A reservoir quantitative prediction method, electronic equipment, storage medium and device

Abstract

The application discloses a reservoir quantitative prediction method, electronic equipment, a storage medium and a device. The method comprises the following steps: establishing a double-factor control initial model based on seismic facies attribute information, fracture attribute information and low-frequency framework information of a target layer; performing synchronous compression transformation frequency division on seismic data of the target layer to obtain a plurality of different frequency band data bodies; sequentially performing Bayesian inversion on each frequency band data body in the order from low to high based on the double-factor control initial model and the frequency band data bodies; iteratively inverting the next frequency band data body based on the inversion result of the previous frequency band data body as the initial constraint; and completing the Bayesian inversion of all the frequency band data bodies to complete reservoir quantitative prediction. The application can effectively reduce the interference caused by horizontal layering seismic reflection on inversion, improve the accuracy of reservoir quantitative prediction, enhance the reservoir detail accuracy and lateral heterogeneity description capability, and provide a reliable reference for high-quality reservoir development area description and well location deployment.

Description

Technical Field

[0001] This invention belongs to the field of petroleum exploration technology, and more specifically, relates to a method for quantitative prediction of reservoirs, electronic equipment, storage medium and device. Background Technology

[0002] With continuous breakthroughs and deepening of oil and gas exploration technology and development, research objectives and tasks are also changing. Complex targets controlled by multiple factors are gradually becoming the main focus of current exploration and development. Multiple wells in the Sichuan Basin have confirmed the development of thin-layered shoal-facies fault-controlled carbonate reservoirs in marine strata. Shoal facies sedimentation is the material basis for reservoir formation, while faults are key to reservoir stimulation and enrichment. Resource assessments suggest that the Maokou Formation shoal-facies sedimentary superimposed faults in the Sichuan Basin cover a large area and possess certain exploration potential. However, these geological anomalies controlled by two factors are highly heterogeneous, thin, and deeply buried. Furthermore, the seismic imaging effect is affected by fault zones, resulting in insufficient seismic data resolution, high ambiguity of single attributes, and significant difficulty in quantitative reservoir prediction, severely restricting further exploration deployment in this field.

[0003] Several methods exist for quantitative reservoir prediction. ① The constrained sparse pulse inversion method involves low-pass filtering of the wellbore impedance curve to supplement the low-frequency components missing in the seismic data, and then fusing it with the relative impedance data volume calculated based on the seismic data to obtain the final absolute impedance volume indicating reservoir anomalies. This method provides inversion results faithful to the seismic data and is fast and efficient, but its resolution and effectiveness are affected by the quality of the seismic data, and its performance in identifying small- and medium-scale reservoirs is poor. ② The geostatistical inversion method uses a linear weighting method to perform local optimization estimation of geological variables, taking into account both thickness and physical property quantitative evaluation, which can effectively improve the resolution of reservoir identification. However, this type of inversion method is limited by the variogram function, has limited effect on improving low-frequency components, and the inversion results are severely affected by seismic reflection. ③ The nonlinear frequency division inversion method explores the amplitude response relationship at different frequencies. When the relationship intersects, frequency band statistics are used to establish a nonlinear mapping relationship between the well logging curve and the waveform to characterize geological targets. This method does not require seismic wavelets and an initial model, but it relies on well logging curves and calculation methods. Currently, the inaccuracy of deep well logging information and the low computational efficiency limit the application of this method. ④ The principle of the phase-controlled inversion method is to combine phase-controlled modeling with parameter inversion. By studying geomorphic and lithological facies and utilizing sequence stratigraphy control, an initial inversion low-frequency trend model is constructed, followed by reservoir parameter inversion under attribute control. Compared with non-phase-controlled deterministic inversion, this type of method can better identify the reservoir development range, but it still suffers from low resolution and lack of detail in characterizing internal structure and physical property changes. Therefore, for special thin-layer geological anomalies under dual-factor control, conventional inversion methods are unable to reflect the small-scale seismic information, which directly affects the prediction accuracy of the reservoir and adds risks to oil and gas exploration.

[0004] The information disclosed in the background section of this invention is intended only to enhance the understanding of the general background of this invention, and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to propose a quantitative reservoir prediction method, electronic device, storage medium, and apparatus, which effectively reduces the interference caused by horizontal layered seismic reflections on inversion, improves the accuracy of quantitative reservoir prediction, enhances the accuracy of reservoir details and the ability to characterize lateral heterogeneity, and provides a reliable reference for the description of high-quality reservoir development areas and well location deployment.

[0006] To achieve the above objectives, the present invention proposes a reservoir quantitative prediction method, electronic device, storage medium, and apparatus.

[0007] According to a first aspect of the present invention, a method for quantitative prediction of reservoirs is proposed, comprising:

[0008] A two-factor control initial model is established based on the seismic phase attribute information, fault phase attribute information, and low-frequency grid information of the target layer;

[0009] The seismic data of the target layer are synchronously compressed, transformed, and frequency-divided to obtain multiple data volumes in different frequency bands;

[0010] Based on the two-factor control initial model and each frequency band data volume, Bayesian inversion is performed sequentially in order of frequency band from low to high. The inversion result of the previous frequency band data volume is used as the initial constraint for the next frequency band data volume for iterative inversion until all frequency band data volumes have completed the Bayesian inversion, thus completing the quantitative prediction of the reservoir.

[0011] Optionally, the expression for the two-factor control initial model is:

[0012] Z(u0)=Merge{αM(u0),βX(u0),γY(u0)};

[0013] Where α, β, and γ are the weighting coefficients for low-frequency framework information, sedimentary facies attribute information, and fracture facies attribute information, respectively; M, X, and Y represent low-frequency framework information, sedimentary facies attribute information, and fracture facies attribute information, respectively; u0 is the data point to be calculated; and Merge is the calculation function that obtains and merges data from multiple datasets.

