A parameter-adaptive pre-stack probabilistic inversion method, equipment, and medium for seismic inversion.
By constructing a generalized hybrid prior model and adaptively updating parameters, the problem that prior distribution in Bayesian inversion is difficult to characterize complex geological backgrounds is solved, and higher accuracy in the inversion of underground medium parameters is achieved.
Patent Information
- Application Number
- CN202611297410.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-25
- Publication Date
- 2026-11-17
AI Technical Summary
In existing Bayesian seismic inversion methods, the construction of prior distributions is mostly based on the Gaussian distribution assumption, which makes it difficult to truly characterize the non-Gaussian and multimodal distribution characteristics of subsurface media under complex geological backgrounds. This results in large errors in the inversion results, affecting the reliability and accuracy of lithology identification and reservoir prediction.
We employ a parameter-adaptive pre-stack probabilistic inversion method for seismic inversion. By constructing a generalized hybrid prior model, we adaptively update the hybrid component number and standard deviation parameters. Combining the likelihood function and posterior probability, we dynamically adjust the prior model to adapt to complex geological conditions.
It improves the adaptability and accuracy of prior modeling, enabling it to more accurately describe the actual statistical distribution characteristics of underground medium parameters under complex geological conditions, and enhances the accuracy and stability of inversion results.
Smart Images

Figure CN123112992A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of seismic exploration technology, and in particular to a parameter-adaptive pre-stack probabilistic inversion method, equipment, and medium for seismic exploration. Background Technology
[0002] Seismic elastic parameter inversion is a core issue in seismic exploration and hydrocarbon reservoir characterization. Elastic parameters such as P-wave velocity, S-wave velocity, and density directly reflect the properties of subsurface rocks and are crucial foundations for lithology identification, hydrocarbon reservoir evaluation, and geological interpretation. Compared to traditional deterministic inversion methods, Bayesian-based seismic elastic parameter inversion can integrate multi-scale information from seismic and well logging data, constrain parameter inversion, and quantify the uncertainty of inversion results, making it widely used in seismic exploration. These Bayesian inversion methods typically employ Markov chain Monte Carlo and other methods within a posterior probabilistic framework to obtain a series of elastic parameter samples that satisfy observational data and prior constraints. Uncertainty analysis of the inversion results is then performed based on sample statistical analysis. The construction of the prior distribution plays a vital role in Bayesian inversion methods. A high-quality prior distribution can effectively guide the convergence direction of the posterior distribution solution and significantly improve the stability and geological rationality of the inversion results, even with incomplete or noisy data. However, constructing the prior distribution of seismic elastic parameters still faces challenges. Due to the influence of various factors such as lithology, pore structure, fluid properties, and sedimentary environment, the elastic parameters of subsurface media often exhibit significant non-Gaussian and multimodal distribution characteristics under complex geological backgrounds. Therefore, how to construct a priori distributions that can accurately characterize the statistical features of the elastic parameters of subsurface media remains a key issue in current Bayesian seismic inversion.
[0003] In existing Bayesian seismic inversion methods, the construction of prior distributions is mostly based on the Gaussian distribution assumption, typically assuming that elastic parameters (P-wave velocity, S-wave velocity, density, etc.) follow a unimodal, symmetrical normal distribution. However, this modeling approach struggles to accurately depict the complexity of subsurface media, especially in areas with diverse lithologies, discontinuous geological structures, or ambiguous lithofacies transitions. In these areas, the distribution of elastic parameters often exhibits significant non-Gaussian, skewed characteristics, and may even show discontinuities and multimodal structures. These statistical characteristics differ significantly from the ideal Gaussian distribution, limiting the applicability of prior distributions built on simple assumptions in Bayesian inversion. This results in substantial errors in practical applications, ultimately affecting the reliability and accuracy of lithology identification and reservoir prediction.
