MCMC-based lithofacies and impedance synchronous seismic inversion method

Abstract

The invention discloses an MCMC-based lithofacies and impedance synchronous seismic inversion method and system, and the method comprises the following steps: carrying out the statistics of a lithofacies transition probability matrix and the mean value and variance of impedance parameters corresponding to different lithofacies based on actual logging data; calculating an impedance parameter vertical variation function based on the well data; randomly generating an initial lithofacies combination based on the statistical lithofacies transition probability matrix, and calculating impedance corresponding to the lithofacies by using the statistical mean value, variance and variation function; generating lithofacies and impedance parameters again based on the randomly generated lithofacies and impedance, designing a target loss function and introducing an L1 norm, respectively calculating loss values corresponding to the lithofacies and impedance parameters twice, and judging whether to accept the newly generated lithofacies and impedance parameters by using a Markov chain acceptance criterion; and after iteration is completed, carrying out statistics on distribution conditions of lithofacies and impedance parameters in the Markov chain in the stable stage, and evaluating uncertainty information of a prediction result. According to the invention, the resolution and precision of impedance prediction are improved.

Description

Technical Field

[0001] The present invention relates to the field of seismic exploration technology, and in particular to a MCMC-based lithofacies and impedance synchronous seismic inversion method. Background Art

[0002] As exploration activities continue to advance, the geological conditions faced become more complex, the target layers gradually go deeper underground, and the requirements for exploration accuracy also increase. Seismic inversion technology plays a vital role in the exploration and development of oil and gas resources. It provides a solid foundation for accurately evaluating the characteristics of oil and gas reservoirs, accurately calculating oil and gas reserves, and formulating reasonable production plans. By comprehensively analyzing seismic data, downhole measured information and existing geological knowledge, seismic inversion can not only reveal the spatial layout and structural morphology of oil and gas reservoirs from a macro level, but also effectively predict the changing trend of reservoir parameters, thereby supporting more sophisticated resource management and development decisions. Elastic parameter prediction is an important part of seismic inversion. It involves the estimation of rock physical properties such as velocity, density, impedance, etc. These parameters reflect the response characteristics of rocks under external forces. Different types of rocks and the fluids (oil, gas or water) contained in them will lead to different elastic responses. Therefore, accurate prediction of elastic parameters can effectively identify potential oil and gas reservoir locations and help understand the mechanical properties of reservoirs. This information is crucial for optimizing drilling decisions because it helps reduce the risk of dry wells while increasing the possibility of discovering commercial-scale oil and gas fields. Lithofacies prediction is another key technology, which aims to infer the distribution of rock types in unknown areas based on existing data. Lithofacies refers to a detailed description of the internal structure of rocks and their composition, which is directly related to the effectiveness of the reservoir - that is, whether the reservoir has good storage capacity and the ability to allow fluid flow. Through lithofacies prediction, geologists can obtain valuable insights into reservoir heterogeneity, which is of great significance for selecting the best well location, designing effective completion schemes, and adopting advanced enhanced oil recovery technologies. In addition, lithofacies prediction also supports a deeper understanding of reservoir connectivity and permeability, thereby promoting a more scientific and reasonable oilfield development strategy. Seismic inversion is usually divided into deterministic inversion and uncertainty inversion. Deterministic inversion focuses on finding one or a group of optimal models that can best explain the observed data. This method is widely used in actual projects, especially when the geological conditions are relatively simple. However, as exploration targets gradually shift to more complex and deeper geological environments, relying solely on a single solution may not be enough to fully understand the reservoir characteristics and their potential risks. In contrast, uncertainty inversion has received increasing attention in recent years, especially under the Bayesian framework-based approach. Bayesian uncertainty inversion not only seeks to find the best model, but more importantly, provides a probability distribution of all possible models, which allows researchers to evaluate the reliability of results under different assumptions. By integrating prior knowledge (such as existing geological information) with currently collected seismic data, the Bayesian method can generate a posterior probability density function that describes in detail the likelihood of various models. The Markov Chain Monte Carlo (MCMC) method is an effective technique for solving seismic inversion problems under the Bayesian framework, especially for dealing with high-dimensional and nonlinear inverse problems.In seismic inversion, MCMC generates a series of model samples by sampling from the posterior probability distribution, which can not only identify the most likely geological model, but also comprehensively evaluate the uncertainty of the model parameters. This method is particularly suitable for applications under complex geological conditions because it can naturally handle multimodal posterior distributions and avoid falling into local optimal solutions, thereby increasing the possibility of finding the global optimal solution. In addition, MCMC allows rich prior information (such as geological knowledge, logging data, etc.) to be integrated into the inversion process, enhancing the reliability and accuracy of the results. After obtaining the elastic parameters, lithofacies prediction can be achieved using supervised or unsupervised data-driven methods.

