Method for evaluating uncertainty of reservoir property parameter prediction

Abstract

The application provides a reservoir physical property parameter prediction uncertainty evaluation method, comprising the following steps: step 1, establishing a statistical rock physics model of a target area, thereby deriving a conditional probability of seismic attributes; step 2, performing statistical analysis on existing physical property parameter data, and establishing a corresponding physical property parameter prior distribution; step 3, based on the statistical rock physics model and the prior probability distribution obtained in steps 1 and 2, using a physical property parameter simulation sampling algorithm of facies variation, performing facies prediction and generating a physical property parameter sample set of the target area; and step 4, according to the physical property parameter sample set obtained in step 3, estimating the expectation, variance and confidence interval of the physical property parameter. The reservoir physical property parameter prediction uncertainty evaluation method realizes the statistical expectation, variance and confidence interval estimation of the shale content, porosity and saturation in the facies variation reservoir, further evaluates the potential risk of the target reservoir development, and provides technical support for the risk analysis of oil and gas exploration and development.

Description

Technical Field

[0001] This invention relates to the field of oilfield development technology, and in particular to an uncertainty evaluation method for predicting reservoir physical parameters. Background Technology

[0002] Seismic methods are among the most effective techniques in oil and gas exploration. Researchers can extract key attributes and important parameters from seismic data that effectively reflect subsurface conditions. Among these, physical properties (clay content, porosity, and saturation) are crucial for describing reservoir characteristics. In practice, due to unavoidable random factors such as noise in data acquisition and assumption biases in seismic interpretation, estimating reservoir physical properties directly or indirectly using seismic data always faces uncertainty. To quantitatively represent this uncertainty, we need to combine rock physics models and statistical analysis methods to establish probabilistic models of reservoir physical properties, develop matching simulation sampling algorithms, and ultimately develop a completely new set of techniques for predicting physical properties and evaluating their uncertainties.

[0003] In the field of seismic-based reservoir characterization, one type of approach is data-driven. These methods often rely on multivariate statistical methods and well logging data to establish probability distributions of key reservoir characteristics (Doyen 1988; Fournier 1989). Due to the lack of reliable physical interpretation and the simplistic structure of statistical models, these methods struggle to address the joint prediction of multiple parameters, including clay content, porosity, and saturation. Another type of approach is model-driven, which establishes empirical models based on rock physics theory or laboratory data, and then uses model inversion to achieve joint prediction of multiple parameters, including clay content, porosity, and saturation (Nie et al. 2004; Fang and Yang 2015). However, these methods are inherently deterministic, do not consider stochastic factors in the process, and cannot assess the uncertainty in the prediction.

[0004] Both of the aforementioned methods have achieved some degree of success; however, a more promising approach is to combine their advantages. In 2006, Bachrach, based on the common rock physics model (Gassmann model), introduced normal and uniform distributions to describe the randomness of seismic property data and well logging data, achieving joint estimation of porosity and gas saturation and analyzing the variance of the parameters. However, this method only applied to a single sandstone and ignored the influence of lithofacies variations. Since 2010, Grana et al., based on the Bayesian statistical framework, have established different statistical rock physics models, achieving predictions of clay content, porosity, and saturation, as well as uncertainty analysis (Grana and Rossa 2010; Figueiredo et al. 2017; Grana 2018; Li et al. 2020). However, these methods all directly estimate physical parameters based on seismic data, resulting in high computational costs for forward modeling and strong initial value dependence in inversion models. Therefore, the posterior probability simulation sampling process for physical property parameters often requires a lot of computation time and is difficult to traverse the probability support set, resulting in low efficiency and inaccuracy in uncertainty evaluation.

[0005] In summary, for reservoirs with facies changes, in order to accurately and efficiently predict physical property parameters and evaluate their uncertainties, it is necessary to establish complex rock physics models based on seismic attribute data and develop sampling algorithms for physical property parameters of facies changes.

