after eliminating the influence of coal seams in a specific embodiment of the present invention;

[0058] Figure 10 This is a schematic diagram of the offset profile before eliminating the influence of the coal seam in a specific embodiment of the present invention;

[0059] Figure 11 This is a schematic diagram of the offset profile after eliminating the influence of the coal seam in a specific embodiment of the present invention. Detailed Implementation

[0060] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0061] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, and / or combinations thereof.

[0062] The following are several specific embodiments of the application of the present invention.

[0063] Example 1

[0064] In a specific embodiment 1 of the present invention, such as Figure 1 As shown, Figure 1This is a flowchart of the matching pursuit method for eliminating strong coal seam reflections based on online dictionary learning according to the present invention. The method includes the following steps:

[0065] (1) Wavelet library construction technology based on online dictionary learning: mainly targeting the seismic waveform characteristics of the target layer, using online dictionary learning technology to extract wavelets that can reflect the local waveform characteristics of the target layer, and constructing a unit phase wavelet library;

[0066] (2) Complex Domain Fast Matching Pursuit Decomposition and Reconstruction Based on Online Dictionary Learning Wavelet Library: Based on the online dictionary learning wavelet library, the seismic signal is decomposed using the complex domain fast matching pursuit technique.

[0067] (3) For the post-stack data, extract single-phase wavelets along the coal seam reflection characteristics; superimpose multiple single-phase wavelets of the coal seam to form a complete coal seam reflection characteristic; subtract the extracted coal seam reflection from the original data to obtain the reflection profile after coal removal.

[0068] (4) For pre-stack gather data after dynamic correction, coal seam reflection characteristics can be extracted based on coal seam reflection travel time, and then coal seam reflection can be eliminated or weakened. After dynamic correction, the pre-stack time shift of the pre-stack data is removed. Under the premise of eliminating coal seam response, the reflection energy of the upper and lower reservoirs of the coal seam can be effectively improved, laying the foundation for subsequent reservoir prediction.

[0069] This invention is based on the theory of matching pursuit seismic signal decomposition. It utilizes online dictionary learning technology to construct a wavelet library for the target layer. Through a complex domain fast matching pursuit algorithm, the seismic signal of the target layer is decomposed. Then, through layer-controlled selective reconstruction technology, the reflection characteristics of the coal seam are extracted by phase and then eliminated to reduce their impact on the characteristic response of the reservoir above and below the coal seam.

[0070] Example 2

[0071] In a specific embodiment 2 of the present invention, the matching pursuit method for eliminating strong coal seam reflections based on online dictionary learning specifically includes the following steps:

[0072] Step 1, Online Dictionary Learning

[0073] There are two types of overcomplete dictionaries: one is based on fixed character types (overcomplete DCT, Contourlet, wavelet dictionary), and the other is overcomplete dictionaries learned from a training set. There are many methods for using overcomplete dictionaries generated through learning, such as MOD-type dictionary learning methods, K-SVD-type dictionary learning methods, and online character type learning methods.

[0074] The online dictionary learning algorithm was proposed by Mairal et al. in 2009. It assumes a finite training set of signals X = [x1, ..., x2]. N ]exist In this context, traditional dictionary learning algorithms optimize the experience cost function:

[0075]

[0076] In the formula, D is A dictionary's basis vectors are composed of its columns. If the dictionary D is "good" representing signal X, then l(x) = ... i ,D) should be very small.

[0077] l(x) i D) is considered the optimal solution to the l1 sparse coding problem:

[0078]

[0079] Here, λ is the regularization parameter, and the problem described above is essentially a basis pursuit or lasso problem. The sparse solution for α is obtained through L1 penalty, but there is no analytical relationship between the value of λ and the corresponding effective sparsity ||α||0. To prevent the dictionary D from becoming arbitrarily large (which would make the value of α arbitrarily small), it is usually listed... The norm constraint is less than or equal to 1. Let C be called the convex matrix set used to prove this constraint:

[0080] However, the experience cost f n (D) The minimization problem is not convex with respect to the dictionary D. For D and the coefficients of the sparse coding α = [α1,L,α] n They can be rewritten as a joint optimization problem, but it is not jointly convex; it is only convex when one of the two variables, D and α, is fixed.