[0004] To address the shortcomings of traditional Gaussian priors, current Bayesian inversion methods have theoretically incorporated more flexible prior construction strategies to enhance their ability to characterize the statistical features of complex subsurface media. For example, some studies employ Gaussian Mixture Models (GMMs) that assume each lithological type follows a single Gaussian distribution, thereby constructing a prior distribution exhibiting multi-modal characteristics by superimposing multiple Gaussian components. Meanwhile, other studies have introduced p-norm constraints within the Bayesian framework to construct generalized prior distributions, which to some extent improves the flexibility and accuracy of probabilistic predictions.
[0005] However, these methods still have certain limitations in practical applications. On the one hand, while generalized prior distributions possess a certain degree of flexibility and can represent distributions with different morphological characteristics, they are essentially still unimodal distributions, and their expressive power is relatively limited for subsurface medium parameters with skewed and multimodal statistical characteristics. On the other hand, although Gaussian mixture models can better characterize multimodal characteristics, they rely on explicit lithological division and parameter initialization, making the modeling process complex and susceptible to noise or lithofacies overlap. Furthermore, the prior distributions in existing Bayesian inversion methods mostly rely on fixed-form probabilistic models, whose key statistical parameters are usually predetermined based on prior data and remain unchanged during the inversion process, lacking the ability to dynamically adjust prior model parameters. Therefore, how to construct a prior model that can effectively characterize the statistical characteristics of elastic parameters under complex geological backgrounds remains a key technical problem to be solved in the probabilistic inversion of seismic elastic parameters.
[0006] Current theories and technologies still face the following key challenges in characterizing the statistical features of elastic parameters in complex subsurface media: 1. The Gaussian prior assumptions widely used in current Bayesian inversion processes are typically based on unimodal, symmetrical distribution models, which are insufficient to effectively characterize the complex statistical features of subsurface media parameters. Especially in areas with multiple lithofacies, ambiguous lithological transitions, or complex geological structures, prior data often exhibit significant non-Gaussianity, skewness, and multimodal characteristics. The inability to effectively characterize prior data based on Gaussian prior assumptions affects the accuracy of inversion results. 2. While Gaussian mixture models can characterize multimodal features to some extent, they typically rely on explicit lithological divisions and parameter initialization. Inversion accuracy is significantly limited under conditions of lithological ambiguity or parameter overlap. 3. Although generalized prior models possess a degree of flexibility, expressing different distribution patterns based on different parameter values, they are essentially still unimodal distributions with limited expressive power, making it difficult to comprehensively characterize complex distribution patterns. 4. Most current prior modeling methods rely on fixed probability distributions, lacking the ability to adapt flexibly to different geological backgrounds and statistical data characteristics. A universal prior modeling framework applicable to complex geological conditions has not yet been formed.
[0007] In summary, a priori probabilistic modeling method suitable for complex geological conditions is needed to overcome the limitations of fixed-form distribution modeling. This method should be able to effectively characterize the complex distribution characteristics such as multimodality and asymmetry of the original sample data, accurately depict the complex distribution patterns of underground medium parameters, and provide more realistic and reliable prior constraints for the probabilistic inversion of seismic elastic parameters. Summary of the Invention
[0008] To address the aforementioned technical problems, this invention proposes a parameter-adaptive pre-stack probabilistic inversion method, equipment, and medium for seismic inversion.
[0009] In a first aspect, the present invention provides a parameter-adaptive pre-stack probabilistic inversion method for seismic inversion, comprising the following steps: S1, Obtain prior data information on the elastic parameters of the underground medium in the target area, and construct a generalized hybrid prior model based on the prior sample information; S2, establish a forward model from elastic parameters to seismic data, construct a likelihood function, and construct the posterior probability together with the generalized hybrid prior model; S3, in the first stage of inversion, adaptively update the number of mixture components in the generalized mixture prior model to determine the optimal number of mixture components; S4. In the second stage of inversion, based on the optimal number of mixture components, the number of mixture components in the generalized mixture prior model is fixed, and the standard deviation parameters of each component are adaptively updated while the elasticity parameters are updated. S5 calculates the corresponding acceptance probability, samples the data, iterates continuously, and outputs the inversion result.