[0006] Chinese patent application CN201510392504.7 discloses a quantitative prediction method for tight sandstone and conglomerate gas reservoirs, comprising the following steps: establishing a geological model of the actual geological characteristics of the study area; obtaining pre-stack geostatistical inversion parameters, impedance probability distribution, and variogram through well analysis within the study area; and calculating the geological body and probability body using the Markov chain Monte Carlo method based on the geological model, combined with lithofacies data, well logging data, and seismic data. This invention establishes geological and rock physical models of sandstone and conglomerate gas reservoirs, obtains well shear wave data, and obtains seismic sensitive parameters of the gas reservoir through analysis of reservoir characteristics. By applying the inversion technology of this invention to obtain specific geological bodies and probability bodies, it solves the technical problems of low seismic vertical resolution, overlapping impedances of the gas reservoir and surrounding rock, and low prediction accuracy of tight sandstone and conglomerate gas reservoirs; this invention also reduces the ambiguity of seismic reservoir prediction.

[0007] Chinese patent application CN201610707298.9 discloses a pre-stack seismic inversion prediction method for shale gas TOC (Total Organic Carbon). This method includes the following steps: Step 1, establishing a shale reservoir TOC inversion objective function; Step 2, pre-stack inversion of shale TOC based on elastic impedance: A prior distribution of the reservoir TOC is established based on statistical analysis of well logging data. Monte Carlo simulation technology is used to randomly sample the established prior distribution, ultimately obtaining the random sample spatial distribution of the reservoir TOC. The maximum posterior probability of the reservoir TOC is estimated, and the TOC value corresponding to the location of this maximum value is the final inversion result. This invention comprehensively applies Bayesian theory, statistical rock physics models, and Monte Carlo random sampling techniques, enabling the simultaneous inversion of several physical property parameters. This eliminates the limitations imposed by other parameters when inverting a single parameter, thereby enhancing the reliability of the inversion.

[0008] Chinese patent application CN201911138179.6 discloses a method for predicting reservoir physical parameters using deep learning. The steps are as follows: 1) Introducing MIC (Micro-Melting) to quantitatively measure the nonlinear correlation between physical parameters and well logging curves, selecting well logging curves that show a significant response to physical parameters; 2) Introducing CEEMDAN to decompose the physical parameter data sequence, obtaining intrinsic mode function (IMF) components and residual RES components, and stabilizing the physical parameter data sequence; 3) Introducing SE (Simplified Estimate) to evaluate the complexity of each IMF component and RES residual, recombining component sequences with similar entropy values ​​to obtain new intrinsic mode components; 4) Normalizing the new intrinsic mode component data and dividing it into training and testing sets; 5) Introducing an LSTM (Linguistic Recurrent Neural Network) to build a prediction model for the reconstructed new components, obtaining predicted values ​​for each new intrinsic mode component; 6) Inversely normalizing the predicted values ​​of each new intrinsic mode component and superimposing them to reconstruct the physical parameter prediction results. This invention reduces redundant information and the number of predicted component models, improving prediction accuracy and speed.

[0009] The existing technologies described above are significantly different from the present invention and have failed to solve the technical problem we want to address. Therefore, we have invented a new method for evaluating the uncertainty of reservoir physical property parameter prediction. Summary of the Invention

[0010] The purpose of this invention is to provide an uncertainty assessment method for predicting reservoir physical property parameters, which estimates the statistical expectation, variance, and confidence interval of clay content, porosity, and saturation in reservoirs with varying lithofacies, and thereby evaluates the potential risks of developing target reservoirs.

[0011] The objective of this invention can be achieved through the following technical measures: a method for evaluating the uncertainty of reservoir physical property parameter prediction, the method comprising:

[0012] Step 1: Establish a statistical rock physics model of the target area, and derive the conditional probability of seismic attributes from it;

[0013] Step 2: Perform statistical analysis on the existing physical property parameter data and establish the corresponding prior distribution of physical property parameters;

[0014] Step 3: Based on the statistical rock physics model and prior probability distribution obtained in Step 1 and Step 2, a petrographic property parameter simulation sampling algorithm for facies change is used to predict the petrographic features and generate a sample set of petrographic parameters for the target area.

[0015] Step 4: Based on the sample set of physical property parameters obtained in Step 3, estimate the expected value, variance, and confidence interval of the physical property parameters.

[0016] The objective of this invention can also be achieved through the following technical measures:

[0017] In step 1, the statistical rock physics model is composed of a deterministic rock physics model and random factors obtained from statistical analysis of well logging data. It is a model-data driven statistical model, and the conditional probabilities of the seismic attributes derived from it are easy to calculate.