[0081]

[0082] However, it is generally undesirable to consider the experience cost f n (D) minimization, but rather the minimization of expected cost:

[0083]

[0084] Step 2, Fast Matching and Pursuit Decomposition in Complex Fields

[0085] Conventional matching pursuit algorithms require scanning the entire atomic library in each match. However, the atomic library is overcomplete, resulting in a large computational burden and slow computation speed. To reduce the scanning workload in matching, the Complex Domain Fast Matching Pursuit Decomposition (CFMP) method can be used to improve computational efficiency. The CFMP method requires obtaining complex seismic records from real seismic records, extracting prior information (such as the time, instantaneous frequency, and instantaneous phase at the maximum amplitude envelope) from the complex signal, and then scanning a small portion of the atomic library under the constraints of the prior information to obtain the best match. After reconstructing the complex signal, the result is returned to the real domain, thereby reducing the computational burden in matching to a certain extent and improving the algorithm efficiency. Let H be a Hilbert space, satisfying... G={g r |||g r ||=1, r∈Γ} are basic functions in H, and each g r All have been standardized, and Γ is the index. The primitive function G is a time-frequency atom library, where each g... r Each time-frequency atom is a single element. Through iterative matching calculations on the complex domain signal, each match yields a complex time-frequency atom. Assuming a complex signal F ∈ H, the goal of the CFMP algorithm is to represent F as a linear combination of complex time-frequency atoms selected from G. The complex signal can be obtained through the Hilbert transform:

[0086] F(t)=f(t)+i·H(f(t)) (5)

[0087] In the formula: f(t) is the real signal; the imaginary part H(f(t)) is the Hilbert transform of the real signal, which is essentially a 90° phase shift of the real part.

[0088] Similarly, assuming g(t) is a time-frequency atom, and H(g(t)) represents the Hilbert transform of the time-frequency atom, then the expression for the complex time-frequency atom is:

[0089] G(t)=g(t)+i·H(g(t)) (6)

[0090] Matching and tracking from the initial residual R 0 Starting with F = F, first calculate the instantaneous amplitude envelope, instantaneous frequency, and instantaneous phase of the residual. Then, assign the time, instantaneous frequency, and instantaneous phase at the maximum value of the amplitude envelope to t. j f j and Through n matching calculations, we obtain:

[0091]

[0092] Rn F is a complex signal. It is a complex time-frequency atom, and R n F is the residual quantity. This represents the inner product of the signal and the residual. It is easy to derive... Orthogonal to R n+1 F, therefore we have:

[0093]

[0094] In order to make ||R n+1 If F|| is minimized, then the selected complex wavelet... Make The optimal selection criterion for the largest, or best, atom is:

[0095]

[0096] The time-frequency atom energy selected by equation (2-5) minimizes the residual obtained in each iteration. When the local properties of the signal are exactly the same as or completely match those of the matching atom, the projection value of the complex signal onto the atom... If it is a real number, then we have Therefore, the matching condition is improved to:

[0097]

[0098] The optimal matching wavelet atom can be calculated using equation (2-6), which yields the optimal time shift, dominant frequency, and phase parameters. To achieve the best matching effect, this paper constructs a time-frequency atom dictionary W within a certain neighborhood of the previously calculated parameters (delay, dominant frequency, and phase) when building the time-frequency atom library, where W = {W γ |||W γ ||=1}, where: the index set γ is an element in the set Γ, and all complex time-frequency atoms are normalized. After applying the matching condition (2-6), the result of the first matching is:

[0099]

[0100] Among them, R 1 F is the complex residual from the first iteration, which can be directly used in the next iteration, saving the computational burden of the Hilbert transform. The calculation process is the same as the previous one. By setting the number of iterations or an error threshold, the final result of complex matching pursuit is expressed as:

[0101]

[0102] Due to various noises and other interference factors in the signal, the matching parameters obtained using instantaneous information contain some errors. This paper uses the least squares method in the complex domain to correct the obtained matching parameters, so that the complex Ricker wavelet determined by the matching parameters can more closely approximate the local information of the seismic signal. Let R... n F = A j Wγ+R n+1 F, where W γ To determine the obtained complex time-frequency atom using the above instantaneous parameters, A j =|A j |exp(i·θ j ) represents the correction term for the wavelet atom, which includes amplitude correction |A j |With phase correction θ j To ensure that the corrected atomic wavelet best matches the seismic signal, ||R n+1 Find the minimum value of F|| using the least squares method over the complex field:

[0103]

[0104] In the formula: W γ The conjugate transpose of ; ε is the damping factor; I is the identity matrix.