[0003] In summary, the current research on reservoir lithofacies and impedance prediction methods has the following problems: (1) The deterministic inversion method can only give a unique solution to the prediction result, but it is affected by noise, seismic forward modeling assumptions, etc. When conducting reservoir evaluation, if only relying on a deterministic model, other potential geological scenarios may be ignored, resulting in misunderstanding or misestimation of reservoir characteristics; (2) Seismic inversion based on the Bayesian framework can obtain analytical solutions when dealing with linear problems, but when facing nonlinear situations, solving the posterior distribution becomes extremely complicated, and the nonlinear relationship makes it difficult to obtain the posterior distribution by direct calculation; (3) The MCMC inversion method has a large randomness in the solution process, which leads to poor stability of its predicted elastic parameters. Since MCMC relies on random sampling to explore the posterior distribution, especially when the sampling is insufficient or the chain is not fully mixed, each run may produce different results. This instability affects the reliable estimation of underground structures; (4) The conventional method of stepwise predicting impedance and then predicting lithofacies itself has cumulative errors, because the errors of each step will accumulate in the subsequent steps. In addition, the impedance value itself depends on the combination of different facies, and ignoring the interdependence between facies and impedance leads to a decrease in the accuracy of facies and impedance prediction. Summary of the invention

[0004] The purpose of the present invention is to provide a MCMC-based lithofacies and impedance synchronous seismic inversion method, which effectively avoids the cumulative error of impedance and lithofacies step-by-step prediction through an impedance iterative optimization algorithm dependent on lithofacies, and simultaneously estimates the impedance and lithofacies of the reservoir, thereby meeting the requirements of high-precision seismic exploration and fine reservoir characterization.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A MCMC-based lithofacies and impedance synchronous seismic inversion method comprises the following steps:

[0007] Step 1: Based on the actual logging data, the lithofacies transition probability matrix is ​​statistically used to generate reasonable lithofacies combinations in the future. By analyzing the logging data, a transition probability matrix reflecting the spatial distribution law of underground lithofacies is constructed, and the mean and variance of the impedance parameters corresponding to different lithofacies are accurately statistically calculated. This provides key basic data for simulation and prediction, ensuring that the lithofacies model is highly consistent with the actual geological conditions.

[0008] Step 2: Calculate the vertical variation function of the impedance parameters based on the well data and establish a vertical continuity model of the impedance parameters. By analyzing the variation of the impedance parameters with depth in the well data, a variation function reflecting the spatial correlation of these parameters is constructed to reduce the randomness introduced in the process of generating the Markov chain and enhance the stability and reliability of the impedance prediction results.

[0009] Step 3: Based on the statistical lithofacies transition probability matrix, the initial lithofacies combination is first randomly generated. Then, the impedance value corresponding to each rock is calculated using the impedance parameter mean, variance and vertical variation function of different lithofacies obtained in the first two steps.

[0010] Step 4: Based on the randomly generated initial lithofacies and impedance parameters, further iteratively generate new lithofacies and impedance data. Design a target loss function to quantify the difference between the predicted results and the actual data. At the same time, introduce the L1 norm constraint in the loss function to promote sparsity and ensure that the model is simple and effective. Calculate the loss values ​​corresponding to the initial and regenerated lithofacies and impedance parameters respectively, and evaluate the improvement effect of the model by comparing the loss values ​​of the two.

[0011] Step 5: According to the acceptance criteria of the Markov chain, by comparing the newly generated lithofacies and impedance parameters with the current model state, decide whether to accept the new parameter set. If the new parameter set can significantly improve the loss function value under the new and old states, it is directly accepted; otherwise, it is accepted or rejected based on a certain probability.