[0010] Optionally, in S1, a generalized mixed prior model The expression is as follows: ; in, Q Indicates the number of mixed components; Indicates the first q The weights of a generalized distribution in the entire model satisfy the following: ; and They represent the first q The mean and standard deviation of a generalized distribution; Indicates the first q The probability of a generalized distribution.
[0011] Optionally, in S1, the generalized distribution The expression is as follows: ; in, , Indicates the standard deviation of the data; Represents the gamma function; p and The parameters representing the generalized distribution are adjusted by... p and Values achieve different morphological distributions
[0012] Optionally, in S2, a forward model from elastic parameters to seismic data is established based on the Aki-Richards approximation formula, and a likelihood function is constructed by combining the observed seismic data d. The expression is as follows: ; in, It is the covariance of seismic data noise. This represents the forward operator defined based on the Aki-Richards approximation formula. W is the seismic wavelet. Indicates the reflection coefficient; The expression for calculating the reflection coefficient is as follows: ; in, θ This represents the average of the incident angle and the transmitted angle; ; ; ; ; ; ; , and Parameters representing the upper medium. , and Parameters representing the underlying medium. Indicates the longitudinal wave velocity. Indicates the transverse wave velocity. Density
[0013] Optionally, in S2, the posterior probability The expression is as follows: ; in, , and These represent conditional probability, prior probability, and marginal probability, respectively; d represents observed seismic data, and the elastic parameter m includes the P-wave velocity. transverse wave velocity and density
[0014] Optionally, in S3, the first stage of the inversion is an adaptive update stage for the number of mixed components, and the process is as follows: Based on the statistical characteristics of the prior data, and given the upper and lower bounds of the number of mixture components, the parameter is perturbed within this range to generate candidate samples of the number of mixture components. The expression for the candidate samples is as follows: ; in, Represents the number of mixed components Q Candidate samples, This represents the number of mixed components in the previous iteration. This is the perturbation update amount, which is within the range of [-1, 1]. Number of mixed components Q A one-sided update strategy is adopted at the value boundary: when Q When the upper boundary is reached, updates are only allowed in the decreasing direction; when Q When the lower boundary is reached, updates are only allowed in the direction of increase; Generate candidate samples Then, based on the candidate samples, the corresponding generalized mixed prior model is reconstructed, and the posterior probability is calculated by combining the likelihood function. Calculate the acceptance probability corresponding to the posterior probability, and determine whether to update the number of mixture components and the elasticity parameter to be inverted based on the acceptance probability, thus determining the optimal number of mixture components.
[0015] Optionally, in S4, the second stage of the inversion is an adaptive update stage for the standard deviation parameter, the process of which includes: Based on the optimal number of mixture components, the number of mixture components of the generalized mixture prior model is fixed, and the standard deviation of each component is adaptively updated. The standard deviation parameter is updated using a random perturbation. A small random perturbation is applied to the standard deviation of the previous iteration to generate a candidate standard deviation, as shown in the following expression: ; in, and These are the candidate standard deviation and the standard deviation of the previous iteration, respectively. The perturbation update amount is the standard deviation; Reconstruct the corresponding generalized mixed prior model based on the candidate standard deviation, and calculate the posterior probability by combining the likelihood function. Calculate the acceptance probability corresponding to the posterior probability, and decide whether to update the standard deviation and the elasticity parameter to be inverted based on the acceptance probability, so as to realize the adaptive optimization and parameter inversion of the generalized mixed prior model parameters.
[0016] Optionally, in S3 and S4, the acceptance probability corresponding to the posterior probability is calculated according to the McMC MH algorithm.
[0017] In a second aspect, the present invention also provides a computer device, including a memory and a processor; The memory stores computer programs; When the processor executes the computer program, it implements the steps of the pre-stack probabilistic inversion method for earthquakes as described above.
[0018] In a third aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that the program, when executed by a processor, implements the pre-stack probabilistic inversion method for seismic inversion as described above.