[0014] Optionally, the step of synchronously compressing and transforming the seismic data of the target layer to obtain multiple data volumes in different frequency bands includes:

[0015] Seismic signals are acquired based on seismic data from the target layer;

[0016] Perform continuous wavelet transform on the seismic signal to obtain the wavelet transform coefficients;

[0017] Calculate all instantaneous frequencies of the seismic signal based on the wavelet transform coefficients;

[0018] Based on all the instantaneous frequencies, coordinate mapping is performed to transform the wavelet transform coefficients from the time-scale domain to the time-frequency domain;

[0019] Frequency bands are divided based on the time-frequency domain to obtain multiple different frequency bands;

[0020] The center frequency of each frequency band is determined based on the instantaneous frequency;

[0021] Based on the center frequency, each frequency band is synchronously compressed and transformed to obtain multiple corresponding time-frequency distributions;

[0022] Perform inverse transformation on multiple time-frequency distributions to obtain multiple data volumes in different frequency bands.

[0023] Optionally, the Bayesian inversion includes:

[0024] The prior probability and likelihood function are obtained based on the multiple frequency band data volumes, the two-factor control initial model, and the inversion result of the previous frequency band data volume.

[0025] Based on the prior probability and likelihood function, the posterior probability distribution is obtained within the Bayesian framework.

[0026] Based on the posterior probability distribution, multiple maxima of the posterior probability are obtained;

[0027] The average of the multiple maxima is taken as the inversion result of each Bayesian inversion.

[0028] Optionally, the expressions for obtaining the prior probability and the likelihood function are respectively:

[0029]

[0030] Where s is the frequency band data volume; z is the impedance data of the initial model under two-factor control or the inversion result of the previous frequency band data volume; m is the seismic wavelet length; L is the number of sampling points in the frequency band data volume; G is the Jacobian matrix obtained by taking the partial derivative of the frequency band data volume with respect to impedance; and C... n and C z Let represent the covariance matrices of the initial models for random noise and two-factor control, respectively; exp be the natural exponential function; T be the matrix transpose sign; p(s|z) be the prior probability; and p(z) be the likelihood function.

[0031] Optionally, the expression for the posterior probability distribution is:

[0032]

[0033] Where p(z|s) is the posterior probability distribution. and These are the variance matrices of the initial models for random noise and two-factor control, respectively.

[0034] Optionally, the expression for calculating the maximum value of the posterior probability is:

[0035]

[0036] in, This represents the minimum value of the objective function J, which is also the maximum value of the posterior probability.

[0037] According to a second aspect of the present invention, a reservoir quantitative prediction device is provided, comprising:

[0038] A module is established to create a two-factor control initial model based on seismic facies attribute information, fault facies attribute information, and low-frequency grid information of the target layer.

[0039] The synchronous compression transform frequency division module is used to perform synchronous compression transform frequency division on the seismic data of the target layer to obtain multiple data volumes in different frequency bands;

[0040] The prediction module is used to perform Bayesian inversion sequentially based on the two-factor control initial model and each frequency band data volume in ascending order of frequency band. It iterates the inversion based on the inversion result of the previous frequency band data volume as the initial constraint for the next frequency band data volume until all frequency band data volumes have completed the Bayesian inversion, thus completing the quantitative prediction of the reservoir.

[0041] According to a third aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0042] At least one processor; and,

[0043] A memory communicatively connected to the at least one processor; wherein,

[0044] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the reservoir quantitative prediction method according to any of the first aspects.

[0045] According to a fourth aspect of the invention, a non-transitory computer-readable storage medium is provided, which stores computer instructions for causing a computer to perform the reservoir quantitative prediction method described in any of the first aspects.

[0046] The beneficial effects of this invention are as follows: This invention combines seismic facies attribute information and fault facies attribute information to identify the longitudinal and lateral distribution characteristics of facies-controlled thin reservoirs under fault influence. Furthermore, it weights and fuses the seismic facies attribute information and fault facies attribute information with low-frequency framework information to establish a two-factor controlled initial model. This enables the enhancement and complementarity of various data, overcomes the problem of multiple solutions for single attributes, indicates the geometric contour information of reservoir distribution, and provides crucial facies-controlled constraint boundary information for seismic inversion. To overcome the drawback of conventional inversion methods not fully utilizing high-frequency seismic data, it performs synchronous compression transformation and frequency division on seismic data to obtain multiple data volumes in different frequency bands. This invention provides a data foundation for subsequent inversions. By utilizing a two-factor controlled initial model and multiple data volumes of different frequency bands in a Bayesian framework for iterative inversions, the inversion results of lower frequency bands are incorporated into the next higher frequency band inversion, enabling information from each frequency band to be transferred step by step. The final inversion results contain information from all frequency bands of the seismic data, reducing inversion uncertainty and improving the reliability of quantitative reservoir prediction. This invention can effectively reduce the interference caused by horizontal layered seismic reflections on inversions, improve the accuracy of quantitative reservoir prediction, enhance the accuracy of reservoir details and the ability to characterize lateral heterogeneity, and provide a reliable reference for the description of high-quality reservoir development areas and well location deployment.

[0047] The system of the present invention has other features and advantages that will be apparent from or will be set forth in detail in the accompanying drawings and following detailed description, which together serve to explain the particular principles of the invention. Attached Figure Description

[0048] The above and other objects, features and advantages of the present invention will become more apparent from the accompanying drawings, in which like reference numerals generally denote like parts.

[0049] Figure 1 A flowchart illustrating the steps of a reservoir quantitative prediction method according to the present invention is shown.

