Well seismic dispersion correction method suitable for volcanic rock large-scale fractured reservoir
Through the combination of support vector machine and multi-scale dynamic equivalent medium model, the accuracy problem of well earthquake-frequency dispersion correction in large-scale fracture reservoirs of volcanic rocks is solved, and high-precision well earthquake-frequency dispersion correction and reservoir fluid prediction are achieved.
Patent Information
- Application Number
- CN202310546087.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-16
- Publication Date
- 2024-11-22
AI Technical Summary
The prior art is difficult to accurately correct the frequency of well earthquake dispersion in large-scale fracture reservoirs of volcanic rocks, resulting in poor matching of well earthquake data, seriously affecting the reservoir fluid prediction accuracy.
Using a conventional well logging curve fracture prediction method based on support vector machine, combining the full-band volcanic rock fracture rock petrophysical model and multi-scale dynamic equivalent medium model, the well logging and seismic frequency band velocity are matched to achieve high-precision wells Vibration dispersion correction.
The accuracy of the earthquake-frequency dispersion correction of wells in large-scale fracture reservoirs of volcanic rocks is improved, the accurate calculation of vertical and transverse wave velocity in earthquake frequency bands is ensured, and the reliability of reservoir fluid prediction is enhanced.
Smart Images

Figure CN119001900A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of reservoir rock physical property research, and in particular to a well seismic dispersion correction method suitable for volcanic rock large-scale fracture reservoirs. Background Art
[0002] Due to the complex pore structure of large-scale fractured volcanic rocks, when elastic waves pass through, pore fluid pressure gradients appear in multiple regions, which intensifies the flow of fluid in the porous medium. Based on the wave-induced flow theory, multi-scale dispersion will be generated, and the dispersion law is relatively complex.
[0003] In actual production, especially in large-scale fractured reservoirs of volcanic rocks, the well-seismic velocities vary greatly. Synthetic seismic records made from logging data always have the problem of mismatching well-seismic data, and the well-seismic calibration effect is poor, which seriously affects the accuracy of subsequent reservoir fluid prediction.
[0004] In recent years, some scholars have carried out research on the method of correction of well-seismic dispersion. Zhao Jianguo et al. measured the full-band P- and S-wave velocities of ancient Ordovician-Cambrian carbonate rocks using low-frequency measurement equipment, and used the seismic band velocities at different pressures to fit the relationship between formation pressure and seismic band velocity. Then, the relationship between formation pressure and logging band velocity was fitted using the logging band velocities at different pressures, and then the logging and seismic velocities were fitted. This method only considered the relationship between formation pressure and well-seismic velocity, and did not consider the influence of large-scale fractures in the formation. Therefore, the prediction effect was less accurate. Deng Jixin et al. used the resonant Q model to Well seismic dispersion correction is performed, where the value of Q is a constant. However, it is known from experiments that if there is dispersion between well seismic in large-scale fractured reservoirs, Q must not be a constant, and the Q value is different in different regions, so the accuracy of well seismic dispersion correction is also low; Luo Xia et al. developed a well seismic dispersion correction method suitable for dense sandstone formations, which achieved well seismic dispersion in the presence of microscopic scale cracks. However, since there are cracks of various scales in large-scale fractured reservoirs of volcanic rocks, it is known through experiments that there are multi-region dispersions. Therefore, directly applying the dense sandstone formation dispersion correction method to carry out dispersion correction of large-scale fractured reservoirs of volcanic rocks will inevitably not be very ideal. The above existing technologies are all greatly different from the present invention and fail to solve the technical problems we want to solve.
[0005] In the Chinese patent application with application number: CN202110509081.8, a method for correction of well seismic dispersion suitable for dense sandstone formations is involved, which includes: step 1, selecting the target reservoir core, and carrying out lithology, physical properties and low-frequency rock physical parameter measurements; step 2, constructing a full-band dense sandstone rock physics model; step 3, calibrating the dynamic equivalent microscale model according to experimental data; step 4, performing well seismic velocity dispersion correction. The method for correction of well seismic dispersion suitable for dense sandstone formations verifies the rationality of the rock physics model by finely calibrating the rock physics experimental data and the rock physics model, and then based on the rock physics model, adjusts the fracture density to make it consistent with the actual logging data of the calculated logging data or less than the error, and then realizes the calculation of the P- and S-wave velocities in the seismic frequency band.
[0006] In the Chinese patent application with application number: CN201010567538.2, a dispersion AVO inversion method is involved, including: obtaining pre-stack gathers, and performing spherical diffusion compensation and dynamic correction on the pre-stack gathers; performing instantaneous spectrum analysis on the pre-stack gathers to generate amplitude spectra at different frequencies; performing amplitude spectrum equalization on the amplitude spectrum; obtaining the stacking velocity of the sampling points in the pre-stack gathers, and generating the layer velocity of the sampling points according to the stacking velocity; according to the amplitude spectrum after spectrum equalization and the layer velocity of the sampling points, generating the longitudinal and transverse wave velocity change rate and the dispersion degree of the sampling points at the set frequency according to the least squares inversion; arranging the generated dispersion degrees according to the gather sequence number to generate a dispersion degree attribute profile. The dispersion AVO inversion method of the embodiment of the invention eliminates the influence of strong amplitude energy clusters caused by differences in elastic interface wave impedance, and can also avoid false "frequency anomalies" that may be caused by stacking processing during spectrum analysis of post-stack seismic data, making the obtained results more reliable.