[0018] In step 2, the prior distribution is obtained by statistical analysis of well logging data using classification or clustering algorithms, including the prior probability of lithofacies and the prior distribution of physical property parameters of fixed lithofacies; among them, the prior distribution of physical property parameters of fixed lithofacies is unimodal and easy to calculate and sample.

[0019] In step 3, the petrographic property parameter simulation sampling algorithm includes probability estimation based on the Monte Carlo method and complex probability density simulation based on Markov chains.

[0020] In step 3, the conditional probability of lithofacies is estimated using the Monte Carlo method based on the conditional expectation of physical property parameters.

[0021] In step 3, for the lithofacies with the highest probability, a Markov chain with uniform random walk is constructed based on the prior distribution of physical property parameters of the fixed lithofacies and the conditional probability of seismic attributes; the stationary distribution of the Markov chain is the posterior distribution of the physical property parameters, thereby generating a sufficient number of physical property parameter samples.

[0022] In step 3, for ease of writing, the statistical parameters are reservoir physical properties including clay content c, porosity φ, and water saturation s, denoted as l=(c,φ,s)∈[L l U l ], U l and L l These represent the upper and lower bounds of the physical property parameters in their physical sense; specific seismic attribute data are selected as observation data and denoted as d; lithofacies are denoted as f.

[0023] In step 3, according to Bayesian statistical methods, the posterior distribution p(l|d,f) of the physical property parameters of a fixed lithofacies is proportional to the product of the prior distribution p(l|f) of the corresponding physical property parameters and the conditional distribution p(d|l,f) of the seismic attributes, i.e.

[0024] p(l|d,f)∝p(d|l,f)p(l|f). (3)

[0025] The posterior probability p(f|d) of lithofacies can be expressed in the form of the conditional expectation of l|f, i.e.

[0026] p(f|d)∝p(f)E l|f [p(d|l,f)] (4)

[0027] Among them, E l|f Let l|f represent the conditional expectation.

[0028] The posterior probability of the lithofacies is calculated by the Monte Carlo method according to formula (6), thereby realizing the prediction of the target lithofacies; a random walk Markov chain is constructed according to formula (5) to simulate the posterior distribution of physical property parameters under the target lithofacies.

[0029] In step 3, the simulation sampling algorithm for the physical parameters of lithofacies changes specifically includes:

[0030] The first step is to analyze the k-th lithofacies f k According to the distribution p(l|f) k Generate N k A sample of physical property parameters {l i}, calculate lithofacies conditional probability

[0031]

[0032] The second step is to select P. fk At its maximum, the corresponding lithofacies f M ;

[0033] Third step, let the target distribution be F(l) = p(d|l,f) M )p(l|f M ), random walk step size δ, from the initial distribution p(l|f M Randomly generated l in ) (t) The current time of the Markov chain is t=0;

[0034] The fourth step is to generate candidate values ​​l. * =l (t) +s,s~Unif(-δ - (l (t) ),δ + (l (t) )),in

[0035]

[0036]

[0037] x = l (t) , The corrected walk steps represent the clay content, porosity, and water saturation, respectively.

[0038] Fifth step, calculate the probability p of the candidate value. t =min(R(l) (t) ,l * ),1), where

[0039]

[0040] Step 6: Generate physical property parameter samples l (t+1) =ql * +(1-q)l (t) , where q follows the parameter p t The 0-1 distribution;

[0041] Step 7: At the current time t = t + 1, return to step 4; continue until the Markov chain reaches the stopping time T, i.e., t > T, then stop running and output the physical property parameter sample set S. l ={l (0) ,l (1) ,...,l (T)};

[0042] Step 8, based on the physical property parameter sample set S l The sample mean is calculated to predict physical property parameters, the sample variance is calculated to quantitatively evaluate the risk of predicting physical property parameters, and the sample high and low quantiles are calculated to represent the possible range of variation of physical property parameters.

[0043] In step 4, let T0 represent the pre-firing period and the expected estimate of the clay content. Expected estimate of porosity And the expected estimate of water saturation They are respectively

[0044]

[0045] The uncertainty evaluation method for reservoir physical parameter prediction in this invention utilizes a deterministic rock physics model and random errors to characterize seismic attribute changes. It employs Bayesian statistical methods to introduce prior distributions to describe the randomness of physical parameters, constructs a statistical rock physics model, analyzes the posterior distribution of physical parameters, and develops an efficient hybrid probability model sampling algorithm by combining Markov chains and Monte Carlo algorithms. Finally, it estimates the corresponding statistical characteristics based on the generated sample sets of clay content, porosity, and saturation.