[0050] Figure 2 A flowchart illustrating the steps of a reservoir quantitative prediction method according to Embodiment 2 of the present invention is shown.

[0051] Figure 3 A schematic diagram showing the results of a synthetic seismic simulation signal and a synchronous compressed time-frequency analysis method according to Embodiment 2 of the present invention is illustrated.

[0052] Figure 4 A schematic diagram showing the result of reconstructing the original signal by selecting a corresponding frequency according to Embodiment 2 of the present invention is illustrated.

[0053] Figure 5 A schematic diagram of a two-dimensional geological model according to Embodiment 2 of the present invention is shown.

[0054] Figure 6 A schematic diagram of a two-dimensional forward inverse polarity profile according to Embodiment 2 of the present invention is shown.

[0055] Figure 7 A schematic diagram of the wave trough amplitude properties reflecting the beach facies distribution according to Embodiment 2 of the present invention is shown.

[0056] Figure 8 A schematic diagram of the dip curvature property reflecting the fracture distribution according to Embodiment 2 of the present invention is shown.

[0057] Figure 9 A schematic diagram of the seismic attribute envelope according to Embodiment 2 of the present invention is shown.

[0058] Figure 10 A schematic diagram of the initial model for two-factor control according to Embodiment 2 of the present invention is shown.

[0059] Figure 11 A schematic diagram of the original seismic profile according to Embodiment 2 of the present invention is shown.

[0060] Figure 12 A schematic diagram of a low-frequency band data profile, a mid-frequency band data profile, and a high-frequency band data profile according to Embodiment 2 of the present invention is shown.

[0061] Figure 13 A comparison diagram is shown between the original seismic data spectrum of Embodiment 2 according to the present invention and the seismic data spectrum obtained by the reservoir quantitative prediction method of this embodiment.

[0062] Figure 14 A schematic diagram of the low-frequency inversion results under the constraints of the two-factor control model according to Embodiment 2 of the present invention is shown.

[0063] Figure 15 A schematic diagram of the mid-frequency inversion result after low-frequency constraint according to Embodiment 2 of the present invention is shown.

[0064] Figure 16 A schematic diagram of the high-frequency inversion result after mid-frequency constraint according to Embodiment 2 of the present invention is shown.

[0065] Figure 17 A schematic diagram of the conventional inversion results without factor constraints according to Embodiment 2 of the present invention is shown. Detailed Implementation

[0066] The invention will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0067] like Figure 1 As shown, a reservoir quantitative prediction method according to the present invention includes:

[0068] A two-factor control initial model is established based on the seismic phase attribute information, fault phase attribute information, and low-frequency grid information of the target layer;

[0069] The seismic data of the target layer are synchronously compressed, transformed, and frequency-divided to obtain multiple data volumes in different frequency bands;

[0070] Based on the two-factor control initial model and each frequency band data volume, Bayesian inversion is performed sequentially in order of frequency band from low to high. The inversion result of the previous frequency band data volume is used as the initial constraint for the next frequency band data volume for iterative inversion until all frequency band data volumes have completed Bayesian inversion, thus completing the quantitative prediction of the reservoir.

[0071] Specifically, this invention extracts seismic facies attribute information and fault facies attribute information from seismic data of the target layer. Combining these two information, the longitudinal and lateral distribution characteristics of facies-controlled thin reservoirs under fault influence can be identified. Furthermore, the attribute information is weighted and fused with low-frequency framework information to establish a two-factor controlled initial model. This model enhances and complements the data, overcomes the multi-solution problem of single attributes, indicates the geometric contour information of the reservoir distribution, and provides crucial facies-controlled constraint boundary information for seismic inversion. The low-frequency framework information is obtained by ordinary kriging interpolation of the low-frequency impedance information of the input well within the stratigraphically constrained sequence body. In the formula, W(u) i ) is the low-frequency impedance value at a known input well point in space, λ iThe optimal coefficients satisfy the conditions of minimizing the error between unknown and known samples and unbiased estimation. M(u0) represents the low-frequency lattice information obtained by Kriging interpolation, and n is the number of optimized sample points, i = 1, 2, ..., n. To overcome the drawback of conventional inversion methods not being able to fully utilize the high frequencies of earthquakes, this invention performs synchronous compression transformation and frequency division on the seismic data of the target layer to obtain multiple data volumes in different frequency bands. According to the frequency range of the actual seismic data, this invention forms multiple seismic data volumes in different frequency bands to improve the identification capability of seismic data. Multiple data volumes in different frequency bands are obtained through synchronous compression transform and frequency division. Under the joint constraints of the initial model and the Bayesian framework, Bayesian inversion is performed on each frequency band data volume in ascending order of frequency. The seismic inversion results of lower frequency band data volumes are used in the seismic inversion of the next frequency band data volume. Specifically, the first Bayesian inversion uses the impedance data of the initial model controlled by two factors and the highest-ranked frequency band data volume as input. The second Bayesian inversion uses the derivative results of the lowest-ranked frequency band data volume and the second-ranked frequency band data volume as input, and so on. This allows the seismic information of each frequency band to be transmitted step by step. The final inversion result contains all frequency band information of the seismic data, reducing the uncertainty of the inversion and improving the reliability of quantitative reservoir prediction. This invention can effectively reduce the interference caused by horizontal layered seismic reflections on the inversion, improve the accuracy of quantitative reservoir prediction, enhance the accuracy of reservoir details and the ability to characterize lateral heterogeneity, and provide a reliable reference for the description of high-quality reservoir development areas and well location deployment.