[0007] In the Chinese patent application with application number: CN201910343232.X, a method for matching well logging and seismic velocity based on seismic rock physics experimental analysis is involved, including: step 1, establishing a velocity variation curve with frequency, i.e., velocity dispersion variation law, through rock physics experimental analysis of seismic frequency bands; step 2, constructing a rock physics model suitable for characterizing the dispersion characteristics of seismic frequency bands; step 3, using the modulus calculation formula of the rock physics model to perform point-by-point mapping dispersion correction of the acoustic wave curve; step 4, calibrating the dispersion-corrected acoustic wave curve by synthetic recording, comparing and analyzing the seismic well bypass and the synthetic seismic record, and outputting the correction curve results of all depths of the entire acoustic well logging curve. This method not only eliminates the matching problem of seismic velocity and logging velocity in scale, but also corrects the velocity dispersion problem caused by the large difference between the two in frequency bands. It has a clear physical connotation and rock physics basis, and greatly improves the accuracy of well-seismic matching and calibration.
[0008] In the Chinese patent application with application number: CN202011033767.6, a pre-stack and post-stack well-seismic joint time-depth calibration method is involved. On the basis of the traditional post-stack seismic well-seismic time-depth calibration, the application of pre-stack track data is combined. For the overall stretch and compression correction of the time-depth relationship, the present invention uses the dynamic correction velocity analysis function of the pre-stack track to distinguish the type of time-depth relationship anomaly and make corrections; for the local stretch and compression correction of the time-depth relationship, the invention uses the phase time of the seismic track beside the well as the standard time, and realizes the local stretch and compression correction of the time-depth relationship by correcting the phase time of the synthetic record. During the time-depth calibration process, the value of the logging curve itself is not changed, which aims to solve the technical problems existing in the prior art that the well-seismic calibration work is difficult and the results are inaccurate because the inaccurate acoustic time difference data is used as the basic data of the time-depth conversion speed, and provides a guarantee for the subsequent high-precision inversion.
[0009] The above existing technologies are greatly different from the present invention and fail to solve the technical problem we want to solve. Therefore, we have invented a new well seismic dispersion correction method suitable for large-scale fracture reservoirs in volcanic rocks. Summary of the invention
[0010] The purpose of the present invention is to provide a well seismic dispersion correction method suitable for volcanic rock large-scale fracture reservoirs, which realizes high-precision velocity dispersion correction of volcanic rock fracture formations.
[0011] The purpose of the present invention can be achieved by the following technical measures: a well seismic dispersion correction method suitable for volcanic rock large-scale fracture reservoirs, the well seismic dispersion correction method suitable for volcanic rock large-scale fracture reservoirs comprises:
[0012] Step 1: Carry out conventional logging curve fracture prediction and fracture characteristic parameter calculation based on support vector machine;
[0013] Step 2: Select the target volcanic rock fracture reservoir core to carry out lithology, physical properties and low-frequency rock physical parameter measurements;
[0014] Step 3, constructing a full-band volcanic rock fracture rock physics model;
[0015] Step 4, calibrating the multi-scale dynamic equivalent medium model using experimental data;
[0016] Step 5: Perform well seismic velocity dispersion correction.
[0017] The purpose of the present invention can also be achieved by the following technical measures:
[0018] In step 1, based on the characterization of fractures by full-bore formation microresistivity imaging logging FMI data, the conventional logging curve fracture prediction and fracture characteristic parameter calculation method based on support vector machine are used. With eight curves including dual lateral resistivity (RD, RS), microspherical focused resistivity (MSFL), longitudinal wave velocity (Vp), density (DEN), natural gamma (gr), neutron (CNL), and wellbore (Cal) as inputs and the corresponding depth high-conductivity fractures of FMI as outputs, the support vector machine (SVM) method is used to identify high-conductivity fractures in volcanic rocks in the study area, and then the fracture characteristic parameters are calculated according to the empirical formula.
[0019] Step 1 includes:
[0020] Step 11, selecting wells in the work area that contain FMI logging data, statistically analyzing fracture characteristics and the response relationship between non-fracture and conventional curves, and establishing training samples;
[0021] Step 12, normalizing the sample data;
[0022] Step 13, selecting the kernel function and penalty factor;
[0023] Step 14, performing high-conductivity crack identification;
[0024] Step 15, calculate crack characteristic parameters.
[0025] In step 11, it is necessary to count no less than 60 conventional curve response values of high-conductivity fractures and no less than 20 conventional curve response values of non-high-conductivity fractures.
[0026] In step 12, the sample data is normalized, the normalization interval is [-1,1], and the normalization calculation formula is as follows:
[0027]
[0028] Among them, x i 、x imax 、x imin They represent the original logging value of the ith conventional curve, the maximum value of the original logging of the ith conventional curve, and the minimum value of the original logging of the ith conventional curve, respectively. i Represents the normalized value of the i-th regular curve.
[0029] In step 13, a Gaussian kernel is selected as the kernel function of the support vector machine method, as shown in the following formula:
[0030]
[0031] Among them, m is the characteristic value of imaging logging, d is the order of the polynomial, σ is the width of the Gaussian distribution, the penalty factor C is 800, and y iIt represents the normalized value of the i-th conventional curve.
[0032] In step 14, through the selection of samples, feature normalization, kernel function, and penalty factor, the support vector machine is trained using the training sample set to obtain a discriminant function for high-conductivity fracture identification based on the support vector machine method, as shown in the following formula:
[0033] n = sgn{∑α i n i k(y i , m)+b} (3)
[0035] where α i is the gradient coefficient, n and n i are the output of the predicted high-conductivity fracture sample and the support vector sample respectively, k(y i , m) is the Gaussian kernel function, and b is the intercept.
[0036] In step 15, after identifying the high-conductivity fractures, the calculation of fracture angle, fracture width, and fracture density can be achieved according to the following formula:
[0037]
[0038] Y < 0, θ is less than 30 degrees; 0 < Y < 0.1, θ is between 30 - 60 degrees; Y > 0.1, θ is greater than 60 degrees (4)
[0040] where Y is the discriminant factor, θ is the fracture dip angle, and currently only qualitative discrimination by region can be performed. R d is the deep lateral resistivity, and R s is the shallow lateral resistivity;
[0041]
[0042] where a is the fracture width in meters, and R mf is the mud filtrate resistivity;
[0043]
[0044] where ε c is the fracture density, and R b is the matrix resistivity.