[0046] This invention fully considers the influence of random factors on the prediction of physical property parameters, making it closer to actual exploration conditions. Test results show that in oil and gas exploration, the method of this patent can simultaneously achieve three tasks: lithology identification, physical property parameter prediction, and uncertainty assessment, demonstrating strong reservoir identification and fluid identification capabilities. The uncertainty assessment method for reservoir physical property parameter prediction estimates the statistical expectation, variance, and confidence interval of clay content, porosity, and saturation in reservoirs with lithological variations, thereby evaluating the potential risks of developing the target reservoir. Based on rock physics modeling and statistical sampling, a novel method for predicting physical property parameters and assessing their uncertainty has been developed for reservoirs with lithological variations, thus providing technical support for risk analysis in oil and gas exploration and development. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of a physical property parameter simulation sampling algorithm in a specific embodiment of the present invention;

[0048] Figure 2 This is a schematic diagram of seismic attribute data in a specific embodiment of the present invention;

[0049] Figure 3 This is a schematic diagram of the lithofacies prediction results in a specific embodiment of the present invention;

[0050] Figure 4 This is a schematic diagram of the expected estimation and confidence interval estimation results of physical property parameters in a specific embodiment of the present invention;

[0051] Figure 5 This is a schematic diagram illustrating the estimation of variance of physical property parameters in a specific embodiment of the present invention;

[0052] Figure 6 This is a flowchart of a specific embodiment of the uncertainty evaluation method for predicting reservoir physical property parameters according to the present invention. Detailed Implementation

[0053] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0054] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, and / or combinations thereof.

[0055] The present invention provides a method for uncertainty evaluation in the prediction of reservoir physical parameters (clay content, porosity, and saturation), comprising: establishing a statistical rock physics model of the target area based on relevant geological and well logging data, combined with rock physics theory; establishing a prior model of the target area based on statistical analysis of existing physical parameter data; performing posterior probability simulation sampling of lithofacies and physical parameters based on the above statistical rock physics model and prior model, thereby generating a physical parameter sample set; and using statistical analysis methods to predict the reservoir physical parameters of the target area and evaluate their uncertainty based on the above sample set. This method, based on rock physics modeling and statistical sampling, develops a novel method for predicting physical parameters and evaluating their uncertainty in reservoirs with varying lithofacies, thus providing technical support for risk analysis in oil and gas exploration and development.

[0056] The following are several specific embodiments of the application of the present invention.

[0057] Example 1

[0058] In a specific embodiment 1 of the present invention, such as Figure 6 As shown, the uncertainty evaluation method for reservoir property parameter prediction of the present invention includes the following steps:

[0059] Step 1: Based on existing data, establish a statistical rock physics model for the target area, from which the conditional probability of seismic attributes can be derived.

[0060] The statistical rock physics model is composed of a deterministic rock physics model and random factors obtained from statistical analysis of well logging data. It is a model-data driven statistical model, and the conditional probabilities of the seismic attributes derived from it are easy to calculate.

[0061] Step 2: Perform statistical analysis on the existing physical property parameter data to establish the corresponding prior distributions of the physical property parameters. The prior distributions are obtained by statistical analysis of the logging data using classification (or clustering) algorithms, including the prior probabilities of lithofacies and the prior distributions of physical property parameters for fixed lithofacies. Among them, the prior distributions of physical property parameters for fixed lithofacies are unimodal and easy to calculate and sample.

[0062] Step 3: Based on the statistical rock physics model and prior probability distribution obtained in Step 1 and Step 2, a petrographic property parameter simulation sampling algorithm is used to predict the petrographic features and generate a sample set of petrographic parameters for the target area.

[0063] The sampling algorithms for simulating petrographic property parameters include probabilistic estimation based on the Monte Carlo method and complex probability density simulation based on Markov chains.

[0064] Based on the conditional expectation of physical property parameters, the Monte Carlo method is used to estimate the conditional probability of lithofacies.

