A probabilistic geological analysis method for the ratio of thin interbedded sand to soil in drilling soil

Abstract

The invention discloses a probabilistic geological analysis method for the sand-to-land ratio of thin interlayers in a drilling soil layer, comprising the following steps: quantitatively constructing a massive geological model according to the sedimentary law and sedimentary phase thickness of the thin interlayer reservoir in the well; calculating the probabilistic distribution characteristics of the sand-to-land ratio of the massive geological model corresponding to different well positions; using the Kriging interpolation method to calculate the probabilistic distribution characteristics of the sand-to-land ratio of the massive geological model at different well positions to construct a probabilistic geological constraint for the sand-to-land ratio; calculating and generating a seismic attribute interpretation model corresponding to the massive geological model based on a propagation matrix forward modeling method and the probabilistic geological constraint of the sand-to-land ratio; fusing the probabilistic geological constraint and the seismic attribute model through the following Bayesian framework to obtain an accurate seismic prediction of the sand-to-land ratio; the invention is suitable for the construction of geological constraints and subsequent quantitative seismic interpretation of the sand-to-land ratio of thin interlayers with different geological characteristics, and thereby guiding subsequent reservoir exploration and development work.

Description

Technical Field

[0001] The present invention belongs to geophysical exploration technology, relates to probabilistic geological constraints of thin interbedded sand-to-soil ratio in geophysical exploration of sandstone and mudstone reservoirs and related quantitative seismic interpretation methods, and in particular to a probabilistic geological analysis method for thin interbedded sand-to-soil ratio in drilling soil layers. Background Art

[0002] The thin interbedded sandstone reservoir has geological characteristics such as variable sedimentary patterns, thin sand body thickness, rapid lateral changes, and interbedded development, which leads to complex seismic reflection characteristics. It is difficult to accurately describe the reservoir using existing geophysical technologies. Widess proposed that a single thin layer smaller than the tuning thickness can be described by a linear relationship between thickness and amplitude (Widess, 1982). As one of the key technologies in seismic sedimentology, seismic attribute slicing technology is also an important tool for thin reservoir prediction. Compared with seismic profile sand body tracking, it has better single sand body identification effect (Zeng, 2001; Li, 2015). Hou et al. constructed the relationship between thin interbedded reservoir parameters and seismic response, and converted the seismic response of thin interbedded layers into sand-to-ground ratio by calculating its tuning factor through support vector machine (Hou, 2019). In addition, random inversion techniques such as seismic waveform inversion are also commonly used methods to identify the combined structure and spatial distribution of thin interbedded layers using reflection waveform characteristics (Pei, 2017). However, the attribute interpretation technology constructed entirely based on the complex seismic mapping relationship of thin interlayer reservoirs in the above methods is difficult to accurately predict thin interlayers. The random inversion method based on random simulation lacks interwell interpolation accuracy and geological rationality, and seismic medium-frequency information cannot meet the needs of accurate interpretation. The multi-solution and uncertainty of the existing interpretation methods cannot be compensated, and the drawbacks of lack of geological constraints have not been improved. It is urgent to introduce reasonable strong geological constraints to improve interpretation accuracy and geological interpretability. Summary of the invention

[0003] The present invention provides a method for constructing and using a probabilistic geological constraint on the sand-to-land ratio of thin interlayers in a drilling soil layer; the method comprises a new method for constructing a probabilistic constraint on the sand-to-land ratio based on a massive geological model that conforms to the sedimentary law, wherein the well point position is controlled by the sedimentary pattern and the sedimentary thickness, and is expanded between wells in a probabilistic form through Kriging interpolation, which is more consistent with the actual thin interlayer characteristics; the probabilistic geological constraint on the sand-to-land ratio obtained can be combined with a seismic attribute interpretation model through a Bayesian framework, and compared with the prediction of the sand-to-land ratio using only seismic attributes, the interpretation accuracy is greatly improved, and the geological interpretability is enhanced.