[0045] In step 2, the volcanic rock core of the target formation is selected to measure the parameters such as porosity, permeability, density, and mineral composition, and then the low-frequency rock physical parameter measurement under formation conditions is carried out, with a measurement frequency of 12-200 Hz, and the saturation states are dry, saturated with water, and saturated with oil; finally, the porosity, permeability, density, and mineral composition data of the core, as well as the low-frequency longitudinal and shear wave velocity data of the core under formation conditions are obtained, providing data support for the subsequent rock physical model verification.
[0046] In step 3, starting from the poroelastic theory, the poroelastic model is expanded by introducing a set of large-scale cracks with directional arrangement. Based on the fluid jet flow theory, a multi-scale dynamic equivalent medium model is proposed, which considers the fluid flow at both the mesoscale and microscale.
[0047] In step 3, the bulk modulus K of the equivalent medium of the multiscale dynamic equivalent medium model eff and shear modulus u eff It can be expressed as:
[0048]
[0049]
[0050] Among them, ω is the angular frequency, K is the skeleton bulk modulus, μ is the shear modulus, ε f is the microcrack density, a is the crack radius, r is the microcrack radius, ε c is the crack density, θ is the inclination angle, is the porosity, λ is the Lame coefficient, Kc, Kp, A, and B are all intermediate parameters without specific physical meanings, A and B are
[0051]
[0052]
[0053] in,
[0054]
[0055] K f is the bulk modulus of the fluid, τ is the time scale, which controls the frequency band range of the dispersion characteristics. It is proportional to the viscosity η of the fluid, inversely proportional to the permeability k, and related to the fracture radius a. Its reciprocal is called the characteristic frequency. For a smaller aspect ratio:
[0056]
[0057] is the particle size, ν is the Poisson's ratio of the solid particle; according to the formula proposed by Connell and Budiansky, the P-wave velocity V is obtainedP and the speed of the S wave V S They are:
[0058]
[0059]
[0060] Among them, Re is the real part and ρ is the density.
[0061] In step 4, the mineral composition analysis data is substituted into the multi-scale equivalent medium model to obtain the full-band rock physical parameters of the core, namely the P-wave velocity and S-wave velocity. By adjusting the fracture density parameters, the P-wave and S-wave velocities in the calculated logging frequency band are matched with the P-wave and S-wave velocities in the logging frequency band. At this time, the calculated low-frequency P-wave and S-wave velocity data are the data to be simulated.
[0062] In step 5, firstly, the lithologic component data is obtained through the logging interpretation data, and then the skeleton modulus K and μ are calculated based on the Voigt-Reuss-Hill model. Then, the porosity data and fracture characteristic parameters are obtained through the logging interpretation data and the aforementioned fracture prediction method, and then the dry rock elastic modulus is calculated based on the Berrymann model. Then, the fluid component data and permeability data are obtained through the logging interpretation data, and the P-wave velocity at the logging frequency is calculated based on the multi-scale equivalent medium model. If the absolute difference between the calculated P-wave velocity Vp1 and the measured P-wave velocity Vp is less than the error, then the P-wave and S-wave velocities in the seismic frequency band can be directly calculated. If the difference between the calculated P-wave velocity and the measured P-wave velocity is greater than the error, then the microcrack density is adjusted and the calculation is performed again according to the above process, and the cycle is repeated until the absolute difference between the calculated P-wave velocity and the measured P-wave velocity is less than the error.
[0063] Step 5 includes:
[0064] Step 51, using the skeleton modulus calculation—Voigt-Reuss-Hill average theory;
[0065] Step 52, calculating the elastic modulus of dry rock;
[0066] Step 53, calculate the full-band longitudinal and transverse wave velocities of fluid-containing rocks.
[0067] In step 51, when the relative content and elastic modulus of each phase constituting the rock medium are known, the upper and lower limits of the effective elastic modulus of the rock medium can be calculated using the Voigt and Reuss formulas; the upper limit Voigt formula represents the ratio of the stress to the strain of the rock medium under the same strain state and the same strain for each phase constituting the rock medium; the lower limit Reuss represents the ratio of the stress to the strain of the rock medium under the same stress state and the same stress for each phase constituting the rock medium. The Hill formula is the arithmetic average result of the Voigt upper limit and the Reuss lower limit:
[0068]
[0069] Where M V 、M R 、M VRH represents the values of bulk modulus K, shear modulus μ, and Young's modulus E obtained by the Voigt, Reuss, and Hill methods; f i is the volume content of the i-th component constituting the rock medium; M i is the elastic modulus of the ith component.
[0070] At step 52, the general form of the self-consistency approximation for the N-phase composite material is:
[0071]
[0072]
[0073] In the ith material, x i is its volume fraction, P and Q represent geometric factors, and the superscript i on P and Q indicates that these factors are in the self-compatible effective modulus and The background medium contains material i; k i ,u i are the bulk modulus and shear modulus of the i-th mineral, respectively. The sum represents the consideration of all phases, including minerals and pores. Dry pores can be simulated by setting the inclusion modulus to zero; fluid-saturated pores can be simulated by setting the inclusion shear modulus to zero. At the same time, these equations are coupled and must be solved by simultaneous iteration.
[0074] In step 53, after the dry rock elastic modulus is calculated, the full-band longitudinal and transverse wave velocities can be calculated based on the multi-scale equivalent medium model in combination with the parameters such as porosity, permeability, and fluid saturation of the logging interpretation results and the predicted fracture characteristic parameters.
[0075] The well seismic dispersion correction method suitable for volcanic rock large-scale fracture reservoirs in the present invention first forms a method for quantitatively characterizing fracture characteristic parameters using conventional logging curves based on FMI imaging logging data; secondly, starting from conducting rock physics experiments on volcanic rock fracture cores, the rationality of the constructed rock physics model is verified by finely calibrating the rock physics experimental data and the rock physics model; then, based on the rock physics model, the microcrack parameters are adjusted to make them consistent with the actual logging data of the calculated logging data or less than the error, thereby realizing the calculation of the P- and S-wave velocities of the volcanic rock large-scale fracture formation under the seismic frequency band.