[0065] For the lithofacies with the highest probability, a Markov chain with uniform random walk is constructed based on the prior distribution of physical property parameters of a fixed lithofacies and the conditional probability of seismic attributes. The stationary distribution of the Markov chain is the posterior distribution of the physical property parameters, thereby generating a sufficient number of physical property parameter samples.

[0066] Step 4: Based on the sample set of physical property parameters obtained in Step 3, estimate the expected value, variance, and confidence interval of the physical property parameters.

[0067] Example 2

[0068] In a specific embodiment 2 of the present invention, in order to solve the problems existing in the prior art, the theoretical methods adopted by the present invention mainly include: Bayesian statistical methods, Monte Carlo algorithms and MCMC theory.

[0069] For ease of writing, the statistical parameters are reservoir physical properties (clay content c, porosity φ, and water saturation s), denoted as l=(c,φ,s)∈[L l U l ]; Select specific seismic attribute data as observation data, denoted as d; lithology is denoted as f.

[0070] Bayesian statistical methods have been widely applied in oil and gas prediction and reservoir characterization. In this invention, Bayesian statistical methods effectively link the two key aspects of lithofacies identification and physical property parameter prediction, suppressing potential interference from lithofacies changes during the probabilistic simulation sampling of physical property parameters. According to Bayesian statistical methods, the posterior distribution p(l|d,f) of physical property parameters for a fixed lithofacies is proportional to the product of the prior distribution p(l|f) of the corresponding physical property parameter and the conditional distribution p(d|l,f) of the seismic attribute, i.e.

[0071] p(l|d,f)∝p(d|l,f)p(l|f). (5)

[0072] The posterior probability p(f|d) of lithofacies can be expressed in the form of the conditional expectation of l|f, i.e.

[0073] p(f|d)∝p(f)E l|f [p(d|l,f)] (6)

[0074] Using formula (6), the posterior probability of the lithofacies is calculated using the Monte Carlo method, thus predicting the target lithofacies. Furthermore, based on formula (5), we construct a random walk Markov chain to simulate the posterior distribution of physical property parameters under the target lithofacies. Figure 1 As shown, the specific algorithm flow for the simulation sampling algorithm of petrographic property parameters is as follows:

[0075] The first step is to analyze any lithofacies f k According to the distribution p(l|f) kGenerate N k A sample of physical property parameters {l i}, calculate lithofacies conditional probability

[0076]

[0077] The second step is to select P. fk At its maximum, the corresponding lithofacies f M ;

[0078] Third step, let the target distribution be F(l) = p(d|l,f) M )p(l|f M ), random walk step size δ, from the initial distribution p(l|f M Randomly generated l in ) (t) ,t=0;

[0079] The fourth step is to generate candidate values ​​l. * =l (t) +s,s~Unif(-δ - (l (t) ),δ + (l (t) )),in

[0080]

[0081]

[0082] Fifth step, calculate the probability p of the candidate value. t =min(R(l) (t) ,l * ),1), where

[0083]

[0084] Step 6: Generate physical property parameter samples l (t+1) =ql * +(1-q)l (t) , where q follows the parameter p t The 0-1 distribution;

[0085] Step 7: t = t + 1, return to step 4; continue until a sufficient number of physical property parameter samples are generated, i.e., t > T, then stop running and output the physical property parameter sample set S. l ={l (0) ,l (1) ,...,l (T)}

[0086] Step 8, based on the physical property parameter sample set S lWe calculate the sample mean to predict physical property parameters, calculate the sample variance to quantitatively evaluate the risk of predicting physical property parameters, and calculate the sample high and low quantiles to represent the possible range of variation of physical property parameters.

[0087] Example 3

[0088] In a specific embodiment 3 of the present invention, the present invention proposes an uncertainty evaluation method for reservoir physical property parameter prediction. Based on the reservoir parameter prediction, the method uses common seismic inversion attribute data (density, P-wave impedance and S-wave impedance) to estimate the reservoir clay content, porosity and water saturation by simulating the conditional distribution of physical property parameters and providing corresponding uncertainty analysis results.