[0019] The application of the technical solution of the present invention has at least the following beneficial effects: (1) This invention provides a parameter-adaptive pre-stack probabilistic inversion method, equipment and medium for seismic inversion. The method of this invention is based on the construction of a generalized hybrid prior model based on the generalized distribution. It introduces the hybrid modeling idea of Gaussian mixture model and forms an overall multi-peak structure by superimposing multiple generalized distribution components. It can effectively characterize the non-Gaussianity, multi-peak and asymmetric statistical characteristics of underground medium parameters under complex geological conditions. It is applicable to various geological backgrounds and complex reservoir conditions. Compared with the traditional single-peak Gaussian prior distribution, it is more in line with the actual geological conditions and improves the adaptability and accuracy of prior modeling.
[0020] (2) The method of the present invention adopts a two-parameter adaptive update strategy to incorporate the number of mixed components and the standard deviation of each component in the generalized mixed prior model into the inversion process for dynamic update. It adopts a phased update mechanism. First, the optimal number of mixed components in the generalized mixed prior model is determined. Then, based on this result, the number of mixed components in the second stage of inversion is fixed, and the standard deviation of each component is adaptively updated to reconstruct the generalized mixed prior model. This strategy reduces the dependence on fixed theoretical distribution assumptions and artificial parameter settings in the traditional prior modeling process, and improves the stability and generalization ability of the prior model. Compared with the prior model adopted in the traditional Bayesian inversion, it can more accurately describe the actual statistical distribution characteristics of underground medium parameters under complex geological conditions. Attached Figure Description
[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.
[0022] Figure 1 This is a flowchart of the steps of the pre-stack probabilistic inversion method for earthquakes in a preferred embodiment of the present invention; Figure 2 This is a schematic diagram of fitting a segment of longitudinal wave data based on Gaussian distribution and generalized mixed distribution; Figure 3 This is a schematic diagram of the elastic parameters obtained from the inversion result of the layered model based on Gaussian distribution; Figure 4 This is a schematic diagram of the elastic parameters of the inversion results obtained based on the Gaussian distribution inversion well model; Figure 5 This is a schematic diagram of the elastic parameters obtained by inverting the layered model using the method of this embodiment; Figure 6 This is a schematic diagram of the elastic parameters of the inversion result obtained by inverting the well model using the method of this embodiment. Detailed Implementation
[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0024] Example 1: like Figure 1 As shown, this embodiment provides a parameter-adaptive pre-stack probabilistic inversion method for seismic inversion, including the following steps (S1 to S5): S1. Obtain prior data on the elastic parameters of the subsurface medium in the target area, and construct a generalized hybrid prior model based on the prior sample information. In this embodiment, the elastic parameters include P-wave velocity, S-wave velocity, and density.
[0025] In this embodiment, the generalized hybrid prior model The expression is as follows: ; in, Q Indicates the number of mixed components; Indicates the first q The weights of a generalized distribution in the entire model satisfy the following: ; and They represent the first q The mean and standard deviation of a generalized distribution; Indicates the first q The probability of a generalized distribution.
[0026] Furthermore, broad distribution The expression is as follows: ; in, , Indicates the standard deviation of the data; Represents the gamma function; p and The parameters representing the generalized distribution are adjusted by... p and The values of achieve different morphological distributions. When p =2, = When the generalized distribution is used, it can represent the Gaussian prior model.
[0027] like Figure 2 As shown, Figure 2 This is a schematic diagram of fitting a segment of P-wave data based on Gaussian distribution and generalized mixture distribution. The P-wave data is represented in the form of a histogram. It can be seen that the generalized mixture distribution fits the data better than the Gaussian distribution.
[0028] S2. Establish a forward model from elastic parameters to seismic data, construct a likelihood function, and construct the posterior probability together with the generalized hybrid prior model.
[0029] In this embodiment, a forward model from elastic parameters to seismic data is established based on the Aki-Richards approximation formula, and a likelihood function is constructed by combining the observed seismic data d. The expression is as follows: ; in, It is the covariance of seismic data noise. This represents the forward operator defined based on the Aki-Richards approximation formula. W is the seismic wavelet. This represents the reflection coefficient.