[0072] In one example, the expression for the two-factor control initial model is:

[0073] Z(u0)=Merge{αM(u0),βX(u0),γY(u0)};

[0074] Where α, β, and γ are the weighting coefficients for low-frequency framework information, sedimentary facies attribute information, and fracture facies attribute information, respectively; M, X, and Y represent low-frequency framework information, sedimentary facies attribute information, and fracture facies attribute information, respectively; u0 is the data point to be calculated; and Merge is the calculation function that obtains and merges data from multiple datasets.

[0075] In one example, the seismic data of the target layer undergoes synchronous compression transformation and frequency division to obtain multiple data volumes in different frequency bands, including:

[0076] Seismic signals are acquired based on seismic data from the target layer;

[0077] Continuous wavelet transform is performed on the seismic signal to obtain the wavelet transform coefficients;

[0078] Calculate all instantaneous frequencies of the seismic signal based on wavelet transform coefficients;

[0079] Based on coordinate mapping of all instantaneous frequencies, the wavelet transform coefficients are transformed from the time-scale domain to the time-frequency domain.

[0080] Frequency bands are divided based on the time-frequency domain, resulting in multiple different frequency bands;

[0081] The center frequency of each frequency band is determined based on the instantaneous frequency;

[0082] Based on the center frequency, synchronous compression transformation is performed on each frequency band to obtain multiple corresponding time-frequency distributions;

[0083] Inverse transformation is performed on multiple time-frequency distributions to obtain multiple data volumes in different frequency bands.

[0084] Specifically, for the seismic signal s(t) of the target layer, a continuous wavelet transform is first performed on it to obtain the wavelet transform coefficients:

[0085]

[0086] In the formula, ψ * () represents the complex conjugate of the wavelet mother function, a is the scaling factor, b is the time shift factor, and W s (a,b) are the wavelet transform coefficients obtained after continuous wavelet transform of the seismic signal, and t is the time point at which the seismic signal participates in the transform.

[0087] Based on Plancherel's theory of equal energy in the time and frequency domains, the wavelet transform coefficients W in the frequency domain are rewritten. s , is represented as:

[0088]

[0089] In the formula, ξ is the angular frequency. It is the Fourier transform of the seismic signal s(t). It is the Fourier transform of ψ(t), where ψ(t) is the wavelet mother function. Let be the complex conjugate Fourier transform of the wavelet mother function, where i is the imaginary unit.

[0090] For any value of (a, b), if W s (a,b)≠0, then in the wavelet transform coefficients W s Taking partial derivatives at any point within the support domain (a,b), all instantaneous frequencies of the seismic signal are calculated:

[0091]

[0092] In the formula, i is the imaginary unit. This indicates taking the partial derivative with respect to the wavelet coefficients.

[0093] Using the instantaneous frequency obtained from the above formula, a coordinate mapping is established, transforming each point (a,b) into (ω). s (a,b),b), realizes the transformation of wavelet coefficients from the time-scale domain to the time-frequency domain, that is, transforms W s (a,b) transforms into W s (ω s (a,b),b).

[0094] Frequency bands are divided according to the time-frequency domain, resulting in multiple different frequency bands. The center frequency of each frequency band is determined based on the instantaneous frequency. Synchronous compression transform is then performed on each frequency band based on its center frequency, yielding multiple corresponding time-frequency distributions. Synchronous compression transform involves compressing the interval near the centroid of any center frequency of the wavelet transform coefficients. The values ​​are reorganized according to the frequency range of the actual seismic data, resulting in multiple time-frequency distributions after synchronous compression:

[0095]

[0096] Among them, a k For discrete scales, Δa k =a k -a k-1 ,Δω k =ω k -ω k-1 ω l Let T be the center frequency, Δa be the interval increment along the scale direction, and Δω be the interval increment over the frequency range. s (ω l b) represents the time-frequency transformation value after synchronous compression.

[0097] Inverse transforms multiple time-frequency distributions are performed to obtain multiple data volumes in different frequency bands; because synchronous compression transform is a rearrangement of complex wavelet coefficients in the frequency direction, according to Parseval's theorem, this method can also be applied to T... s (ω l b) Achieving reversible reconstruction to obtain the initial signal, if considered in discrete form, its inverse expression is:

[0098]

[0099] The regularization constant is defined in the formula. This is the inverse transform expression for the regularization constant.

[0100] This invention performs synchronous compression multi-band analysis on seismic data, and by selecting a certain continuous frequency band and performing an inverse Fourier transform, the time-domain seismic signal corresponding to the corresponding frequency band can be obtained.

[0101] For example, based on the frequency range of actual seismic data, wavelet transform coefficients can be divided into low-frequency, mid-frequency, and high-frequency bands, and then seismic information in the low-frequency, mid-frequency, and high-frequency bands can be obtained respectively to improve the identification capability of seismic data.

[0102] In one example, Bayesian inversion includes:

[0103] The prior probability and likelihood function are obtained based on the inversion results of multiple frequency band data volumes, the two-factor control initial model, and the previous frequency band data volume.

[0104] The posterior probability distribution is obtained within a Bayesian framework based on prior probability and likelihood function.

[0105] Find multiple maxima of the posterior probability based on the posterior probability distribution;

[0106] The average of multiple maxima is taken as the inversion result for each Bayesian inversion.

[0107] Specifically, prior probabilities and likelihood functions are calculated based on multiple frequency band data volumes, the initial two-factor control model, and the inversion results of the previous frequency band data volume. The initial inversion does not require the inversion results of the previous frequency band data volume. Assuming that the impedance and noise values ​​within the established initial two-factor control model follow a Gaussian distribution, the prior probabilities and likelihood functions are calculated based on their statistical mean, variance, and other information.