[0089] First, a deterministic rock physics model needs to be established by referring to the properties of common minerals and fluids in the target area and relevant geological data. Then, based on existing well logging data, a statistical rock physics model is established by statistically analyzing the fitting error of the deterministic model, from which the conditional probability p(d|l,f) of the seismic attribute can be derived. Simultaneously, a classification (or clustering) algorithm is used to statistically analyze the existing physical property parameter data to determine the lithofacies category and its prior probability p(f), thereby establishing the corresponding prior distribution p(l|f) of the physical property parameters. It is important to note that we should select appropriate seismic attributes and establish an efficient rock physics model to ensure that the conditional probability p(d|l,f) of the seismic attribute is easy to calculate; when establishing the prior model, a suitable statistical model should be selected to ensure that the prior probability p(l|f) of the physical property parameters is unimodal and easy to calculate and sample.

[0090] Secondly, based on the aforementioned statistical rock physics model and prior probability distribution, the following approach is adopted: Figure 1 The illustrated petrographic variation physical property parameter simulation sampling algorithm enables petrographic prediction and generates a sample set of physical property parameters for the target area. Furthermore, based on this sample set, the expected value, variance, and confidence interval of the physical property parameters are estimated.

[0091] Figure 2 This is seismic inversion attribute data used to test the patented method. The data consists of Poisson's ratio, P-wave impedance, and S-wave impedance, simulated from well logging data with known physical property parameters. The black line represents data synthesized using a deterministic rock physics model, while the gray line represents data generated using a statistical rock physics model.

[0092] Figure 3 The images show the predicted lithofacies. The left image shows the actual lithofacies, and the right image shows the lithofacies predicted by the method of this patent. Gray represents sandstone, and black represents mudstone. Comparing the two images, it can be seen that sand bodies in the 2595-2600ms and 2625-2630ms ranges were correctly identified, indicating that the method of this patent can effectively identify sandstone.

[0093] Figure 4 (a)-(c) show the expected and confidence interval estimates of clay content, porosity, and water saturation, respectively. The solid black lines represent the actual physical property parameters, and the dashed black lines represent the expected estimates. As can be seen from the figures, the solid and dashed black lines are generally consistent, but not perfectly aligned in details. This indicates that quantitative predictions of physical property parameters always involve uncertainty, and simple expected estimates cannot characterize this uncertainty. The gray shaded area in the figures represents the 95% confidence interval of the physical property parameters. We find that the solid black line falls entirely within the gray shaded area, and the width of the shaded area varies. This demonstrates that the method of this patent can accurately estimate the confidence interval of physical property parameters, thereby providing the range of random variations in the physical property parameters and intuitively describing the uncertainty of the physical property parameters.

[0094] Figure 5 Figures (a)-(c) show the variance estimates for clay content, porosity, and water saturation, respectively. As shown, the predicted variance for water saturation is higher in the sandstone regions at 2595-2600 ms and 2625-2630 ms, indicating greater uncertainty in the prediction of water saturation in these regions. This aligns with qualitative observations in practical work. This demonstrates that the method presented in this patent can effectively estimate the variance of physical property parameters, thereby providing a quantitative evaluation index for the uncertainty of physical property parameter predictions.

[0095] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0096] Except for the technical features described in the specification, all other technologies are known to those skilled in the art.

Claims

1. A method for evaluating the uncertainty of reservoir physical property parameter prediction, characterized in that, The uncertainty assessment methods for the prediction of reservoir physical parameters include: Step 1: Establish a statistical rock physics model of the target area, and derive the conditional probability of seismic attributes from it; Step 2: Perform statistical analysis on the existing physical property parameter data and establish the corresponding prior distribution of physical property parameters; Step 3: Based on the statistical rock physics model and prior probability distribution obtained in Step 1 and Step 2, a petrographic property parameter simulation sampling algorithm for facies change is used to predict the petrographic features and generate a sample set of petrographic parameters for the target area. Step 4: Based on the sample set of physical property parameters obtained in Step 3, estimate the expected value, variance, and confidence interval of the physical property parameters.

2. The uncertainty evaluation method for reservoir property parameter prediction according to claim 1, characterized in that, In step 1, the statistical rock physics model is composed of a deterministic rock physics model and random factors obtained from statistical analysis of well logging data. It is a model-data driven statistical model, and the conditional probabilities of the seismic attributes derived from it are easy to calculate.

3. The uncertainty evaluation method for reservoir property parameter prediction according to claim 1, characterized in that, In step 2, the prior distribution is obtained by statistical analysis of well logging data using classification or clustering algorithms, including the prior probability of lithofacies and the prior distribution of physical property parameters of fixed lithofacies; among them, the prior distribution of physical property parameters of fixed lithofacies is unimodal and easy to calculate and sample.