[0012] Step 6. Repeat steps 4 and 5 until the preset number of iterations is reached. During this process, the facies and impedance parameters are continuously updated according to the acceptance criteria of the Markov chain to ensure that each iteration moves towards the optimization goal. When the iteration is completed, the distribution of facies and impedance parameters in the Markov chain in the stable stage is statistically analyzed to determine the final prediction result. At the same time, the uncertainty information of the prediction result is evaluated by analyzing the variability and convergence in the chain.

[0013] In some embodiments, in step 1, the lithofacies transition probability matrix is:

[0014]

[0015] In the formula, each row is the current state lithofacies, and each column is the probability of the current state lithofacies converting to other lithofacies.

[0016] In some embodiments, in step 2, the variogram is:

[0017]

[0018] Where, v is the variogram value calculated based on well data; h is the hysteresis distance; N is the length of logging data; m(x i ) is the impedance parameter value at a certain position i.

[0019] In some embodiments, step 3 includes the following steps:

[0020] S31. Randomly generate initial lithofacies combination:

[0021] p(ξ i+1 )=p(ξ i )γ (3);

[0022] In the formula, p(ξ i+1 ) is the state of the lithofacies at time i+1; p(ξ i ) is the state of the lithofacies at time i; γ is the lithofacies transition probability matrix; given the lithofacies state at time 0, a lithofacies sequence of any length can be obtained according to formula (3);

[0023] S32. By performing Kronecker product on the matrix constructed by the variogram and the impedance variance matrix of each lithofacies, an impedance curve satisfying the characteristics of the logging curve is simulated and generated:

[0024]

[0025] Where p(ξ) is the current lithofacies state; N(zp; μ zp|τ , α zp|τ ) is the impedance distribution corresponding to the lithofacies state of p(ξ); μ zp|τ is the mean value of the corresponding impedance under the p(ξ) lithofacies state; α zp|τ Ω is the covariance obtained by Kronecker product of impedance variance and variogram; n It is the possible lithofacies space.

[0026] In some embodiments, the calculation expression for the loss value corresponding to the lithofacies and impedance parameters generated twice in step 4 is:

[0027] LOSS=||d-Gm||2+||Dm||1 (5);

[0028] Where d is the observed seismic data; G is the forward operator for obtaining synthetic seismic records from impedance m; and D is the differential operator.

[0029] In some embodiments, in step 5, by comparing the newly generated lithofacies and impedance parameters with the current model state, it is decided whether to accept the new parameter set; if the new parameter set can improve the loss function value under the new and old states, it is directly accepted; otherwise, it is accepted or rejected according to a certain probability;

[0030] The loss function value under the new and old states is:

[0031]

[0032] In the formula, p(d|zp t+1 )p(zp t+1 ) is the logarithm of the loss value in the new state; p(d|zp t )p(zp t ) is the logarithm of the loss value in the old state. It should be noted that the smaller the loss value, the larger the logarithm.

[0033] In some embodiments, if the new state loss value is less than the old state loss value, Greater than 1, otherwise less than 1; at the same time, a number between 0 and 1 is randomly generated If η is greater than If the new state is selected, the lithofacies and impedance are accepted; otherwise, the original state is retained.

[0034] A MCMC-based lithofacies and impedance synchronous seismic inversion system, comprising:

[0035] Statistics module: Based on the actual logging data, it is used to calculate the lithofacies transition probability matrix for the subsequent generation of lithofacies combinations, and to calculate the mean and variance of the impedance parameters corresponding to different rocks;

[0036] Impedance parameter calculation module: Based on well data, it is used to calculate the vertical variation function of impedance parameters and establish the vertical continuity of impedance parameters;

[0037] Random generation module: The lithofacies transition probability matrix is ​​used to randomly generate the initial lithofacies combination, and the impedance of the lithofacies is calculated using the statistical mean, variance and variogram;

[0038] Regeneration module: Based on the randomly generated lithofacies and impedance, it is used to regenerate lithofacies and impedance parameters, and calculate the loss values ​​corresponding to the lithofacies and impedance parameters generated twice respectively using the target loss function and L1 norm constraint;

[0039] Judgment module: It is used to judge whether to accept the newly generated lithofacies and impedance parameters according to the Markov chain acceptance criterion;