[0076] The well seismic dispersion correction method of the present invention, which is suitable for large-scale fracture reservoirs in volcanic rocks, has advantages that other technologies do not have. Its specific advantages and characteristics are manifested in the following aspects:
[0077] First, based on the characterization of fracture characteristics by FMI imaging data, a fracture identification method based on conventional logging curves was developed, with a fracture identification accuracy of 85%, laying a data foundation for describing the propagation law of elastic waves in volcanic fracture reservoirs;
[0078] Second, a triple-pore rock physics model that is more suitable for volcanic rock fracture reservoirs and takes both microscopic and mesoscopic into consideration was constructed, and the logging velocity was used as a constrained cyclic iteration to make the calculated seismic frequency band velocity more accurate. BRIEF DESCRIPTION OF THE DRAWINGS
[0079] Figure 1 This is a schematic diagram of actual rock pores in a specific embodiment of the present invention;
[0080] Figure 2 This is a schematic diagram of rock equivalent pores in a specific embodiment of the present invention;
[0081] Figure 3 This is a comparison diagram between the 7# core P-wave velocity measurement data and the model in a specific embodiment of the present invention;
[0082] Figure 4 This is a comparison diagram between the 7# core shear wave velocity measurement data and the model in a specific embodiment of the present invention;
[0083] Figure 5 A flow chart of conventional logging curves for identifying fractures in a specific embodiment of the present invention;
[0084] Figure 6 It is a flow chart of a specific embodiment of the well seismic dispersion correction method suitable for volcanic rock large-scale fracture reservoirs of the present invention;
[0085] Figure 7 It is a schematic diagram of FMI imaging logging of a certain well in a specific embodiment of the present invention;
[0086] Figure 8 This is a diagram showing the effect of frequency dispersion correction of a certain well in a specific embodiment of the present invention;
[0087] Fig. 9 This is a diagram showing the effect of frequency dispersion correction of a certain well in a specific embodiment of the present invention;
[0088] Fig.10 This is a diagram showing the effect of dispersion correction for a certain well in a specific embodiment of the present invention. DETAILED DESCRIPTION
[0089] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0090] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations and / or combinations thereof.
[0091] The well seismic dispersion correction method suitable for large-scale fracture reservoirs of volcanic rocks of the present invention comprises:
[0092] Step 1: Based on the characterization of fractures using full-bore formation microresistivity imaging (FMI) data, conventional logging curve fracture prediction and fracture characteristic parameter calculation based on support vector machine are carried out.
[0093] Step 2: Select the target volcanic rock fracture reservoir core to carry out lithology, physical properties and low-frequency rock physical parameter measurements.
[0094] Step 3: Construct a full-band volcanic rock fracture rock physics model.
[0095] Step 4: Calibrate the multi-scale dynamic equivalent medium model using experimental data.
[0096] Step 5: Correction of well seismic velocity dispersion.
[0097] The following are several specific embodiments of the present invention.
[0098] Example 1
[0099] In a specific embodiment 1 of the present invention, Figure 6 As shown, Figure 6 The flowchart of the well seismic dispersion correction method suitable for large-scale fracture reservoirs of volcanic rocks of the present invention is as follows. The well seismic dispersion correction method suitable for large-scale fracture reservoirs of volcanic rocks comprises:
[0100] (1) Based on the characterization of fractures using full-bore formation microresistivity imaging (FMI) logging data, conventional logging curve fracture prediction and fracture characteristic parameter calculation based on support vector machine are carried out, such as Figure 5 As shown, Figure 5 The specific process for predicting fractures is as follows: ① Select wells with imaging logging interpretation, and establish the high-conductivity fracture depth section based on the imaging interpretation results; ② Optimize the parameters of conventional curves that are sensitive to fractures, statistically analyze the fracture characteristics and the response relationship between non-fractures and conventional curves, and establish training samples; ③ Establish a fracture recognition model based on the training samples of the support vector machine algorithm; ④ Select the wells that need to be calculated, further select eight conventional curves that are sensitive to fractures, and perform normalization processing; ⑤ Carry out fracture prediction based on the recognition model, and calculate the fracture characteristic parameters based on the empirical formula.
[0101] Full-bore formation microresistivity imaging (FMI) logging is a high-resolution formation imaging logging tool that performs microresistivity scanning measurements through button electrodes attached to the well wall, and converts the difference in formation resistivity near the well wall into clear image light and dark changes. When geological bodies such as cracks and dissolved pores appear in the formation, the resistivity measurement value changes, so FMI can clearly characterize the characteristic parameters of formation fractures. However, FMI logging is expensive, and the price of a single well measurement is more than 100 times that of conventional curve logging. Therefore, in most cases, at most 1 to 2 wells in a work area contain imaging logging data. Conventional logging curves, such as dual lateral resistivity, micro-spherical focused resistivity, acoustic wave velocity, density, borehole diameter, natural gamma and other logging data, will also produce different responses on these logging curves when cracks appear in the formation. However, due to the existence of multiple solutions, the accuracy of fracture identification is low. Therefore, a conventional logging curve fracture prediction and fracture characteristic parameter calculation method based on support vector machine is developed based on FMI data characterization of fractures. The eight curves of dual lateral resistivity (RD, RS), micro-spherical focused resistivity (MSFL), longitudinal wave velocity (Vp), density (DEN), natural gamma (gr), neutron (CNL), and borehole diameter (Cal) are used as input, and the corresponding depth high-conductivity fractures of FMI are used as output. The (support vector machine) SVM method is used to identify high-conductivity fractures in volcanic rocks in the study area, and then the fracture characteristic parameters are calculated according to the empirical formula.