[0030] Furthermore, the expression for calculating the reflection coefficient is as follows: ; in, θ This represents the average of the incident angle and the transmitted angle; ; ; ; ; ; ; , and Parameters representing the upper medium. , and Parameters representing the underlying medium. Indicates the longitudinal wave velocity. Indicates the transverse wave velocity. Indicates density.
[0031] Furthermore, posterior probability The expression is as follows: ; in, , and These represent conditional probability, prior probability, and marginal probability, respectively; d represents observed seismic data, and the elastic parameter m includes the P-wave velocity. transverse wave velocity and density .
[0032] S3, in the first stage of inversion, adaptively update the number of mixture components in the generalized mixture prior model to determine the optimal number of mixture components.
[0033] In this embodiment, the first stage of the inversion is an adaptive update stage for the number of mixed components, and the process is as follows: Based on the statistical characteristics of the prior data, and given the upper and lower bounds of the number of mixture components, the parameter is perturbed within this range to generate candidate samples of the number of mixture components. The expression for the candidate samples is as follows: ; in, Represents the number of mixed components Q Candidate samples, This represents the number of mixed components in the previous iteration. This is the perturbation update amount, which is within the range of [-1, 1]. Number of mixed components Q A one-sided update strategy is adopted at the value boundary: when Q When the upper boundary is reached, updates are only allowed in the decreasing direction; when Q When the lower boundary is reached, updates are only allowed in the direction of increase; Generate candidate samples Then, based on the candidate samples, the corresponding generalized mixed prior model is reconstructed, and the posterior probability is calculated by combining the likelihood function. Calculate the acceptance probability corresponding to the posterior probability, and determine whether to update the number of mixture components and the elasticity parameter to be inverted based on the acceptance probability, thus determining the optimal number of mixture components.
[0034] Specifically, the process of determining the optimal number of mixture components includes: The acceptance probability of a candidate sample in this iteration is calculated and compared with a random number between 0 and 1. If the acceptance probability is greater than the random number, the current elasticity parameter and the number of mixed components are updated as candidate samples; if the acceptance probability is less than the random number, the current elasticity parameter and the number of mixed components remain unchanged, and the process proceeds to the next iteration. This process is repeated within a preset range of mixed component numbers. Through continuous iterative search, the number of mixed components gradually converges towards a state with a higher posterior probability, ultimately determining the optimal number of mixed components, which is then fixed for the adaptive update of the standard deviation parameter in the second stage.
[0035] S4. In the second stage of inversion, based on the optimal number of mixture components, the number of mixture components in the generalized mixture prior model is fixed, and the standard deviation parameters of each component are adaptively updated while the elasticity parameters are updated.
[0036] The second stage of the inversion is the adaptive update stage of the standard deviation parameter, which includes the following process: Based on the optimal number of mixture components, the number of mixture components of the generalized mixture prior model is fixed, and the standard deviation of each component is adaptively updated. The standard deviation parameter is updated using a random perturbation. A small random perturbation is applied to the standard deviation of the previous iteration to generate a candidate standard deviation, as shown in the following expression: ; in, and These are the candidate standard deviation and the standard deviation of the previous iteration, respectively. The perturbation update amount is the standard deviation; Reconstruct the corresponding generalized mixed prior model based on the candidate standard deviation, and calculate the posterior probability by combining the likelihood function. Calculate the acceptance probability corresponding to the posterior probability, and decide whether to update the standard deviation and the elasticity parameter to be inverted based on the acceptance probability, so as to realize the adaptive optimization and parameter inversion of the generalized mixed prior model parameters.
[0037] Specifically, the process of adaptive optimization and parameter inversion of the parameters of the generalized hybrid prior model includes: The acceptance probability of a candidate sample in this iteration is calculated and compared with a random number between 0 and 1. If the acceptance probability is greater than the random number, the current elasticity parameter and standard deviation are updated to be candidate samples; if the acceptance probability is less than the random number, the current elasticity parameter and standard deviation are kept unchanged, and the process proceeds to the next iteration. Through multiple iterations, the standard deviation parameter is adjusted synchronously with the model parameters, achieving adaptive optimization of the generalized mixed prior model.