[0108]

[0109] In the above formula, s represents the frequency band data volume; z represents the impedance data of the initial model under two-factor control or the inversion result of the previous frequency band data volume; m represents the seismic wavelet length; L represents the number of sampling points in the frequency band data volume; G represents the Jacobian matrix obtained by taking the partial derivative of the frequency band data volume with respect to impedance; and C represents the frequency band data volume with respect to impedance. n and C z Let represent the covariance matrices of the initial models for random noise and two-factor control, respectively; exp be the natural exponential function; T be the matrix transpose sign; p(s|z) be the prior probability; and p(z) be the likelihood function.

[0110] The first Bayesian inversion uses impedance data from the initial model controlled by two factors and the highest-ranked frequency band data volume as input. The second Bayesian inversion uses the derivative results of the lowest-ranked frequency band data volume and the second-ranked frequency band data volume as input, and so on. This allows seismic information from each frequency band to be transmitted step by step. The final inversion result contains information from all frequency bands of the seismic data, reducing the uncertainty of the inversion and improving the reliability of quantitative reservoir prediction.

[0111] The combined likelihood function and prior probability, within the Bayesian framework, yield the posterior probability distribution:

[0112] p(z|s)∝p(z)p(s|z);

[0113] According to Bayes' rule, ignoring the constant term, the posterior probability distribution can be written as:

[0114]

[0115] In the formula, and This represents the variance matrix of the random noise and the model.

[0116] Taking the logarithm of the above equation and finding the maximum value of the posterior probability p(z|s) is equivalent to finding the minimum value of the objective function J, resulting in the following objective function:

[0117]

[0118] but

[0119] By continuously adjusting the model parameters A perturbation is applied to maximize the posterior probability density value, and the average of multiple posterior probabilities is output as the inversion result. Here, I is the identity matrix. G T It is the transpose of the Jacobian matrix.

[0120] In one example, the expressions for calculating the prior probability and the likelihood function are as follows:

[0121]

[0122] Where s is the frequency band data volume; z is the impedance data of the initial model under two-factor control or the inversion result of the previous frequency band data volume; m is the seismic wavelet length; L is the number of sampling points in the frequency band data volume; G is the Jacobian matrix obtained by taking the partial derivative of the frequency band data volume with respect to impedance; and C... n and C z Let represent the covariance matrices of the initial models for random noise and two-factor control, respectively; exp be the natural exponential function; T be the matrix transpose sign; p(s|z) be the prior probability; and p(z) be the likelihood function.

[0123] In one example, the expression for the posterior probability distribution is:

[0124]

[0125] Where p(z|s) is the posterior probability distribution. and These are the variance matrices of the initial models for random noise and two-factor control, respectively.

[0126] In one example, the expression for calculating the maximum posterior probability is:

[0127]

[0128] in, This represents the minimum value of the objective function J, which is also the maximum value of the posterior probability.

[0129] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the invention. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present invention can be combined with each other.

[0130] Example 1

[0131] This embodiment provides a method for quantitative prediction of reservoirs, including:

[0132] A two-factor controlled initial model is established based on seismic facies attribute information, fault facies attribute information, and low-frequency framework information of the target layer. Seismic facies attribute information and fault facies attribute information are extracted from the seismic data of the target layer. Combining these two information, the longitudinal and lateral distribution characteristics of facies-controlled thin reservoirs under fault influence can be identified. The attribute information and low-frequency framework information are then weighted and fused to establish a two-factor controlled initial model, which can enhance and complement the data, overcome the multi-solution problem of single attributes, indicate the geometric contour information of the reservoir distribution, and provide crucial facies-controlled constraint boundary information for seismic inversion. The low-frequency framework information is obtained by ordinary kriging interpolation of the low-frequency impedance information of the input well within the stratigraphically constrained sequence body. In the formula, W(u) i ) is the low-frequency impedance value at a known input well point in space, λ i The optimal coefficients satisfy the conditions of minimizing the error between unknown and known samples and unbiased estimation. M(u0) is the low-frequency framework information obtained by Kriging interpolation, and n is the number of optimized sample points, i = 1, 2, ..., n. The expression of the two-factor control initial model is: Z(u0) = Merge{αM(u0),βX(u0),γY(u0)}; where α, β, and γ are the weighting coefficients of low-frequency framework information, sedimentary facies attribute information, and fracture facies attribute information, respectively; M, X, and Y represent low-frequency framework information, sedimentary facies attribute information, and fracture facies attribute information, respectively; u0 is the data point to be calculated; and Merge is the calculation function that obtains and fuses data from multiple datasets.

[0133] The process of synchronously compressing and transforming seismic data from the target layer to obtain multiple data volumes in different frequency bands includes: acquiring seismic signals based on the seismic data from the target layer; performing continuous wavelet transform on the seismic signals to obtain wavelet transform coefficients; calculating all instantaneous frequencies of the seismic signals based on the wavelet transform coefficients; performing coordinate mapping based on all instantaneous frequencies to transform the wavelet transform coefficients from the time-scale domain to the time-frequency domain; dividing the time-frequency domain into multiple frequency bands; determining the center frequency of each frequency band based on the instantaneous frequencies; performing synchronous compression transform on each frequency band based on the center frequencies to obtain multiple corresponding time-frequency distributions; and performing inverse transform on the multiple time-frequency distributions to obtain multiple data volumes in different frequency bands. For the seismic signal s(t) from the target layer, a continuous wavelet transform is first performed to obtain the wavelet transform coefficients: In the formula, ψ * () represents the complex conjugate of the wavelet mother function, a is the scaling factor, b is the time shift factor, and W s (a,b) represent the wavelet transform coefficients obtained after continuous wavelet transform of the seismic signal, where t is the time point at which the seismic signal participates in the transform. Based on Plancherel's theory of equal energy in the time and frequency domains, the wavelet transform coefficients W in the frequency domain are rewritten. s , is represented as: In the formula, ξ is the angular frequency. It is the Fourier transform of the seismic signal s(t). It is the Fourier transform of ψ(t), where ψ(t) is the wavelet mother function. Let be the complex conjugate Fourier transform of the wavelet mother function, where i is the imaginary unit. For any value of (a, b), if W s (a,b)≠0, then in the wavelet transform coefficients W s Taking partial derivatives at any point within the support domain (a,b), all instantaneous frequencies of the seismic signal are calculated:

[0134]

[0135] In the formula, i is the imaginary unit. This represents taking the partial derivative with respect to the wavelet coefficients. Using the instantaneous frequency obtained from the above formula, a coordinate mapping is established, transforming each point (a, b) into (ω... s (a,b),b), realizes the transformation of wavelet coefficients from the time-scale domain to the time-frequency domain, that is, transforms W s (a,b) transforms into W s (ω s (a,b),b). Frequency bands are divided according to the time-frequency domain, resulting in multiple different frequency bands. The center frequency of each band is determined based on the instantaneous frequency. Synchronous compression transform is then performed on each band based on its center frequency, yielding multiple corresponding time-frequency distributions. Synchronous compression transform involves compressing the interval near the centroid of any center frequency of the wavelet transform coefficients. The values ​​are reorganized according to the frequency range of the actual seismic data, resulting in multiple time-frequency distributions after synchronous compression:

[0136]

[0137] Among them, a k For discrete scales, Δa k =a k -a k-1 ,Δω k =ω k -ω k1 ω l Let T be the center frequency, Δa be the interval increment along the scale direction, and Δω be the interval increment over the frequency range. s (ω l b) represents the time-frequency transform value after synchronous compression. An inverse transform is performed on multiple time-frequency distributions to obtain multiple data volumes in different frequency bands; because synchronous compression transform is a rearrangement of complex wavelet coefficients in the frequency direction, according to Parseval's theorem, this method can also be applied to T... s (ω l b) Achieving reversible reconstruction to obtain the initial signal, if considered in discrete form, its inverse expression is:

[0138]

[0139] The regularization constant is defined in the formula. This is the inverse transform expression for the regularization constant.

[0140] Based on the two-factor control initial model and each frequency band data volume, Bayesian inversion is performed sequentially in ascending order of frequency. The inversion result of the previous frequency band data volume is used as the initial constraint for the next frequency band data volume, and iterative inversion is performed until all frequency band data volumes have completed Bayesian inversion, thus achieving quantitative reservoir prediction. The Bayesian inversion includes: calculating the prior probability and likelihood function based on multiple frequency band data volumes, the two-factor control initial model, and the inversion results of the previous frequency band data volume; obtaining the posterior probability distribution within the Bayesian framework based on the prior probability and likelihood function; finding multiple maxima of the posterior probability based on the posterior probability distribution; and taking the average of these maxima as the inversion result for each Bayesian inversion. Prior probabilities and likelihood functions are calculated based on multiple frequency band data volumes, the initial two-factor control model, and the inversion results of the previous frequency band data volume. The initial inversion does not require the inversion results of the previous frequency band data volume. Assuming that the impedance and noise values ​​within the established initial two-factor control model follow a Gaussian distribution, the prior probabilities and likelihood functions are calculated based on their statistical mean, variance, and other information.

[0141]

[0142] In the above formula, s represents the frequency band data volume; z represents the impedance data of the initial model under two-factor control or the inversion result of the previous frequency band data volume; m represents the seismic wavelet length; L represents the number of sampling points in the frequency band data volume; G represents the Jacobian matrix obtained by taking the partial derivative of the frequency band data volume with respect to impedance; and C represents the frequency band data volume with respect to impedance. n and C z Let represent the covariance matrices of the random noise and the two-factor controlled initial model, respectively. exp is the natural exponential function, T is the matrix transpose, p(s|z) is the prior probability, and p(z) is the likelihood function. The first Bayesian inversion uses the impedance data from the two-factor controlled initial model and the highest-ranked frequency band data volume as input. The second Bayesian inversion uses the derived results of the lowest-ranked frequency band data volume and the second-ranked frequency band data volume as input, and so on. This allows seismic information from each frequency band to be progressively transmitted, resulting in an inversion result that includes all frequency band information from the seismic data. This reduces the uncertainty of the inversion and improves the reliability of quantitative reservoir prediction.

[0143] The combined likelihood function and prior probability, within the Bayesian framework, yield the posterior probability distribution:

[0144] p(z|s)∝p(z)p(s|z);

[0145] According to Bayes' rule, ignoring the constant term, the posterior probability distribution can be written as:

[0146]

[0147] In the formula, and This represents the variance matrix of the random noise and the model.

[0148] Taking the logarithm of the above equation and finding the maximum value of the posterior probability p(z|s) is equivalent to finding the minimum value of the objective function J, resulting in the following objective function:

[0149]

[0150] but

[0151] By continuously adjusting the model parameters A perturbation is applied to maximize the posterior probability density value, and the average of multiple posterior probabilities is output as the inversion result. Here, I is the identity matrix. G T It is the transpose of the Jacobian matrix.

[0152] Example 2

[0153] like Figure 2 As shown, this embodiment provides a method for quantitative prediction of reservoirs, including:

[0154] ①Prepare earthquake data;

[0155] ② Extract sedimentary facies and structural attributes from seismic data;

[0156] ③ Utilize the low-frequency lattice fusion properties to establish a two-factor phase control model;

[0157] ④ Perform synchronous compression transformation on the seismic data to obtain seismic volume data in different frequency bands;

[0158] ⑤ Incorporate seismic data from different frequency bands into Bayesian iterative inversion under phase control constraints;

[0159] ⑥ Predict the development zone of thin-layer fault-controlled reservoirs and identify favorable reservoirs.