4. The uncertainty evaluation method for reservoir property parameter prediction according to claim 1, characterized in that, In step 3, the petrographic property parameter simulation sampling algorithm includes probability estimation based on the Monte Carlo method and complex probability density simulation based on Markov chains.

5. The uncertainty evaluation method for reservoir property parameter prediction according to claim 4, characterized in that, In step 3, the conditional probability of lithofacies is estimated using the Monte Carlo method based on the conditional expectation of physical property parameters.

6. The uncertainty evaluation method for reservoir property parameter prediction according to claim 5, characterized in that, In step 3, for the lithofacies with the highest probability, a Markov chain with uniform random walk is constructed based on the prior distribution of physical property parameters of the fixed lithofacies and the conditional probability of seismic attributes; the stationary distribution of the Markov chain is the posterior distribution of the physical property parameters, thereby generating a sufficient number of physical property parameter samples.

7. The uncertainty evaluation method for reservoir property parameter prediction according to claim 6, characterized in that, In step 3, for ease of writing, the statistical parameters are reservoir physical properties including clay content c, porosity φ, and water saturation s, denoted as l=(c,φ,s)∈[L l U l ], U l and L l These represent the upper and lower bounds of the physical property parameters in their physical sense; specific seismic attribute data are selected as observation data and denoted as d; lithofacies are denoted as f.

8. The uncertainty evaluation method for reservoir property parameter prediction according to claim 7, characterized in that, In step 3, according to Bayesian statistical methods, the posterior distribution p(l|d,f) of the physical property parameters of a fixed lithofacies is proportional to the product of the prior distribution p(l|f) of the corresponding physical property parameters and the conditional distribution p(d|l,f) of the seismic attributes, i.e. p(l|d,f)∝p(d|l,f)p(l|f) (1) The posterior probability p(f|d) of lithofacies can be expressed in the form of the conditional expectation of l|f, i.e. p(f|d)∝p(f)E l|f [p(d|l,f)] (2) Among them, E l|f Denotes the conditional expectation of l|f; The posterior probability of the lithofacies is calculated by the Monte Carlo method according to formula (2) to realize the prediction of the target lithofacies; a random walk Markov chain is constructed according to formula (1) to simulate the posterior distribution of physical property parameters under the target lithofacies.

9. The uncertainty evaluation method for reservoir property parameter prediction according to claim 8, characterized in that, In step 3, the simulation sampling algorithm for the physical parameters of lithofacies changes specifically includes: The first step is to analyze the k-th lithofacies f k According to the distribution p(l|f) k Generate N k A sample of physical property parameters {l i }, calculate lithofacies conditional probability The second step is to select... At its maximum, the corresponding lithofacies f M ; The third step is to set the target distribution. Random walk step size δ, from initial distribution p(l|f M Randomly generated l in ) (t) =(c (t) ,φ (t) ,s (t) ), the current time of the Markov chain is t=0; The fourth step is to generate candidate values ​​l. * =l (t) +s,s~Unif(-δ - (l (t) ,δ + (l (t) ),in x = l (t) , These represent the corrected walk steps for clay content, porosity, and water saturation, respectively. Fifth step, calculate the probability p of the candidate value. t =min(R(l) (t) ,l * ),1), where Step 6: Generate physical property parameter samples l (t+1) =ql * +(1-q)l (t) , where q follows the parameter p t The 0-1 distribution; Step 7: The Markov chain is at its current time t = t + 1, return to step 4; until the Markov chain reaches its stopping time T, i.e., t > T, stop running and output the physical property parameter sample set S. l ={l (0) ,l (1) ,...,l (T) }; Step 8, based on the physical property parameter sample set S l The sample mean is calculated to predict physical property parameters, the sample variance is calculated to quantitatively evaluate the risk of predicting physical property parameters, and the sample high and low quantiles are calculated to represent the possible range of variation of physical property parameters.

10. The uncertainty evaluation method for reservoir property parameter prediction according to claim 1, characterized in that, In step 4, let T0 represent the pre-firing period and the expected estimate of the clay content. Expected estimate of porosity And the expected estimate of water saturation They are respectively

Images