[0038] It should be noted that the acceptance probability corresponding to the posterior probability in this embodiment is calculated based on the McMC MH algorithm.
[0039] S5 calculates the corresponding acceptance probability, samples the data, iterates continuously, and outputs the inversion result.
[0040] like Figure 3 and Figure 4 As shown, the results obtained from the Gaussian prior distribution inversion show a poor match between the predicted and actual values; for example... Figure 5 and Figure 6 As shown, the results obtained by the two-parameter adaptive generalized hybrid prior inversion method of this embodiment show a good match between the predicted and true values, and the accuracy is significantly improved. This indicates that the method of this embodiment has an advantage in inversion accuracy compared with the traditional Gaussian prior distribution.
[0041] This embodiment provides a parameter-adaptive pre-stack probabilistic inversion method for seismic inversion. Based on a generalized distribution, it constructs a generalized hybrid prior model, introducing the hybrid modeling concept of Gaussian mixture models. By superimposing multiple generalized distribution components to form an overall multimodal structure, it can effectively characterize the non-Gaussianity, multimodality, and asymmetric statistical characteristics of subsurface medium parameters under complex geological conditions. Furthermore, this embodiment employs a two-parameter adaptive update strategy, incorporating the number of mixture components and the standard deviation of each component in the generalized hybrid prior model into the inversion process for dynamic updates. A staged update mechanism is adopted, reducing the reliance on fixed theoretical distribution assumptions and manually set parameters in traditional prior modeling, thus improving the stability and generalization ability of the prior model. Compared to the prior models used in traditional Bayesian inversion, it can more accurately describe the actual statistical distribution characteristics of subsurface medium parameters under complex geological conditions.
[0042] In addition, this embodiment also provides a computer device, including a memory and a processor; The memory stores computer programs; When the processor executes the computer program, it implements the steps of the pre-stack probabilistic inversion method for earthquakes as described above.
[0043] It should be noted that computer program code for performing the operations of this application can be written in one or more programming languages or a combination thereof. These programming languages include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0044] The flowcharts in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in the flowchart may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the flowchart, and combinations of blocks in the flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0045] In addition, this embodiment also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the pre-stack probabilistic inversion method described above.
[0046] The readable storage medium provided in this application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., a computer program) for executing the above-described pre-stack probabilistic inversion method. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the pre-stack probabilistic inversion method provided in the above embodiments, and will not be repeated here.
[0047] The above description is only a preferred embodiment of the present invention and does not limit the scope of the present invention. All equivalent structural transformations made under the inventive concept of the present invention using the contents of the present invention specification and drawings, or direct / indirect applications in other related technical fields, are included within the protection scope of the present invention.
Claims
1. A parameter-adaptive pre-stack probabilistic inversion method for seismic inversion, characterized in that, Includes the following steps: S1, Obtain prior data information on the elastic parameters of the underground medium in the target area, and construct a generalized hybrid prior model based on the prior sample information; S2, establish a forward model from elastic parameters to seismic data, construct a likelihood function, and construct the posterior probability together with the generalized hybrid prior model; S3, in the first stage of inversion, adaptively update the number of mixture components in the generalized mixture prior model to determine the optimal number of mixture components; S4. In the second stage of inversion, based on the optimal number of mixture components, the number of mixture components in the generalized mixture prior model is fixed, and the standard deviation parameters of each component are adaptively updated while the elasticity parameters are updated. S5 calculates the corresponding acceptance probability, samples the data, iterates continuously, and outputs the inversion result.
2. The seismic pre-stack probabilistic inversion method according to claim 1, characterized in that, In S1, the generalized mixed prior model The expression is as follows: ; in, Q Indicates the number of mixed components; Indicates the first q The weights of a generalized distribution in the entire model satisfy the following: ; and They represent the first q The mean and standard deviation of a generalized distribution; Indicates the first q The probability of a generalized distribution.