[0004] The present invention is achieved through the following technical solutions:

[0005] A probabilistic geological analysis method for thin interbedded sand-to-soil ratio in a drilling soil layer comprises the following steps:

[0006] Massive geological models are constructed based on the sedimentary laws and sedimentary phase thickness of thin inter-reservoirs in the wells;

[0007] Calculate the probability distribution characteristics of sand-to-ground ratio of massive geological models corresponding to different well locations;

[0008] The Kriging interpolation method is used to calculate the probability distribution characteristics of sand-to-land ratio of massive geological models at different well locations to construct probabilistic geological constraints on sand-to-land ratio;

[0009] Generate seismic attribute interpretation model based on propagation matrix forward modeling and sand-to-soil ratio probabilistic geological constraints corresponding to massive geological models;

[0010] The following Bayesian framework is used to integrate the probabilistic geological constraints and seismic attribute model to obtain accurate earthquake prediction of sand-to-ground ratio:

[0011] ψ=argmax(p(r|ω)p(ω))

[0012] Where: ψ represents the predicted sand-to-ground ratio; r is the seismic attribute; ω corresponds to the sand-to-ground ratio; p(r|ω) represents the two-dimensional probability density function obtained in (3); and p(ω) is the probability distribution of the geological constraints corresponding to all sand-to-ground ratios obtained in (2).

[0013] Furthermore, the massive geological model is quantified according to the depositional law and sedimentary phase thickness of the thin inter-reservoir in the well:

[0014] Quantification of sedimentary patterns of drilled thin interlayer reservoirs: Based on the interpretation results of well logging sedimentary facies in the work area, the sedimentary characteristics of the specific thin interlayer reservoirs are described using a one-dimensional Markov chain model, and the essential characteristics of the Markov chain are generated by quantifying the sedimentary pattern using the transition probability matrix. The transition probability matrix P is expressed as:

[0015]

[0016] Among them: element P ij (i-th row, j-th column) represents the transition probability from state i to state j;

[0017] Massive geological model construction and sand-to-ground ratio probability distribution: The initial state is continuously adjusted according to the transfer probability matrix P obtained from different well locations to realize massive geological sedimentary sequence simulation; the thickness distribution of each sedimentary phase in the well is further statistically analyzed, the corresponding cumulative distribution function is calculated and randomly sampled and assigned to the sequence simulation results to realize massive geological model construction, and finally the sand-to-ground ratio probability distribution characteristics of the massive geological model corresponding to different well locations are calculated.

[0018] Furthermore, the seismic attribute interpretation model is generated by the propagation matrix forward modeling method and the sand-to-ground ratio probabilistic geological constraints corresponding to the massive geological model:

[0019] The elastic parameter distribution of each sedimentary phase in the logging section corresponding to the formation is statistically analyzed, and the 0° incident PP wave reflection seismic record of the propagation matrix method forward algorithm and 90° phase Ricker wavelet simulation model is assigned to the massive geological model;

[0020] The propagation matrix forward modeling method can simulate the effect of thin layer thickness on reflection coefficient, as well as interlayer energy loss and multiple wave effects; the 90° phase-shifted wavelet can realize the symmetrical response of the reservoir and waveform; the 90° phase-shifted Ricker wavelet can be obtained by Hilbert transforming the zero-phase wavelet; in the case of P wave incidence, the thin layer reflection coefficient and transmission coefficient r = [R PP ,R RS ,T PP ,T PS ] T as follows:

[0021] r=-(A 1 -BA 2 ) -1 i P

[0022] Among them A 1 , A 2 and B represent the propagation matrices of the upper, lower and middle thin layers, i P represents the incident vector;

[0023] Based on the propagation matrix forward modeling method, the minimum amplitude and arc length attributes of the massive geological model are extracted as sensitive attributes of sand-to-sand ratio, and the two-dimensional conditional probability density function corresponding to different sand-to-sand ratios is constructed using kernel function non-parametric probability density calculation.

[0024] Among them, the kernel function is used to smooth the seismic attribute values ​​corresponding to different sand-to-ground ratios, and the Gaussian kernel function is used as a filtering template. The Gaussian kernel function value at the position (i, j) can be expressed as:

[0025]

[0026] Where: σ represents the standard deviation, and the filter template size is (2k+1)*(2k+1).