[0040] Iterative output module: repeat steps S4-S5 until a given number of iterations is reached, statistically analyze the distribution of lithofacies and impedance in the stable Markov chain, and obtain the final lithofacies and impedance prediction results and their uncertainty information.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. The present invention considers the coupling relationship between lithofacies and impedance, and predicts reservoir lithofacies and impedance parameters at the same time, thus avoiding the cumulative error caused by the step-by-step prediction method;

[0043] 2. The present invention is based on the Bayesian framework and adopts the MCMC method to solve the inverse problem. It gives the uncertainty information of the result while giving the result, so as to realize the reliability evaluation of the result;

[0044] 3. The present invention considers the vertical variation function of the geological sedimentation process in the MCMC iteration process, so that there is a coupling relationship between adjacent points, reduces the randomness of the generated lithofacies and impedance parameters, and accelerates the convergence of the Markov chain;

[0045] 4. Add L1 norm constraint to the target loss to promote the sparsity of impedance parameters and improve the resolution and accuracy of impedance prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a schematic diagram of the overall process of Embodiment 1 of the present invention;

[0047] Figure 2 Schematic diagram of well data tested in Example 1 of the present invention; wherein, Figure (a) is lithofacies data; Figure (b) is impedance data; and Figure (c) is seismic record data;

[0048] Figure 3 This is a schematic diagram of the distribution of the impedances corresponding to various rock phases based on well data statistics according to the first embodiment of the present invention;

[0049] Figure 4 Schematic diagram of the inversion result of the first embodiment of the present invention; wherein, Figure (a) is the lithofacies iterative update process; Figure (b) is the final predicted lithofacies; Figure (c) is the probability of each lithofacies; Figure (d) is the impedance prediction result and the random realization of the iterative process;

[0050] Figure 5 It is a schematic diagram of the structure of the system of embodiment 2 of the present invention. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0052] Embodiment 1

[0053] See also Figure 1 - Figure 4 , a MCMC-based lithofacies and impedance simultaneous seismic inversion method, is proposed based on the following questions:

[0054] 1. Although the deterministic inversion method can provide a unique solution to the prediction results, this method is easily affected by data noise and the accuracy of seismic forward modeling assumptions. In the actual application of reservoir evaluation, if you rely entirely on a single deterministic model, you may ignore a variety of potential geological scenarios and the complex and changeable characteristics of the reservoir. This limitation may lead to a one-sided understanding or misjudgment of reservoir properties, which in turn affects a series of decision-making processes such as resource assessment and development planning. Therefore, when conducting reservoir analysis, combining probabilistic methods and other auxiliary information to consider more possibilities is an important way to improve the accuracy and reliability of evaluation.

[0055] 2. The seismic inversion method based on the Bayesian framework can provide direct analytical solutions when dealing with linear problems, which makes parameter estimation relatively simple and intuitive. However, when encountering nonlinear situations, solving the posterior probability distribution becomes extremely complicated. Nonlinear relationships not only increase the difficulty of calculation, but may also cause the posterior distribution to exhibit multimodal or multi-peak characteristics, making it impossible to obtain accurate results through simple mathematical operations. At this point, the traditional direct calculation method is no longer applicable, and it is usually necessary to use numerical methods such as Monte Carlo simulation and Markov chain to approximate the posterior distribution in order to obtain more reliable geological parameter estimates.

[0056] 3. The MCMC inversion method introduces significant random elements in the solution process, which makes the predicted elastic parameters relatively less stable. Since the method relies on random sampling to explore complex posterior probability distributions, each run may produce different results in the case of insufficient sampling or insufficient mixing of the Markov chain. This variability in results reduces the reliability and consistency of the estimation of underground structural characteristics.

[0057] 4. The conventional method of predicting impedance step by step and then predicting lithofacies is prone to cumulative errors, because small deviations in each step will gradually amplify in subsequent analysis, ultimately affecting the accuracy of the overall result. In addition, this method often ignores the inherent interdependence between impedance values ​​and lithofacies - impedance is the result of a combination of different lithofacies, and lithofacies in turn affect impedance characteristics. This neglect leads to the failure to fully reflect the complex interaction between the two in the prediction process, thereby reducing the accuracy of lithofacies and impedance predictions.