[0105] A obtained through equation (2-9) j After obtaining the amplitude and phase correction terms, the matched wavelet atom becomes G. γ =|A j |W γ (t), where the matching parameter is φ j =α j +θ j The phase after correction. The optimal matched wavelet atom can be calculated from equation (2-9), that is, the optimal time shift, dominant frequency, phase, and other parameters are calculated. Prior information of the complex signal is obtained, including: amplitude envelope, instantaneous frequency, and instantaneous phase. The formulas for calculating the amplitude envelope, instantaneous frequency, and instantaneous phase of the complex signal are as follows:

[0106]

[0107]

[0108]

[0109] Where f(t) is the real part of the signal, f * (t) represents the imaginary part of the signal. The amplitude coefficient after n iterations is:

[0110]

[0111] Example 3

[0112] In a specific embodiment 3 of the present invention, the technology of the present invention was applied to actual data from a certain oil field. Figure 2 A schematic diagram of data reconstruction based on reflected energy and frequency. Figure 3 The bottom image shows the original seismic profile, and the top image shows the single-phase reflection characteristic profile of the coal seam reflection peak extracted along the layer. It can be seen that the method described in this patent can extract the coal seam reflection along the layer by phase. Figure 4 The bottom image shows the original seismic profile, and the top image shows the single-phase reflection characteristic profile of the coal seam reflection wave valley extracted along the layer. It can be seen that the method described in this patent can extract the coal seam reflection along the layer by phase. Figure 5 The bottom image shows the original seismic profile, and the top image shows the single-phase reflection characteristic profile of the coal seam reflection peak extracted along the layer. It can be seen that the method described in this patent can extract the coal seam reflection along the layer by phase. Figure 6 Below is the original seismic profile, and above is the coal seam reflection formed by combining the extracted different phase data. It can be seen that the method described in this patent can effectively extract coal seam reflection and avoid interlayer reflection in the coal seam. Figure 7 The bottom image shows the original seismic profile, while the top image shows the seismic profile after eliminating strong coal seam reflections. As can be seen from the profiles, the energy reflected from the upper and lower parts of the coal seam is relatively stronger, and the interlayers in the coal seam are also highlighted. Figure 8 This is the pre-stack gather before the coal seam influence is eliminated, where the coal seam reflection is strong. Figure 9 The pre-stack gather after eliminating the influence of the coal seam shows that the reflection of strong coal seams is significantly weakened. Figure 10 It is an offset profile before the influence of the coal seam is eliminated. Due to the strong reflection of the coal seam, it has a significant impact on the upper and lower reflective layers of the coal seam. Figure 11 This is the offset profile after eliminating the influence of the coal seam. Due to the elimination of strong coal seam reflections, the imaging effect of the upper and lower reflection layers, which were originally affected by strong coal seam reflections, is significantly improved. The above results demonstrate that, based on the wavelet library constructed using online dictionary learning, the techniques of fast matching pursuit decomposition in the complex domain and selective single-phase reconstruction along the layer can effectively eliminate the influence of strong interference, thus verifying the effectiveness of this technical solution in eliminating strong coal seam reflections and improving effective signal energy and resolution.

[0113] In summary, the strong coal seam reflection elimination method based on fast matching pursuit decomposition using online dictionary learning, as described in this invention, is applied to seismic data processing. It constructs a wavelet library based on online dictionary learning and then utilizes a fast matching pursuit decomposition method based on this online wavelet library in the complex domain to eliminate strong coal seam reflections. This matching pursuit method for eliminating strong coal seam reflections, based on the wavelet library constructed using online dictionary learning, effectively eliminates the influence of strong interference through fast matching pursuit decomposition in the complex domain and selective single-phase reconstruction along the layer. This verifies the effectiveness of this technical solution in eliminating strong coal seam reflections and improving effective signal energy and resolution.

[0114] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0115] Except for the technical features described in the specification, all other technologies are known to those skilled in the art.

Claims

1. A matching pursuit method for eliminating strong coal seam reflections based on online dictionary learning, characterized in that, This matching-tracking method for eliminating strong coal seam reflections, based on online dictionary learning, includes: Step 1: Based on the seismic waveform characteristics of the target layer, construct a single-phase wavelet library using online dictionary learning; Step 2: Based on the online dictionary learning of the single-phase wavelet library, perform matched tracking signal decomposition on the target layer reflection signal; Step 3: Subtract the extracted coal seam reflection from the original data to obtain the reflection profile after coal removal; Step 4: Extract the characteristic wavelet of coal seam reflection from the pre-stack gather based on the coal seam reflection travel time, then eliminate or weaken the coal seam reflection, and perform pre-stack time offset of the pre-stack data after reaction correction.

2. The method for eliminating strong coal seam reflections based on online dictionary learning according to claim 1, characterized in that, In step 1, based on the seismic waveform characteristics of the target layer, online dictionary learning technology is used to extract wavelets that can reflect the local waveform characteristics of the target layer, thus forming a single-phase wavelet library.