[0102] ① Select wells in the work area that contain FMI logging data, statistically analyze fracture characteristics and the response relationship between non-fracture and conventional curves, and establish training samples
[0103] The 8 conventional logging curves with high response to fractures are Vp, gr, RD, MSFL, RS, DEN, CNL, and CAl. Since only high-conductivity fractures have a greater impact on well seismic dispersion, it is only necessary to count the response relationship between high-conductivity fracture characteristics and conventional curves. As many sample data as possible is a prerequisite for improving prediction accuracy. Generally, the sample data must not be less than 80. Here, it is necessary to count no less than 60 high-conductivity fracture conventional curve response values and no less than 20 non-high-conductivity fracture conventional curve response values.
[0104] ② Sample data normalization processing
[0105] In order to prevent a sample feature from being too large or too small, thus playing an unbalanced role in training, and in kernel calculations, inner product operations or exponential operations are used. Unbalanced data may cause calculation difficulties. Generally, the sample data needs to be normalized, and the normalization interval is [-1,1]. The normalization calculation formula is as follows:
[0106]
[0107] Among them, x i 、x imax 、x imin They represent the original logging value of the ith conventional curve, the maximum value of the original logging of the ith conventional curve, and the minimum value of the original logging of the ith conventional curve, respectively. i Represents the normalized value of the i-th regular curve.
[0108] ③Choice of kernel function and penalty factor
[0109] The ultimate goal of the support vector machine is to find a suitable classification function to predict unknown samples. The determination of the classification function mainly involves the selection of the kernel function and the determination of the penalty factor C. These two parameters have a great influence on the accuracy and applicability of the model. The most commonly used kernel functions are linear kernel, polynomial kernel, and Gaussian kernel. In general, the Gaussian kernel is selected as the kernel function of the support vector machine method, as shown in the following formula:
[0110]
[0111] Among them, m is the imaging logging characteristic value, d is the order of the polynomial, σ is the width of the Gaussian distribution, and the penalty factor C is 800.
[0112] ④ Identification of high-conductivity cracks
[0113] Through the selection of samples, feature normalization, kernel function and penalty factor, the support vector machine is trained using the training sample set to obtain the discriminant function of high-conductivity crack identification based on the support vector machine method, as shown in the following formula:
[0114] n=sgn{∑αi n i k(y i ,m)+b} (3)
[0116] where α i is the gradient coefficient, n, n i are the output of the predicted high-conductivity fracture samples and the support vector samples respectively, k(y i ,m) is the Gaussian kernel function, and b is the intercept.
[0117] ⑤ Calculate fracture characteristic parameters
[0118] After identifying the high-conductivity fractures, the calculation of fracture angle, fracture width, and fracture density can be achieved according to formulas (4, 5, 6):
[0119]
[0120] Y < 0, θ is less than 30 degrees; 0 < Y < 0.1, θ is between 30 - 60 degrees; Y > 0.1, θ is greater than 60 degrees (4)
[0122] where Y is the discrimination factor, θ is the fracture dip angle, and currently only qualitative discrimination can be performed in different regions. R d is the deep lateral resistivity, and R s is the shallow lateral resistivity.
[0123]
[0124] where a is the fracture width in meters, and R mf is the mud filtrate resistivity.
[0125]
[0126] where ε c is the fracture density, and R b is the matrix resistivity.
[0127] (2) Select the core of the target volcanic rock fracture reservoir to carry out lithology, physical properties, and low-frequency rock physics parameter measurements
[0128] Select the core of the volcanic rock in the target formation to measure parameters such as porosity, permeability, density, and mineral composition, and then carry out low-frequency rock physics parameter measurements under formation conditions. The measurement frequency is 12 - 200 Hz, and the saturation states are dry, water-saturated, and oil-saturated. Finally, obtain the porosity, permeability, density, mineral composition data of the core, as well as the low-frequency P-wave and S-wave velocity data of the core under formation conditions, providing data support for subsequent verification of the rock physics model.
[0129] (3) Construct a full-frequency volcanic rock fracture rock physics model
[0130] The volcanic fractured reservoir rock is actually a kind of fluid-containing fractured porous medium. The existence of pores and fluids and the flow between the fluids in the pores and the fluids in the fractures have an important influence on the propagation of seismic waves. At this time, the classic rock physics model Gassmann equation guiding seismic exploration is no longer valid.
[0131] Starting from the poroelastic theory, the present invention expands the poroelastic model, introduces a group of large-scale fractures arranged in a directional manner, and proposes a multi-scale dynamic equivalent medium model based on the fluid jet flow theory. The model considers the fluid flow at both the mesoscopic and microscopic scales. In order to describe the elastic wave propagation of fractured media in the full frequency band, it is necessary to consider the complex pores ( Figure 1 ) is simplified by assuming that ( Figure 2 ) There are three types of pores: circular pores, large-scale cracks, and microcracks. The circular pores are relatively rigid and incompressible; large-scale cracks (without fillings, i.e., high-conductivity cracks) and flat pores (microcracks) are relatively weak. For the convenience of mathematical derivation, it is assumed that microcracks, pores, and microcracks and pores can be interconnected. Each crack can be connected to multiple microcracks or pores, but each microcrack and each pore is connected to at most one crack, and the cracks are not connected to each other. Based on this triple porosity hypothesis, the flow of fluids between pores, microcracks, and cracks in rocks containing microcracks, pores, and large-scale cracks is described. The derivation of the model formula is complicated, so only the results are given here.