[0058] The specific steps include:

[0059] S1, such as Figure 2 As shown in Figure 1, based on the actual logging data, the lithofacies transition probability matrix is ​​statistically used to generate lithofacies combinations in the future. By analyzing the logging data, a lithofacies transition probability matrix reflecting the spatial distribution law of underground lithofacies is constructed, and the mean and variance of the impedance parameters corresponding to different lithofacies are accurately statistically calculated, as shown in Figure 1. Figure 3 This provides key basic data for simulation and prediction, ensuring that the lithofacies model is highly consistent with the actual geological conditions.

[0060] The impedance mean μ and variance δ corresponding to different lithofacies τ are recorded as: (τ1, μ1, δ1), (τ2, μ2, δ2), (τ3, μ3, δ3). This statistical method helps to more accurately understand the conversion relationship and physical characteristics between lithofacies and improve the reliability and accuracy of reservoir evaluation and prediction.

[0061] The lithofacies transition probability matrix is:

[0062]

[0063] In the formula, each row represents the current state of lithofacies 1, 2, and 3, and each column represents the probability of the next state changing from the current state of lithofacies to other lithofacies. For example, the first row and the second column represent the probability of changing from lithofacies 1 to lithofacies 2 is 0.05.

[0064] S2. Calculate the vertical variation function of impedance parameters based on well data and effectively establish a vertical continuity model of impedance parameters. By analyzing the variation law of impedance parameters with depth in well data, a variation function reflecting the spatial correlation of these parameters is constructed to reduce the randomness introduced in the process of generating Markov chains and enhance the stability and reliability of impedance prediction results.

[0065] The variogram is:

[0066]

[0067] Where, v is the variogram value calculated based on well data; h is the hysteresis distance; N is the length of logging data; m(x i) is the impedance parameter value at a certain position i.

[0068] S3. Based on the statistical lithofacies transition probability matrix, the initial lithofacies combination is first randomly generated.

[0069] p(ξ i+1 )=p(ξ i )γ (3);

[0070] In the formula, p(ξ i+1 ) is the state of the lithofacies at time i+1; p(ξ i ) is the state of the lithofacies at time i; γ is the lithofacies transition probability matrix; given the lithofacies state at time 0, a lithofacies sequence of any length can be obtained according to formula (3).

[0071] Then, using the impedance parameter mean, variance and vertical variogram of different lithofacies obtained in the first two steps, the matrix constructed by the variogram is Kronecker-multiplied with the impedance variance matrix of each lithofacies (Formula 4) to simulate and generate an impedance curve that meets the characteristics of the logging curve:

[0072]

[0073] Where p(ξ) is the current lithofacies state; N(zp; μ zp|τ , α zp |τ) is the impedance distribution corresponding to the lithofacies state p(ξ); μ zp|τ is the mean value of the corresponding impedance under the p(ξ) lithofacies state; α zp|τ Ω is the covariance obtained by Kronecker product of impedance variance and variogram; n It is the possible lithofacies space.

[0074] S4. Based on the randomly generated lithofacies and impedance, further iteratively generate new lithofacies and impedance data. Design a target loss function to quantify the difference between the predicted results and the actual data. At the same time, introduce the L1 norm constraint to promote sparsity and ensure the simplicity and effectiveness of the model (Formula 5). Calculate the loss values ​​(LOSS) corresponding to the lithofacies and impedance parameters generated initially and again. t and LOSS t+1 ), and evaluate the improvement effect of the model by comparing the loss values ​​of the two.

[0075] LOSS=||d-Gm||2+||Dm||1 (5);

[0076] Where d is the observed seismic data; G is the forward operator for obtaining synthetic seismic records from impedance m; and D is the differential operator.

[0077] S5. According to the Markov chain acceptance criterion, by comparing the newly generated lithofacies and impedance parameters with the current model state, decide whether to accept the new parameter set. If the new parameter set can significantly improve the loss function value under the new and old states (Formula 6), it is directly accepted; otherwise, it is accepted or rejected based on a certain probability.

[0078]

[0079] In the formula, p(d|zp t+1 )p(zp t+1 ) is the logarithm of the loss value in the new state; p(d|zp t )p(zp t ) is the logarithm of the loss value in the old state. It should be noted that the smaller the loss value, the larger the logarithm.

[0080] If the new state loss value is less than the old state loss value, Greater than 1, otherwise less than 1; at the same time, a number between 0 and 1 is randomly generated If η is greater than If the new state is selected, the lithofacies and impedance are accepted; otherwise, the original state is retained.