3. The method for eliminating strong coal seam reflections based on online dictionary learning according to claim 2, characterized in that, In step 1, assume a finite training set X = [x 1,…, x N ]exist In this context, the dictionary learning algorithm achieves its goal by optimizing the experience cost function: In the formula, D is A dictionary, where each column consists of basis vectors, l(x i D) is the loss function; if the dictionary D can well represent the signal X, then the loss function l(x, D) is the loss function. i D) should be very small; l(x) i D) is considered the optimal solution to the l1 sparse coding problem: Here, λ is the regularization parameter, and the above problem is also a solution of basis pursuit. The sparse solution of α is obtained through L1 penalty, but there is no analytical relationship between the value of λ and the corresponding effective sparsity ||α||0.

4. The method for eliminating strong coal seam reflections based on online dictionary learning according to claim 1, characterized in that, In step 2, a single-phase wavelet library is learned based on an online dictionary, and a fast matching pursuit decomposition algorithm in the complex domain is used to perform matching pursuit signal decomposition on the target layer reflection signal.

5. The method for eliminating strong coal seam reflections based on online dictionary learning according to claim 4, characterized in that, In step 2, the Fast Complex Domain Matching Pursuit (CFMP) method obtains complex seismic records based on real seismic records. It extracts prior information from the complex signal, including the time, instantaneous frequency, and instantaneous phase at the maximum amplitude envelope. Then, under the constraint of the prior information, it scans a small portion of the atom library to obtain the best match. After reconstructing the complex signal, the result is returned to the real domain, thereby reducing the amount of computation in the matching process and improving the algorithm efficiency to a certain extent.

6. The method for eliminating strong coal seam reflections based on online dictionary learning according to claim 5, characterized in that, In step 2, the formulas for calculating the amplitude envelope, instantaneous frequency, and instantaneous phase of the complex signal are as follows: Where f(t) is the real part of the signal, f * (t) represents the imaginary part of the signal; based on this prior information, the matching atom of the response can be quickly found in the dictionary.

7. The method for eliminating strong coal seam reflections based on online dictionary learning according to claim 6, characterized in that, In step 2, let H be the Hilbert space, and G = {g r |||g r ||=1,r∈M} are basic functions in H, and each g r All signals are standardized, and M is the total number of time-frequency atoms in the time-frequency atom library. Assuming a complex signal F ∈ H, the goal of the CFMP algorithm is to represent F as a linear combination of complex time-frequency atoms selected from G. The complex signal is obtained through the Hilbert transform. F(t)=f(t)+i·H(f(t)) (6) In the formula: f(t) is the real signal; the imaginary part H(f(t)) is the Hilbert transform of the real signal, which is essentially a 90° phase shift of the real part.

8. The method for eliminating strong coal seam reflections based on online dictionary learning according to claim 7, characterized in that, In step 2, assuming g(t) is a time-frequency atom and H(g(t)) represents the Hilbert transform of the time-frequency atom, the expression for the complex time-frequency atom is: G(t)=g(t)+i·H(g(t)) (7) Matching and tracking from the initial residual R 0 Starting with F = F, first calculate the instantaneous amplitude envelope, instantaneous frequency, and instantaneous phase of the residual. Then, assign the time, instantaneous frequency, and instantaneous phase at the maximum value of the amplitude envelope to t. j f j and Through n matching calculations, we obtain: R n F is a complex signal; where... It is a complex time-frequency atom, and R n F is the residual quantity. Let represent the inner product of the signal and the residual; therefore, we have: In order to make ||R n+1 If F|| is minimized, then the selected complex wavelet is required. Satisfaction makes The optimal selection criterion for the largest, or best, atom is: The time-frequency atom selected by equation (10) can minimize the residual obtained in each iteration; the best matching wavelet atom can be calculated by equation (10), that is, the optimal amplitude, time shift, main frequency and phase parameters can be calculated.

9. The method for eliminating strong coal seam reflections based on online dictionary learning according to claim 1, characterized in that, In step 3, for the post-stack data, single-phase wavelets characterizing coal seam reflection are extracted along the layers based on the coal seam reflection characteristics. Multiple single-phase wavelets of the coal seam are superimposed to form a complete coal seam reflection characteristic. The extracted coal seam reflection is subtracted from the original data to obtain the reflection profile after coal removal.

10. The method for eliminating strong coal seam reflections based on online dictionary learning according to claim 1, characterized in that, In step 4, for the pre-stack gather data after dynamic correction, the characteristic wavelet of coal seam reflection in the pre-stack gather can be extracted based on the coal seam reflection travel time. Then, the coal seam reflection is eliminated or weakened. After dynamic correction, the pre-stack time shift of the pre-stack data is performed to effectively improve the reflection energy of the upper and lower reservoirs of the coal seam while eliminating the coal seam response.

Images