[0027] Beneficial Effects

[0028] 1. Compared with the deterministic sand-to-formation ratio distribution result obtained by conventional interpolation, the present invention obtains the probability distribution result corresponding to different sand-to-formation ratios. The well point position is controlled by the sedimentation mode and sedimentation thickness, and is expanded to the wells through Kriging interpolation in a probabilistic form, which is more consistent with the actual thin interbed characteristics; the probabilistic geological constraints of the sand-to-formation ratio obtained can be combined with the seismic attribute interpretation model through the Bayesian framework. Compared with the sand-to-formation ratio prediction using only seismic attributes, the interpretation accuracy is greatly improved and the geological interpretability is enhanced.

[0029] 2. The present invention provides a method for constructing and using probabilistic geological constraints on the sand-to-land ratio of thin interbeds controlled by sedimentary patterns, which belongs to geophysical exploration technology. This invention aims at the difficult problem of predicting the sand-to-land ratio of thin interbeds of sand and mud. By quantifying the sedimentary patterns of multiple wells and constructing massive models, the probability distribution of the sand-to-land ratio of multiple well points is constructed and the inter-well expansion is performed using Kriging interpolation, so as to realize the probabilistic geological constraints on the sand-to-land ratio controlled by sedimentary patterns; and further, as a priori information, the two-dimensional probability density function of the sensitive attributes for the sand-to-land ratio constructed based on the forward simulation of the massive model is combined to perform Bayesian statistical classification of the sand-to-land ratio. The fusion of probabilistic geological constraints and seismic attribute statistical characteristics can greatly improve the prediction accuracy of the sand-to-land ratio. The present invention is suitable for the construction of geological constraints and subsequent quantitative seismic interpretation of the sand-to-land ratio of thin interbeds with different geological characteristics, and is used to guide subsequent reservoir exploration and development work. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 The process of constructing and using the probabilistic geological constraints for thin interbedded sand-land ratio;

[0031] Figure 2 (a) Stratigraphic sedimentary sequence and (b) three examples of simulated geological models for the oil group of sample well 3; where light gray represents mudstone, medium gray represents sandy mudstone, dark gray represents muddy sandstone, and black represents sandstone;

[0032] Figure 3 Comparison of the sand-to-formation ratio statistics of 5000 geological models corresponding to the sedimentary pattern of sample well 3 (gray histogram and black solid probability density line) and the average sand-to-formation ratio of the actual formation (black dashed line);

[0033] Figure 4 Statistical histogram (gray column) and probability density function (black solid line) of the sand-to-ground ratio of the massive geological model corresponding to the sedimentary patterns of the six sample wells: (a) sample well 1; (b) sample well 2; (c) sample well 3; (d) sample well 4; (e) sample well 5; (f) sample well 6;

[0034] Figure 5 Probabilistic constraints for the sand-to-surface ratio of the target reservoir: (a) 10% sand-to-surface ratio; (b) 20% sand-to-surface ratio; (c) 30% sand-to-surface ratio; (d) 40% sand-to-surface ratio; (e) 50% sand-to-surface ratio;

[0035] Figure 6 Deterministic interpolation volume of sand-to-soil ratio of target reservoir;

[0036] Figure 7 (a) Geological model (light grey - mudstone, medium grey - sandy mudstone, dark grey - muddy sandstone, black - sandstone); (b) elastic parameters (P-wave velocity, S-wave velocity and density); (c) corresponding seismic records;

[0037] Figure 8(a) Intersection of the minimum amplitude and arc length attributes achieved according to the model results, with the color code being the sand-to-ground ratio; (b) Two-dimensional probability density functions of the minimum amplitude and arc length attributes corresponding to different sand-to-ground ratios;

[0038] Fig. 9 Target thin interbedded seismic attribute volume: minimum amplitude;

[0039] Fig.10 Target thin interbedded seismic attribute volume: arc length;

[0040] Fig.11 The prediction results of sand-to-soil ratio without probabilistic geological constraints;

[0041] Fig.12 The sand-to-soil ratio prediction results obtained after probabilistic geological constraints;

[0042] Fig.13 Comparison of the predicted sand-to-surface ratio and the actual drilled sand-to-surface ratio at the well location, where the light grey column represents the predicted sand-to-surface ratio before constraint, the dark grey column represents the predicted sand-to-surface ratio after constraint, and the black column represents the actual drilled sand-to-surface ratio;

[0043] Table 1 Markov chain transition probability matrix corresponding to the strata. DETAILED DESCRIPTION

[0044] The present invention provides a method for constructing and using probabilistic geological constraints for the sand-to-land ratio of thin interbeds in drilling soil layers. The method is to control the well point location by the sedimentation pattern and sedimentation thickness, and expand it to the wells in a probabilistic form through Kriging interpolation, which is more consistent with the actual thin interbed characteristics; the obtained probabilistic geological constraints for the sand-to-land ratio can be combined with the seismic attribute interpretation model through the Bayesian framework. Compared with the prediction of the sand-to-land ratio using only seismic attributes, the interpretation accuracy is greatly improved and the geological interpretability is enhanced.