[0081] S6. Repeat steps S4-S5 until the given number of iterations is reached. During this process, the lithofacies and impedance parameters are continuously updated according to the acceptance criteria of the Markov chain to ensure that each iteration moves towards the optimization goal. When the iteration is completed, the distribution of lithofacies and impedance parameters in the Markov chain in the stable stage is statistically analyzed to determine the final prediction result. At the same time, the uncertainty information of the prediction result is evaluated by analyzing the variability and convergence in the chain.

[0082] like Figure 4 As shown, Figure 4 (a) is the random realization of lithofacies during the iteration process. As the number of iterations increases, the lithofacies simulation results gradually become stable. Figure 4 (b) is the average lithofacies result obtained after the lithofacies tend to be stable. Figure 4 (c) is the probability corresponding to the predicted rock; Figure 4 (d) is the impedance prediction result, where the background represents the possible impedance result range. By analyzing the variability and convergence in the chain, the uncertainty information of the prediction result can be well evaluated.

[0083] Embodiment 2

[0084] A MCMC-based simultaneous seismic inversion system for lithofacies and impedance, such as Figure 5 As shown, including:

[0085] Statistics module: Based on the actual logging data, it is used to calculate the lithofacies transition probability matrix for the subsequent generation of lithofacies combinations, and to calculate the mean and variance of the impedance parameters corresponding to different rocks;

[0086] Impedance parameter calculation module: Based on well data, it is used to calculate the vertical variation function of impedance parameters and establish the vertical continuity of impedance parameters;

[0087] Random generation module: The lithofacies transition probability matrix is ​​used to randomly generate the initial lithofacies combination, and the impedance of the lithofacies is calculated using the statistical mean, variance and variogram;

[0088] Regeneration module: Based on the randomly generated lithofacies and impedance, it is used to regenerate lithofacies and impedance parameters, and calculate the loss values ​​corresponding to the lithofacies and impedance parameters generated twice respectively using the target loss function and L1 norm constraint;

[0089] Judgment module: It is used to judge whether to accept the newly generated lithofacies and impedance parameters according to the Markov chain acceptance criterion;

[0090] Iterative output module: Repeat the methods in the regeneration module and the judgment module until the given number of iterations is reached, statistically analyze the distribution of lithofacies and impedance in the stable Markov chain, and obtain the final lithofacies and impedance prediction results and their uncertainty information.

[0091] The MCMC-based lithofacies and impedance synchronous seismic inversion system of the present invention can be installed in a computer device. The computer device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a program for generating a lithofacies and impedance synchronous seismic inversion system based on MCMC. The memory includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. The processor is the control core of the electronic device, and uses various interfaces and lines to connect the various components of the entire computer device, and executes various functions of the computer device and processes data by running or executing programs or modules stored in the memory, and calling data stored in the memory.

[0092] The module described in the present invention refers to a series of computer program segments that can be executed by a processor of a computer device and can complete fixed functions, and is stored in a memory of the computer device.

Claims

1. A MCMC-based lithofacies and impedance synchronous seismic inversion method, characterized in that: The following steps are involved: S1. Based on the actual logging data, the lithofacies transition probability matrix is ​​statistically used to generate lithofacies combinations in the future, and the mean and variance of impedance parameters corresponding to different rocks are statistically analyzed; S2. Calculate the vertical variation function of impedance parameters based on well data and establish the vertical continuity of impedance parameters; S3, using the lithofacies transfer probability matrix to randomly generate the initial lithofacies combination, and using the statistical mean, variance and variogram to calculate the impedance of the lithofacies; S4, based on the randomly generated lithofacies and impedance, the lithofacies and impedance parameters are generated again, the target loss function is designed, and the L1 norm constraint is introduced to calculate the loss values ​​corresponding to the lithofacies and impedance parameters generated twice respectively; S5. judging whether to accept the newly generated lithofacies and impedance parameters according to the Markov chain acceptance criterion; S6. Repeat steps S4-S5 until a given number of iterations is reached, and the distribution of lithofacies and impedance in the stable Markov chain is statistically analyzed to obtain the final lithofacies and impedance prediction results and their uncertainty information.