[0132] Bulk modulus K of the equivalent medium in the multiscale dynamic equivalent medium model eff and shear modulus u eff It can be expressed as:
[0133]
[0134]
[0135]
[0136] Among them, ω is the angular frequency, K is the skeleton bulk modulus, μ is the shear modulus, ε f is the microcrack density, a is the crack radius, r is the microcrack radius, ε c is the crack density, θ is the inclination angle, is the porosity, λ is the Lame coefficient, Kc, Kp, A, and B are all intermediate parameters without specific physical meanings, A and B are
[0137]
[0138]
[0139] in,
[0140]
[0141] K f is the bulk modulus of the fluid, τ is the time scale, which controls the frequency band range where the dispersion characteristics appear. It is proportional to the viscosity η of the fluid, inversely proportional to the permeability k, and related to the fracture radius a. Its reciprocal is called the characteristic frequency. For smaller aspect ratios:
[0142]
[0143] is the particle size, ν is the Poisson's ratio of the solid particle. According to the formula proposed by Connell and Budiansky, the P-wave velocity V is obtained P and the speed of the S wave V S They are:
[0144]
[0145]
[0146] Among them, Re is the real part and ρ is the density.
[0147] (4) Experimental data calibration of multi-scale dynamic equivalent medium model
[0148] Core 7# was selected for calibration verification. From the mineral composition analysis data, it can be seen that the tuff content of core 7# is 21.5%, the sandstone content is 63%, and the clay content is 15.5%. The cracks are well developed, and low-angle cracks are visible. The crack radius is measured to be 0.25 mm, the crack density is 0.035, and the particle diameter is 0.02 mm. The microcracks are invisible to the naked eye and cannot be quantitatively evaluated. From the physical property measurement, it can be known that the core density is 2.474g / cm 3 , porosity 0.0114, permeability 0.378mD, the test state is 100% saturated with water, and the fluid viscosity is 1 centipoise. Substituting the above data into the multi-scale equivalent medium model can obtain the full-band rock physical parameters of the 7# core - longitudinal wave velocity and shear wave velocity.
[0149] Figure 3 , Figure 4 The comparison results of the P-wave and S-wave velocities of the 7# core and the multi-scale equivalent medium model simulation are shown. Since the microcrack density is a microscopic parameter and is not easy to obtain, and the microcrack density parameter in the rock physics model is a parameter that must be input, the fracture density parameter can be adjusted to make the calculated P-wave and S-wave velocities in the well logging frequency band match the P-wave and S-wave velocities in the well logging. At this time, the calculated low-frequency P-wave and S-wave velocity data is the data to be simulated. Figure 3 , Figure 4It can be seen that the P-wave and S-wave velocities measured in the low-frequency band of core 7# are in good agreement with those calculated by the rock physics model, with the P-wave velocity agreement exceeding 85% and the S-wave velocity agreement exceeding 80%. This verifies that the multi-scale equivalent medium model can better describe the propagation of elastic waves in volcanic rock fracture reservoirs in the full-frequency band, thus laying the foundation for subsequent well-seismic dispersion correction.
[0150] (5) Well seismic velocity dispersion correction
[0151] First, the lithologic component data is obtained through the logging interpretation data, and then the skeleton modulus K and μ are calculated based on the Voigt-Reuss-Hill model. Then, the porosity data and fracture characteristic parameters are obtained through the logging interpretation data and the aforementioned fracture prediction method. Then, the dry rock elastic modulus is calculated based on the Berrymann model. Then, the fluid component data and permeability data are obtained through the logging interpretation data. The P-wave velocity Vp1 at the logging frequency is calculated based on the multi-scale equivalent medium model. If the absolute difference between the calculated P-wave velocity and the measured P-wave velocity Vp is less than the error, then the P-wave and S-wave velocities in the seismic frequency band can be directly calculated. If the difference between the calculated P-wave velocity and the measured P-wave velocity is greater than the error, then the microcrack density is adjusted and the calculation is performed again according to the above process, and the cycle is repeated until the absolute difference between the calculated P-wave velocity and the measured P-wave velocity is less than the error.
[0152] ① Calculation of skeleton modulus—Voigt-Reuss-Hill average theory
[0153] When the relative content and elastic modulus of each phase of the rock medium are known, the upper and lower limits of the effective elastic modulus of the rock medium can be calculated using the Voigt and Reuss formulas (Formula 11). The upper limit of the Voigt formula represents the ratio of the stress to the strain of the rock medium under the same strain state and the same strain for each phase of the rock medium; the lower limit of Reuss represents the ratio of the stress to the strain of the rock medium under the same stress state and the same stress for each phase of the rock medium. The Hill formula is the arithmetic average of the Voigt upper limit and the Reuss lower limit.
[0154]
[0155] Where M V 、M R 、M VRH represents the values of bulk modulus K, shear modulus μ, and Young's modulus E obtained by the Voigt, Reuss, and Hill methods; f i is the volume content of the i-th component constituting the rock medium; M i is the elastic modulus of the ith component.
[0156] ② Calculation of elastic modulus of dry rock
[0157] Berryman (1980b, 1995) gave the general form of the self-consistency approximation for N-phase composites:
[0158]
[0159]
[0160] In the ith material, x i is its volume fraction, P and Q represent geometric factors, and the superscript i on P and Q indicates that these factors are in the self-compatible effective modulus and The background medium contains material i. k i ,u i are the bulk modulus and shear modulus of the i-th mineral, respectively. The sum represents the consideration of all phases, including minerals and pores. Dry pores can be simulated by setting the inclusion modulus to zero. Fluid-saturated pores can be simulated by setting the inclusion shear modulus to zero. At the same time, these equations are coupled and must be solved by simultaneous iteration.
[0161] ③ Calculation of full-band longitudinal and transverse wave velocities of fluid-containing rocks
[0162] After calculating the elastic modulus of dry rock, the full-band longitudinal and transverse wave velocities can be calculated based on the multi-scale equivalent medium model, combined with the porosity, permeability, fluid saturation and other parameters of the logging interpretation data and the predicted fracture characteristic parameters.