[0045] like Figure 1 As shown, the present invention adopts the following steps:

[0046] (1) The sedimentary pattern of the thin interbedded reservoirs in the drilled wells is quantified, based on which a massive geological model of different well locations is constructed, and the probability distribution of the sand-to-soil ratio is statistically analyzed. The sedimentary pattern of the thin interbedded reservoirs is quantified, including the sedimentary law and the thickness of the sedimentary phase.

[0047] ① Quantification of sedimentary patterns of drilled thin interlayer reservoirs: Based on the interpretation results of well logging sedimentary phases in the work area, a one-dimensional Markov chain model is used to describe the sedimentary characteristics of specific thin interlayer reservoirs, and the sedimentary patterns are quantified using the transition probability matrix. The essential characteristics of the Markov chain can be expressed by its transition probability matrix P:

[0048]

[0049] Among them: element P ij(i-th row, j-th column) represents the transition probability from state i to state j.

[0050] ② Massive geological model construction and sand-to-land ratio statistics: The transfer probability matrix obtained for different well locations is used to simulate the massive geological sedimentary sequence by continuously adjusting the initial state; the thickness distribution of each sedimentary phase in the well is further statistically analyzed, the corresponding cumulative distribution function is calculated, and the sequence simulation results are randomly sampled to realize the construction of the massive geological model, ensuring that the sedimentary law and the thickness distribution of each lithofacies in the model are consistent with the actual stratigraphic characteristics, and improving the accuracy of model construction. Finally, the probability distribution characteristics of the sand-to-land ratio of the massive geological model corresponding to different well locations are calculated.

[0051] (2) Construction of probabilistic geological constraints on sand-to-formation ratio: Based on the sand-to-formation ratio distribution of multiple massive geological models corresponding to different well locations obtained in (1), the probability value corresponding to a certain sand-to-formation ratio is extracted, and the inter-well interpolation of the probabilities corresponding to different sand-to-formation ratios is realized by using the Kriging interpolation method. Finally, the probability distribution results corresponding to multiple sand-to-formation ratios can be obtained, and the probabilistic geological constraints on sand-to-formation ratios can be realized. This result can be used for the initial quantitative analysis of the uncertainty of the distribution of sand-to-formation ratios in the target reservoir and as a priori constraints for the subsequent accurate prediction of sand-to-formation ratios by combining seismic information. Compared with the deterministic sand-to-formation ratio distribution results obtained by conventional interpolation, this method obtains the probability distribution results corresponding to different sand-to-formation ratios. This probability distribution is controlled by the sedimentary mode and sedimentary thickness in (1), and is more consistent with the actual thin interbed characteristics.

[0052] (3) Construction of interpretation model based on propagation matrix forward modeling and interpretation model: Statistically calculate the elastic parameter distribution of each sedimentary phase in the logging section corresponding to the formation, and assign it to the 0° incident PP wave reflection seismic record in the model ② using the propagation matrix forward modeling algorithm and the 90° phase Ricker wavelet simulation model. The propagation matrix method can simulate the influence of thin layer thickness on the reflection coefficient, as well as the interlayer energy loss and multiple wave effects. The 90° phase-shifted wavelet can realize the symmetrical response of the reservoir and waveform. The 90° phase-shifted Ricker wavelet can be obtained by Hilbert transforming the zero-phase wavelet. In the case of P wave incidence, the thin layer reflection coefficient and transmission coefficient r=[R PP ,R RS ,T PP ,T PS ] T as follows:

[0053] r=-(A 1 -BA 2 ) -1 i P (2)

[0054] Among them A 1 , A 2 and B represent the propagation matrices of the upper, lower and middle thin layers, i P Represents the incident vector.