3. The seismic pre-stack probabilistic inversion method according to claim 2, characterized in that, In S1, the generalized distribution is expressed as follows: ; in, , Indicates the standard deviation of the data; Represents the gamma function; p and The parameters representing the generalized distribution are adjusted by... p and The values achieve different morphological distributions.
4. The seismic pre-stack probabilistic inversion method according to claim 3, characterized in that, In S2, a forward model from elastic parameters to seismic data is established based on the Aki-Richards approximation formula, and a likelihood function is constructed by combining the observed seismic data d. The expression is as follows: ; in, It is the covariance of seismic data noise. This represents the forward operator defined based on the Aki-Richards approximation formula. W is the seismic wavelet. Indicates the reflection coefficient; The expression for calculating the reflection coefficient is as follows: ; in, θ This represents the average of the incident angle and the transmitted angle; ; ; ; ; ; ; , and Parameters representing the upper medium. , and Parameters representing the underlying medium. Indicates the longitudinal wave velocity. Indicates the transverse wave velocity. Indicates density.
5. The seismic pre-stack probabilistic inversion method according to claim 4, characterized in that, In S2, the posterior probability The expression is as follows: ; in, , and These represent conditional probability, prior probability, and marginal probability, respectively; d represents observed seismic data, and the elastic parameter m includes the P-wave velocity. transverse wave velocity and density .
6. The seismic pre-stack probabilistic inversion method according to claim 5, characterized in that, In S3, the first stage of the inversion is the adaptive update stage of the number of mixed components, and the process is as follows: Based on the statistical characteristics of the prior data, and given the upper and lower bounds of the number of mixture components, the parameter is perturbed within this range to generate candidate samples of the number of mixture components. The expression for the candidate samples is as follows: ; in, Represents the number of mixed components Q Candidate samples, This represents the number of mixed components in the previous iteration. This is the perturbation update amount, which is within the range of [-1, 1]. Number of mixed components Q A one-sided update strategy is adopted at the value boundary: when Q When the upper boundary is reached, updates are only allowed in the decreasing direction; when Q When the lower boundary is reached, updates are only allowed in the direction of increase; Generate candidate samples Then, based on the candidate samples, the corresponding generalized mixed prior model is reconstructed, and the posterior probability is calculated by combining the likelihood function. Calculate the acceptance probability corresponding to the posterior probability, and determine whether to update the number of mixture components and the elasticity parameter to be inverted based on the acceptance probability, thus determining the optimal number of mixture components.
7. The seismic pre-stack probabilistic inversion method according to claim 6, characterized in that, In S4, the second stage of the inversion is the adaptive update stage of the standard deviation parameter, and the process includes: Based on the optimal number of mixture components, the number of mixture components of the generalized mixture prior model is fixed, and the standard deviation corresponding to each component is adaptively updated. The standard deviation parameter is updated using a random perturbation method. A small random perturbation is applied to the standard deviation of the previous iteration to generate a candidate standard deviation, as shown in the following expression: ; in, and These are the candidate standard deviation and the standard deviation of the previous iteration, respectively. The perturbation update amount is the standard deviation; Reconstruct the corresponding generalized mixed prior model based on the candidate standard deviation, and calculate the posterior probability by combining the likelihood function. Calculate the acceptance probability corresponding to the posterior probability, and decide whether to update the standard deviation and the elasticity parameter to be inverted based on the acceptance probability, so as to realize the adaptive optimization and parameter inversion of the generalized mixed prior model parameters.
8. The seismic pre-stack probabilistic inversion method according to claim 7, characterized in that, In S3 and S4, the acceptance probability corresponding to the posterior probability is calculated according to the McMC MH algorithm.
9. A computer device, characterized in that, Including memory and processor; The memory stores computer programs; When the processor executes the computer program, it implements the steps of the pre-stack probabilistic inversion method as described in any one of claims 1 to 8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the pre-stack probabilistic inversion method for seismic inversion as described in any one of claims 1 to 8.