[0163] Example 2
[0164] In a specific embodiment 2 of the present invention, the correction effect will be analyzed in detail by taking a certain work area of Shengli Oilfield as an example. The work area includes a FMI imaging data well Su 13 well. Figure 7As shown in Figure 1, the second track is the resistivity dynamic imaging map, and the third track is the imaging interpretation result. The dark gray tadpole map indicates high-conductivity fractures, the light gray tadpole map indicates high-resistance fractures, and the white tadpole map indicates the formation interface. High-conductivity fractures are fractures that need to be identified. The well is selected to statistically analyze the response relationship between high-conductivity fractures and non-high-conductivity fractures and conventional curves to establish training samples. Table 1 shows the established training samples. For the convenience of display, only some samples are displayed. In Table 1, the first column is the fracture mark. When the actual depth imaging interpretation result is a high-conductivity fracture, the fracture mark is 1, and when the imaging interpretation result is a non-high-conductivity fracture, the fracture mark is 0; the second column is the depth, and the third to tenth columns are respectively P-wave velocity (Vp), natural gamma (gr), deep lateral resistivity (RD), microsphere focusing resistivity (MSFL), shallow lateral resistivity (RS), neutron (CNL), density (DEN), and wellbore (CAL). Table 2 is the remaining sample detection table, which is mainly used to test the prediction consistency of the SVM algorithm. In Table 2, the first column is the result of fracture prediction by SVM algorithm, 1 represents high conductivity fracture (unfilled fracture), 0 represents non-high conductivity fracture; the second column is the result of imaging interpretation, 1 represents high conductivity fracture, 0 represents non-high conductivity fracture; the third column is depth, and the fourth to eleventh columns are characteristic curves, which are P-wave velocity (Vp), natural gamma (gr), deep lateral resistivity (RD), microsphere focused resistivity (MSFL), shallow lateral resistivity (RS), neutron (CNL), density (DEN), and wellbore (CAL). As can be seen from Table 2, there are only 3 inconsistencies between the fracture prediction results in the first column and the imaging interpretation results in the second column, and the overall consistency rate exceeds 85%, which verifies the reliability of the SVM algorithm.
[0165] Table 1. Su 13 well training sample table
[0166] Crack Sign depth Vp gr RD MSFL RS CNL DEN CA1 1 1747.5 3017.882 73.371 3.564 5.356 4.151 25.982 1.876 0.0809 1 1748.75 3086.566 74.8396 2.3261 10.3812 2.638 27.4944 2.0585 0.0815 1 1752.3 4617.252 72.1175 8.406 16.8165 10.1926 15.2731 2.6055 0.0738 1 1755.6 3495.474 69.2792 3.9223 20.0441 4.1691 16.5123 2.5006 0.0734 1 1757.3 3165.786 77.5709 3.3505 6.0258 3.5445 21.5434 2.3451 0.0756 1 1764 3826.358 84.83 4.04 4.495 3.697 22.133 2.445 0.0736 1 1765.2 3779.128 96.7572 3.671 12.62 3.276 22.0171 2.3697 0.0719 1 1770.75 3510.735 89.5896 4.7103 5.9904 4.3521 24.161 2.3602 0.0759 1 1773 3969.732 74.6 5.984 9.555 5.671 19.587 2.49 0.0704 1 1780.2 4407.256 102.6001 5.5909 5.8289 5.7127 26.1559 2.4461 0.0726 1 1783.4 3566.246 90.0969 6.6153 7.8199 6.6838 18.5247 2.4904 0.0676 1 1790.3 4109.558 101.5069 4.9372 4.273 5.8338 25.6031 2.4098 0.0722 1 1791 4436.165 86.099 4.548 7.51 5.356 19.619 2.496 0.0729 1 1798 5022.079 77.595 11.853 51.206 12.089 15.476 2.56 0.0676 1 1799.3 3259.428 92.413 5.4505 5.3583 5.325 22.4031 2.4859 0.0682 1 1886 5193.212 133.036 56.495 107.853 69.505 11.233 2.485 0.0638 1 1887.6 5135.724 139.9497 52.3681 734.2693 65.6846 8.8726 2.5167 0.0638 1 1901.7 4914.646 120.0651 23.2217 98.7654 26.4785 13.9221 2.4704 0.065 1 1904.7 4489.568 98.1108 7.413 10.0006 8.518 19.0346 2.5089 0.0661 1 1911.8 4034.063 88.333 7.5199 9.1155 8.851 20.1862 2.5078 0.0652 1 1931.9 4273.537 99.4606 6.7882 14.0677 6.2048 22.455 2.4448 0.0744 0 1969 5304.375 107.312 55.559 194.823 54.612 15.141 2.476 0.0614 0 1969.5 5222.218 109.401 50.444 75.647 50.746 15.994 2.462 0.0614 0 1956 5517.441 110.907 70.343 239.153 81.422 14.169 2.518 0.0614 0 1956.5 5215.338 118.699 68.304 42.042 77.855 14.082 2.514 0.0613 0 1956.7 5226.055 117.269 69.6011 45.8262 78.8154 14.6125 2.5072 0.0614 0 1957 5365.63 119.086 70.296 76.015 79.675 15.152 2.495 0.0614 0 1957.5 5441.885 112.052 67.772 102.513 76.719 14.722 2.457 0.0613 0 1958 5051.71 124.373 38.219 44.799 43.001 14.336 2.49 0.0614 0 1958.5 4833.032 116.673 37.257 149.599 40.344 16.284 2.447 0.0621 0 1897 4858.842 100.026 18.01 10.731 19.819 18.252 2.517 0.0647 0 1897.5 5109.208 87.643 24.435 46.413 27.896 16.898 2.473 0.0647 0 1897.7 5125.933 83.9756 28.4712 107.8136 32.5634 16.8373 2.4561 0.0648
[0167] Table 2 Remaining sample detection table