[0055] Based on the forward modeling records, the minimum amplitude and arc length attributes are extracted as sensitive attributes of the sand-to-sand ratio, and the two-dimensional conditional probability density function corresponding to different sand-to-sand ratios is constructed using the kernel function nonparametric probability density estimation. The kernel function is used to smooth the seismic attribute values ​​corresponding to different sand-to-sand ratios, and the Gaussian kernel function is used as a filtering template. The Gaussian kernel function value at the (i, j) position can be expressed as:

[0056]

[0057] Where σ represents the standard deviation, and the filter template size is (2k+1)*(2k+1).

[0058] (4) Bayesian framework joint interpretation: Construct a Bayesian framework to combine the probabilistic geological constraints in (2) and the sand-to-ground ratio interpretation model in (3) to make accurate seismic predictions of the sand-to-ground ratio. The introduction of probabilistic geological constraints can greatly improve the prediction accuracy of the sand-to-ground ratio. The Bayesian classification criterion is expressed as

[0059] ψ=argmax(p(r|ω)p(ω)) (4)

[0060] where ψ represents the predicted sand-to-ground ratio; r is the seismic attribute; ω corresponds to the sand-to-ground ratio; p(r|ω) represents the two-dimensional probability density function obtained in (3); and p(ω) is the probability distribution of the geological constraints corresponding to all sand-to-ground ratios obtained in (2).

[0061] The present invention provides a method for constructing and using a probabilistic geological constraint on the sand-to-land ratio of a thin interbedded soil layer in a drilling well. The method comprises a new probabilistic constraint construction method for the sand-to-land ratio based on a massive geological model that conforms to the law of sedimentation. Compared with the deterministic sand-to-land ratio distribution result obtained by conventional interpolation, the method obtains a probabilistic distribution result corresponding to different sand-to-land ratios. The well point position is controlled by the sedimentation mode and sedimentation thickness, and is expanded between wells in a probabilistic form through Kriging interpolation, which is more consistent with the actual thin interbed characteristics. The obtained probabilistic geological constraint on the sand-to-land ratio can be combined with a seismic attribute interpretation model through a Bayesian framework. Compared with the prediction of the sand-to-land ratio using only seismic attributes, the interpretation accuracy is greatly improved and the geological interpretability is enhanced.

[0062] The specific operation process is as follows Figure 1 shown.

[0063] (1) Eight typical wells were selected in the work area, six of which were used as sample wells and two as verification wells. The lithology and grain size characteristics of the selected (thin interlayer) target layers were first analyzed and the sedimentary facies were divided accordingly. The sedimentary sequence of sample well 3 is as follows: Figure 2 As shown in a, light gray represents mudstone, medium gray represents sandy mudstone, dark gray represents muddy sandstone, and black represents sandstone.

[0064] ① Based on this, the sedimentary characteristics description based on the one-dimensional Markov chain model was carried out, and the transition probability matrix (shown in Formula 1) was statistically obtained. The transition probability matrix of sample well 3 is shown in Table 1. Table 1 Markov chain transition probability matrix corresponding to the formation.

[0065] Transition probability matrix Mudstone Sandy mudstone Mudstone sandstone Mudstone 0 1 0 0 Sandy mudstone 0.5 0 0.3 0.2 Mudstone 0 1 0 0 sandstone 0 1 0 0

[0066] Here, only lithology transfer is considered without considering thickness characteristics, so the transfer probability (diagonal elements) of the same sedimentary phase is 0. The numerical distribution of elements in the matrix controls the directionality of deposition and can accurately characterize the main sedimentary mode of the formation. By continuously adjusting the initial state, a massive simulation without thickness information that conforms to this sedimentary mode can be achieved. 5,000 random simulations were performed on the target layer sections of the 6 sample wells, and a total of 30,000 sequence models were recorded.