[0160] A single-channel seismic signal synthesized from Ricker wavelets with dominant frequencies of 45Hz, 35Hz, 25Hz, and 20Hz was established. The signal was then calculated using the synchronous compression transform method, and the feasibility of frequency division reconstruction was analyzed. Figure 3 The results of synthesizing the earthquake simulation signal and the synchronous compression time-frequency analysis method show that the synchronous compression transformation can distinguish the four frequency components involved, and the signal frequency value and energy width are well characterized. Figure 4 This is the result of reconstructing the original signal using the corresponding frequencies. As can be seen, the synchronous compression method can clearly resolve the four participating signal components, and the reconstructed result is basically consistent with the original components, with minimal error.

[0161] Guided by geological models of reservoirs controlled by both shoal facies and faults, a two-dimensional geological model was established, such as... Figure 5 and Figure 6 As shown. The basic parameters of the model are set according to the main marker stratigraphic segments of the marine region of the Sichuan Basin. Two sets of reservoirs are set in the model, one along the fault line and the other laterally, corresponding to the actual reservoir development. Forward modeling shows that when the reservoir is developed, the seismic reflection of the three members of the Maokou Formation is characterized by fault wave development, weakened lateral wave trough energy, and medium to weak wave peak reflection characteristics in the reverse polarity profile. Based on the seismic response characteristic analysis, attribute profiles are extracted from the seismic data of the Maokou Formation in a certain area of ​​southern Sichuan. Figure 7 To reflect the wave trough amplitude properties of the beach facies distribution, the extent to which the lateral facies is modified is indicated; Figure 8 To reflect the dip angle and curvature properties of the fracture distribution, and to represent the extent of dissolution associated with physical fracturing of the longitudinal section, Figure 9 The seismic attribute envelope represents the reservoir spatial information embodied by the integrated longitudinal and transverse attribute controls. By weighted fusion of the attribute volume envelope as a spatial constraint phase with a low-frequency framework model, a two-factor controlled initial model can be established. This model is well-adapted to geological models and can provide facies boundary information for subsequent inversion, such as... Figure 10 As shown.

[0162] Figure 11 This is an actual seismic profile of the Maokou Formation in southeastern Sichuan Basin. Figure 12 From top to bottom, these are seismic profiles at different scales obtained after synchronous compression transformation and decomposition, namely, low-frequency data profile, mid-frequency data profile, and high-frequency data profile. Figure 13 This is a comparison chart of the original seismic data spectrum and the spectrum obtained from the decomposition. The comparison shows that the low-frequency profile mainly reflects the large stratigraphic framework, while the mid- and high-frequency profiles, while maintaining certain stratigraphic framework information, show more obvious vertical detail reflection characteristics within the Maokou Formation, resulting in improved resolution. This indicates that as the decomposition frequency increases, the seismic data's ability to identify small geological bodies gradually improves, which helps to improve the prediction resolution of thin reservoirs. The method of this invention was used to conduct a method experiment on the marine Maokou Formation in the Sichuan Basin. After establishing a two-factor initial model, low-frequency seismic data was first used as input for inversion. Then, this low-frequency inversion result was used as a supplementary low-frequency source when mid-frequency seismic data was used for inversion. After mid-frequency seismic data was used for inversion, this mid-frequency inversion result was used as a supplementary low-frequency source when high-frequency seismic data was used for inversion. Under this iterative constraint, the different scale information contained in the seismic data was fully utilized, and the inversion results were refined step by step, improving the prediction resolution and stability. The inversion results obtained using the method of this invention are compared with conventional inversion results. Figure 14 This is the low-frequency inversion result under the constraints of a two-factor control model. Figure 15 The mid-frequency inversion result is constrained by low frequencies. Figure 16 The results are high-frequency inversion results constrained by the intermediate frequency, which are the final calculation results of the method of this invention. It can be seen that after calculation using the method of this invention, the thin-layer morphology of this type of fault-controlled shoal facies dual-factor controlled reservoir gradually becomes clearer, and the development range and internal conditions are more obvious. Figure 17 The conventional inversion results, obtained without factor constraints, cannot accurately characterize reservoir boundary features, resulting in relatively low inversion accuracy. In contrast, the inversion prediction results obtained by the method of this invention enhance the accuracy of reservoir details and the ability to characterize lateral heterogeneity compared to conventional inversion prediction results, making it more suitable for complex structural regions.

[0163] The results achieved by this invention enable dual-factor joint control, which is more consistent with geological models. Low-frequency information is gradually integrated into medium and high frequencies, effectively improving the quantitative description accuracy of thin reservoirs. This helps to delineate high-quality reservoir development areas, reduce interpretation risks, and provide key support for subsequent well location deployment. It has certain application value in research areas with similar reservoirs and geological conditions.

[0164] Example 3

[0165] This embodiment provides an apparatus for quantitative reservoir prediction, including:

[0166] A module is established to create a two-factor control initial model based on seismic facies attribute information, fault facies attribute information, and low-frequency grid information of the target layer.

[0167] The synchronous compression transform frequency division module is used to perform synchronous compression transform frequency division on the seismic data of the target layer to obtain multiple data volumes in different frequency bands;

[0168] The prediction module is used to perform Bayesian inversion on the initial model controlled by two factors and each frequency band data volume in ascending order of frequency. It uses the inversion result of the previous frequency band data volume as the initial constraint for the next frequency band data volume to perform iterative inversion until all frequency band data volumes have completed Bayesian inversion, thus completing the quantitative prediction of the reservoir.

[0169] Example 4

[0170] This embodiment also provides an electronic device, which includes:

[0171] At least one processor; and,

[0172] A memory communicatively connected to the at least one processor; wherein,

[0173] The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the reservoir quantitative prediction method in Embodiment 1.