2. The MCMC-based lithofacies and impedance synchronous seismic inversion method according to claim 1, characterized in that: In S1, the lithofacies transition probability matrix is: In the formula, each row is the current state lithofacies, and each column is the probability of the current state lithofacies converting to other lithofacies.

3. The MCMC-based lithofacies and impedance synchronous seismic inversion method according to claim 1, characterized in that: In S2, the variogram is: Where, v is the variogram value calculated based on well data; h is the lag distance; N is the length of logging data; m(ξ i ) is the impedance parameter value at a certain position i.

4. The MCMC-based lithofacies and impedance synchronous seismic inversion method according to claim 1, characterized in that: S3 includes the following steps: S31. Randomly generate initial lithofacies combination: p(ξ i+1 )=p(ξ i )c(3); In the formula, p(ξ i+1 ) is the state of the lithofacies at time i+1; p(ξ i ) is the state of the lithofacies at time i; γ is the lithofacies transition probability matrix; given the lithofacies state at time 0, a lithofacies sequence of any length can be obtained according to formula (3); S32. By performing Kronecker product on the matrix constructed by the variogram and the impedance variance matrix of each lithofacies, an impedance curve satisfying the characteristics of the logging curve is simulated and generated: Where p(ξ) is the current lithofacies state; N(zp;ξ zp|τ ,α zp|τ ) is the impedance distribution corresponding to the lithofacies state of p(ξ); μ zp|τ is the mean value of the corresponding impedance under the p(ξ) lithofacies state; α zp|τ Ω is the covariance obtained by Kronecker product of impedance variance and variogram; n It is the possible lithofacies space.

5. The MCMC-based lithofacies and impedance synchronous seismic inversion method according to claim 1, characterized in that: The calculation expression of the loss value corresponding to the lithofacies and impedance parameters generated twice in S4 is: LOSS=‖d-Gm‖2+‖Dm‖1 (5); Where d is the observed seismic data; G is the forward operator for obtaining synthetic seismic records from impedance m; and D is the differential operator.

6. The MCMC-based lithofacies and impedance synchronous seismic inversion method according to claim 1, characterized in that: In S5, by comparing the newly generated lithofacies and impedance parameters with the current model state, it is decided whether to accept the new parameter set; if the new parameter set can improve the loss function value under the new and old states, it is directly accepted; otherwise, it is accepted or rejected according to a certain probability; The loss function value under the new and old states is: In the formula, p(d|zp t+1 )p(zp t+1 ) is the logarithm of the loss value in the new state; p(d|zp t )p(zp t ) is the logarithm of the loss value in the old state.

7. The MCMC-based lithofacies and impedance synchronous seismic inversion method according to claim 6, characterized in that: If the new state loss value is less than the old state loss value, Greater than 1, otherwise less than 1; at the same time, a number between 0 and 1 is randomly generated If η is greater than If the new state is selected, the lithofacies and impedance are accepted; otherwise, the original state is retained.

8. A MCMC-based lithofacies and impedance synchronous seismic inversion system, using a MCMC-based lithofacies and impedance synchronous seismic inversion method as claimed in any one of claims 1 to 7, characterized in that: include: Statistics module: Based on the actual logging data, it is used to calculate the lithofacies transition probability matrix for the subsequent generation of lithofacies combinations, and to calculate the mean and variance of the impedance parameters corresponding to different rocks; Impedance parameter calculation module: Based on well data, it is used to calculate the vertical variation function of impedance parameters and establish the vertical continuity of impedance parameters; Random generation module: The lithofacies transition probability matrix is ​​used to randomly generate the initial lithofacies combination, and the impedance of the lithofacies is calculated using the statistical mean, variance and variogram; Regeneration module: Based on the randomly generated lithofacies and impedance, it is used to regenerate lithofacies and impedance parameters, and calculate the loss values ​​corresponding to the lithofacies and impedance parameters generated twice respectively using the target loss function and L1 norm constraint; Judgment module: It is used to judge whether to accept the newly generated lithofacies and impedance parameters according to the Markov chain acceptance criterion; Iterative output module: repeat steps S4-S5 until a given number of iterations is reached, statistically analyze the distribution of lithofacies and impedance in the stable Markov chain, and obtain the final lithofacies and impedance prediction results and their uncertainty information.

Images