[0067] ②The next step is to assign thickness characteristics to the geological model. The thickness distribution of each sedimentary phase in the logging section corresponding to the target layer of the 6 sample wells is statistically analyzed to obtain the corresponding cumulative distribution function, and the corresponding sedimentary phase thickness value is randomly sampled for the cumulative probability. It is assigned to the corresponding sedimentary phase of the model in ① to realize the sedimentary mode and thickness simulation of the geological model, ensure that the thickness distribution of each lithofacies in the model conforms to the actual stratigraphic characteristics, and improve the accuracy of model construction. Ultimately, a large number of geological models that conform to the laws of stratigraphic deposition and the thickness distribution of sedimentary phases can be realized. Figure 2 b shows three examples of geological models randomly constructed using the sedimentation pattern and total sediment thickness distribution of the target interval of the sample well 3. Figure 3 The sand-to-formation ratio distribution of 5000 random models corresponding to sample well 3 is compared with the actual drilling sand-to-formation ratio of the well target layer, which shows good consistency and verifies the accuracy of the random model.

[0068] (2) The target reservoirs of the six sample wells are processed as described in (1) to obtain the sand-to-land ratio distribution of the massive geological model corresponding to the well locations, as shown in Figure 4 As shown. Further, the probability values ​​corresponding to a certain sand-to-land ratio (10%, 20%, 30%, 40% and 50%) are extracted from the sand-to-land ratio distribution corresponding to the 6 sample wells as the probability values ​​corresponding to the sand-to-land ratio at each sample well point, and the Kriging interpolation method is used to realize the inter-well probability interpolation, and finally the probability distribution results corresponding to the sand-to-land ratio can be obtained, as shown respectively. Figure 5 As shown in the figure, the probabilistic geological constraints of sand-to-land ratio are finally realized. The deterministic sand-to-land ratio distribution results obtained by conventional interpolation are as follows: Figure 6As shown, the actual drilling sand-to-formation ratio at the well point location of the target reservoir has limited characterization ability for the distribution and change law of the sand-to-formation ratio at the location near the well, and the change between wells can generally only be adjusted by the interpolation algorithm parameters, which is greatly affected by the interpolation algorithm itself. The present invention obtains the probability distribution results corresponding to different sand-to-formation ratios, and the sand-to-formation ratio at any spatial point in the target reservoir can be described by the probability distribution. It can be used for the initial uncertainty quantitative analysis of the distribution of the sand-to-formation ratio of the target reservoir and as a priori constraint for the subsequent accurate prediction of the sand-to-formation ratio by combining seismic and other information. The probability distribution is controlled by the sedimentation mode and sedimentation thickness in (1), which is more consistent with the actual thin interbed characteristics, and has a wider range of applications than the deterministic sand-to-formation ratio interpolation results.

[0069] (3) The elastic parameter distribution of each sedimentary phase in the logging section corresponding to the formation is statistically analyzed and randomly assigned to the corresponding sedimentary phase in the model (30,000) constructed in (1), and the propagation matrix method forward algorithm (shown in Formula 2) and the 90° phase Ricker wavelet are further used to simulate the 0° incident PP wave reflection seismic record of each model. The 90° phase-shifted Ricker wavelet can be obtained by Hilbert transforming the zero-phase wavelet. Figure 7 A random geological model, elastic parameter distribution and corresponding seismic records are shown. The troughs and sand bodies have a good correspondence, which verifies the accuracy of the method. Based on the forward modeling records, the minimum amplitude and arc length attributes are extracted as sand-to-sand sensitive attributes. The intersection relationship is shown in Figure 8 The kernel function nonparametric probability density estimation (Formula 3) can be used to construct the two-dimensional conditional probability density function corresponding to different sand-to-land ratios, as shown in Fig. Figure 8 As shown in b.

[0070] (4) The Bayesian framework (Formula 4) is combined with the probabilistic geological constraints in (2) and the two-dimensional conditional probability density function interpretation model of the sand-to-ground ratio in (3) to perform accurate seismic prediction of the target thin interbedded sand-to-ground ratio. The minimum amplitude attribute volume and arc length attribute volume are extracted as follows: Fig. 9 , Fig.10 The prediction results of sand-to-land ratio completely relying on the two-dimensional conditional probability density function are shown in Fig.11 As shown in the figure, the prediction results of sand-to-land ratio obtained by combining probabilistic geological constraints are as follows: Fig.12 As shown in Figure 2, the predicted sand-to-formation ratio of the sample well and the verification well is compared with the actual sand-to-formation ratio. Fig.13 As shown in the figure, the prediction results and actual drilling results of the sand-to-formation ratio of the sample wells and verification wells verified the effectiveness of the technology. The overall average prediction error was reduced from 36% to 12%. It can be seen that the introduction of probabilistic geological constraints greatly improved the prediction accuracy of the sand-to-formation ratio.