[0174] An electronic device according to embodiments of the present disclosure includes a memory and a processor. The memory is used to store non-transitory computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0175] The processor may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In one embodiment of this disclosure, the processor is used to execute computer-readable instructions stored in the memory.

[0176] Those skilled in the art will understand that, in order to solve the technical problem of how to achieve a good user experience, this embodiment may also include well-known structures such as communication buses and interfaces, and these well-known structures should also be included within the protection scope of this disclosure.

[0177] For a detailed description of this embodiment, please refer to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.

[0178] Example 5

[0179] This embodiment provides a non-transitory computer-readable storage medium that stores computer instructions for causing a computer to execute the reservoir quantitative prediction method in Embodiment 1.

[0180] A computer-readable storage medium according to embodiments of the present disclosure stores non-transitory computer-readable instructions. When these non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the methods described in the foregoing embodiments of the present disclosure are performed.

[0181] The aforementioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or portable hard drive), media with built-in rewritable non-volatile memory (e.g., memory card), and media with built-in ROM (e.g., ROM cartridge).

[0182] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. A method for quantitative prediction of reservoirs, characterized in that, include: A two-factor control initial model is established based on the seismic phase attribute information, fault phase attribute information, and low-frequency grid information of the target layer; The seismic data of the target layer are synchronously compressed, transformed, and frequency-divided to obtain multiple data volumes in different frequency bands; Based on the two-factor control initial model and each frequency band data volume, Bayesian inversion is performed sequentially in order of frequency band from low to high. The inversion result of the previous frequency band data volume is used as the initial constraint for the next frequency band data volume for iterative inversion until all frequency band data volumes have completed the Bayesian inversion, thus completing the quantitative prediction of the reservoir.

2. The reservoir quantitative prediction method according to claim 1, characterized in that, The expression for the initial model of the two-factor control is: Z(u0)=Merge{αM(u0),βX(u0),γY(u0)}; Where α, β, and γ are the weighting coefficients for low-frequency framework information, sedimentary facies attribute information, and fracture facies attribute information, respectively; M, X, and Y represent low-frequency framework information, sedimentary facies attribute information, and fracture facies attribute information, respectively; u0 is the data point to be calculated; and Merge is the calculation function that obtains and merges data from multiple datasets.

3. The reservoir quantitative prediction method according to claim 1, characterized in that, The process of synchronously compressing, transforming, and frequency-dividing the seismic data of the target layer to obtain multiple data volumes in different frequency bands includes: Seismic signals are acquired based on seismic data from the target layer; Perform continuous wavelet transform on the seismic signal to obtain the wavelet transform coefficients; Calculate all instantaneous frequencies of the seismic signal based on the wavelet transform coefficients; Based on all the instantaneous frequencies, coordinate mapping is performed to transform the wavelet transform coefficients from the time-scale domain to the time-frequency domain; Frequency bands are divided based on the time-frequency domain to obtain multiple different frequency bands; The center frequency of each frequency band is determined based on the instantaneous frequency; Based on the center frequency, each frequency band is synchronously compressed and transformed to obtain multiple corresponding time-frequency distributions; Perform inverse transformation on multiple time-frequency distributions to obtain multiple data volumes in different frequency bands.

4. The reservoir quantitative prediction method according to claim 1, characterized in that, The Bayesian inversion includes: The prior probability and likelihood function are obtained based on the multiple frequency band data volumes, the two-factor control initial model, and the inversion result of the previous frequency band data volume. Based on the prior probability and likelihood function, the posterior probability distribution is obtained within the Bayesian framework. Based on the posterior probability distribution, multiple maxima of the posterior probability are obtained; The average of the multiple maxima is taken as the inversion result of each Bayesian inversion.

5. The reservoir quantitative prediction method according to claim 4, characterized in that, The expressions for obtaining the prior probability and the likelihood function are as follows: Where s is the frequency band data volume; z is the impedance data of the initial model under two-factor control or the inversion result of the previous frequency band data volume; m is the seismic wavelet length; L is the number of sampling points in the frequency band data volume; G is the Jacobian matrix obtained by taking the partial derivative of the frequency band data volume with respect to impedance; and C... n and C z Let represent the covariance matrices of the initial models for random noise and two-factor control, respectively; exp be the natural exponential function; T be the matrix transpose sign; p(s|z) be the prior probability; and p(z) be the likelihood function.

6. The reservoir quantitative prediction method according to claim 5, characterized in that, The expression for the posterior probability distribution is: Where p(z|s) is the posterior probability distribution. and These are the variance matrices of the initial models for random noise and two-factor control, respectively.

7. The reservoir quantitative prediction method according to claim 6, characterized in that, The expression for calculating the maximum value of the posterior probability is: in, This represents the minimum value of the objective function J, which is also the maximum value of the posterior probability.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the reservoir quantitative prediction method according to any one of claims 1-7.

9. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing a computer to perform the reservoir quantitative prediction method according to any one of claims 1-7.

10. A reservoir quantitative prediction device, characterized in that, include: A module is established to create a two-factor control initial model based on seismic facies attribute information, fault facies attribute information, and low-frequency grid information of the target layer. The synchronous compression transform frequency division module is used to perform synchronous compression transform frequency division on the seismic data of the target layer to obtain multiple data volumes in different frequency bands; The prediction module is used to perform Bayesian inversion sequentially based on the two-factor control initial model and each frequency band data volume in ascending order of frequency band. It iterates the inversion based on the inversion result of the previous frequency band data volume as the initial constraint for the next frequency band data volume until all frequency band data volumes have completed the Bayesian inversion, thus completing the quantitative prediction of the reservoir.

Images