Claims

1. A probabilistic geological analysis method for the ratio of thin interbedded sand to soil in drilling soil. It is characterized in that The steps include: Massive geological models are constructed based on the sedimentary laws and sedimentary phase thickness of thin inter-reservoirs in the wells; Calculate the probability distribution characteristics of sand-to-ground ratio of massive geological models corresponding to different well locations; The Kriging interpolation method is used to calculate the probability distribution characteristics of sand-to-land ratio of massive geological models at different well locations to construct probabilistic geological constraints on sand-to-land ratio; Based on the propagation matrix forward modeling method and the sand-to-sand ratio probabilistic geological constraints corresponding to the massive geological model, the seismic attribute interpretation model is calculated and generated; wherein: the seismic attribute interpretation model is generated by the propagation matrix forward modeling method and the sand-to-sand ratio probabilistic geological constraints corresponding to the massive geological model: The elastic parameter distribution of each sedimentary phase in the logging section corresponding to the formation is statistically analyzed, and the 0° incident PP wave reflection seismic record of the propagation matrix method forward algorithm and 90° phase Ricker wavelet simulation model is assigned to the massive geological model; The propagation matrix forward modeling method can simulate the effect of thin layer thickness on reflection coefficient, as well as interlayer energy loss and multiple wave effects; the 90° phase-shifted wavelet can realize the symmetrical response of the reservoir and waveform; the 90° phase-shifted Ricker wavelet can be obtained by Hilbert transforming the zero-phase wavelet; in the case of P wave incidence, the thin layer reflection coefficient and transmission coefficient r = [R PP ,R RS ,T PP ,T PS ] T as follows: r=-(A 1 -BA 2 ) -1 i P Among them A 1 , A 2 and B represent the propagation matrices of the upper, lower and middle thin layers respectively, i P represents the incident vector; Based on the propagation matrix forward modeling method, the minimum amplitude and arc length attributes of the massive geological model are extracted as sensitive attributes of sand-to-sand ratio, and the two-dimensional conditional probability density function corresponding to different sand-to-sand ratios is constructed using kernel function non-parametric probability density calculation. Among them, the kernel function is used to smooth the seismic attribute values ​​corresponding to different sand-to-ground ratios, and the Gaussian kernel function is used as a filtering template. The Gaussian kernel function value at the position G(i,j) can be expressed as: Where: σ represents the standard deviation, and the filter template size is (2k+1)*(2k+1); The following Bayesian framework is used to integrate the probabilistic geological constraints and seismic attribute model to obtain accurate earthquake prediction of sand-to-ground ratio: ψ=argmax(p(r|ω)p(ω)) Where: ψ represents the predicted sand-to-ground ratio; r is the seismic attribute; ω corresponds to the sand-to-ground ratio; p(r|ω) represents the two-dimensional probability density function; p (ω) Represents the probability distribution of geological constraints corresponding to all sand-to-land ratios obtained.

2. A probabilistic geological analysis method for the thin interlayer sand-to-soil ratio in a drilling soil layer according to claim 1, It is characterized in that The massive geological model is quantified according to the depositional law and sedimentary phase thickness of the thin inter-reservoir in the well: Quantification of sedimentary patterns of drilled thin interlayer reservoirs: Based on the interpretation results of well logging sedimentary facies in the work area, the sedimentary characteristics of the specific thin interlayer reservoirs are described using a one-dimensional Markov chain model, and the essential characteristics of the Markov chain are generated by quantifying the sedimentary pattern using the transition probability matrix. The transition probability matrix P is expressed as: Among them: element P ij represents the transition probability from state i to state j; Massive geological model construction and sand-to-ground ratio probability distribution: The initial state is continuously adjusted according to the transfer probability matrix P obtained from different well locations to realize massive geological sedimentary sequence simulation; the thickness distribution of each sedimentary phase in the well is further statistically analyzed, the corresponding cumulative distribution function is calculated and randomly sampled and assigned to the sequence simulation results to realize massive geological model construction, and finally the sand-to-ground ratio probability distribution characteristics of the massive geological model corresponding to different well locations are calculated.

Images