[0168] Predicting cracks Crack Sign depth Vp gr RD MSFL RS CNL DEN CA1 1 1 1901.7 4914.646 120.0651 23.2217 98.7654 26.4785 13.9221 2.4704 0.065 1 1 1904.7 4489.568 98.1108 7.413 10.0006 8.518 19.0346 2.5089 0.0661 1 1 1911.8 4034.063 88.333 7.5199 9.1155 8.851 20.1862 2.5078 0.0652 1 1 1931.9 4273.537 99.4606 6.7882 14.0677 6.2048 22.455 2.4448 0.0744 1 1 1935.6 4154.332 72.0263 9.2369 11.333 9.0285 19.9981 2.5046 0.0634 1 1 1950.3 4994.488 80.2152 35.7917 51.1636 42.0794 16.3287 2.5059 0.0621 1 1 2014.5 4690.674 107.328 15.895 11.827 16.51 18.059 2.497 0.0596 1 1 2000.1 4269.012 97.767 15.7405 18.8967 15.9187 19.8206 2.5528 0.0611 0 1 1962.8 4983.14 101.387 34.948 62.0454 33.0451 15.1161 2.5013 0.0614 1 1 1965.8 4944.779 123.854 37.0159 68.5342 33.0951 15.4494 2.4969 0.0616 1 1 1970.3 5177.665 113.912 41.5157 53.8799 40.816 13.7775 2.4484 0.0614 1 1 1994.5 4843.015 100.35 17.834 18.353 17.575 16.706 2.554 0.0623 0 0 2020.1 4473.839 117.1492 23.6358 76.6005 24.1656 17.9369 2.5367 0.0605 0 0 2021 4683.539 117.781 26.205 85.83 26.308 17.168 2.536 0.0606 0 0 2021.1 4693.645 117.781 25.7933 56.1674 25.8087 17.4006 2.5332 0.0606 0 0 2021.2 4715.479 117.781 25.268 47.4531 25.2843 17.3282 2.533 0.0606 1 0 2020.5 4675.92 117.781 26.391 30.789 27.104 16.256 2.541 0.0604 0 0 2020.6 4651.611 117.781 26.4699 22.9723 27.1855 15.8365 2.541 0.0604 0 0 2020.7 4633.233 117.781 26.491 28.1116 27.1405 15.8717 2.5405 0.0604 1 0 2021.6 4702 117.781 23.4142 49.0952 23.7912 17.2972 2.5406 0.0606 0 0 2021.9 4680.588 117.781 21.9266 31.2728 22.5464 17.8867 2.545 0.0606
[0169] Based on the support vector machine method, the above sample extraction part is normalized and pattern recognition training is carried out, and the remaining samples are used as test samples for verification, with an accuracy rate of 85%. The following takes three wells as examples to analyze the correction effect in detail. Figure 8 This is the calibration effect diagram of Su 13 Well. The first line in the figure is the P-wave velocity, the dotted line is the predicted 40Hz P-wave velocity, the solid line is the well logging P-wave velocity, the second line is the predicted 40Hz S-wave velocity, the third line is the porosity, the fourth line is the tuff content, and the fifth line is the fracture opening. Figure 8It can be seen that the 1940m-1950m stratum of Su 13 well has a high tuff content and a porosity of 8%. It is predicted that cracks are well developed and the crack opening is about 0.5 mm. The calculated 40Hz P-wave velocity is about 300m / s different from the measured P-wave velocity in well logging, and the dispersion amplitude is close to 10%. The dispersion cannot be ignored. Fig. 9 This is the dispersion correction effect diagram of Su 1-25 well. The first track in the figure is permeability, the second track is well logging P-wave velocity (solid line) and predicted 40Hz P-wave velocity (dashed line), the fourth track is porosity, the fifth track is tuff content, and the sixth track is fracture opening. It can be seen from the figure that the well is tuff formation from 2000 to 2010m, with a porosity of about 5%. It is predicted that fractures are developed and the opening reaches about 200 microns. After calculation, the P-wave velocity (solid line) of this section of well logging and the P-wave velocity (dashed line) of the seismic frequency band (40Hz) are significantly different, with a maximum of 250m / s and a dispersion amplitude of 7%; at 2010-2020 meters, there are also cracks, but the crack opening is only tens of microns, so the dispersion amplitude is small. The maximum difference between the P-wave velocity of this section of well logging and seismic is less than 100m / s, and the dispersion amplitude is 2%; below 2020 meters, there is no development of fractures, and there is no dispersion of P- and S-wave velocities, so no correction is required.
[0170] Example 3
[0171] In a specific embodiment 3 of the present invention, a work area of Shengli Oilfield is taken as an example to analyze the correction effect in detail. The lithology of this block is mainly granite, with developed fractures. The support vector machine algorithm of the present invention is used to predict fractures. On this basis, the well-seismic dispersion correction of this block is realized based on the multi-scale model, which lays a foundation for subsequent reservoir prediction, seismic forward and inversion, etc. Fig.10 This is the dispersion correction effect diagram of Well 672-1. The figure shows the measured P-wave velocity (solid line) and the predicted P-wave velocity (dashed line) of the seismic frequency band (40Hz). The second track is the predicted S-wave velocity of the seismic frequency band 40Hz, the third track is the density, and the fourth track is the fracture opening. From the figure, we can see that the well has developed fractures at 1280-1290m, and the fracture opening is about 300 microns. After calculation, the P-wave dispersion amplitude is about 8%. At the same time, it is also found that there is dispersion in places without fractures. Comprehensive analysis shows that the velocity is low at this place, which can generally be considered that micro-cracks are developed, so there is also dispersion between wells and earthquakes, but the dispersion amplitude is small.
[0172] Finally, it should be noted that the above is only 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 aforementioned embodiments, those skilled in the art can still modify the technical solutions recorded in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
[0173] Except for the technical features described in the specification, all other technical features are known technologies to those skilled in the art.
Claims
1. A well seismic dispersion correction method suitable for large-scale fracture reservoirs in volcanic rocks, characterized in that: The well-seismic dispersion correction method suitable for large-scale fractured reservoirs in volcanic rocks includes: Step 1: Conduct conventional logging curve fracture prediction based on support vector machines and calculate fracture characteristic parameters. Step 2: Select cores from the target volcanic rock fractured reservoir to measure lithology, physical properties, and low-frequency rock physics parameters. Step 3: Construct a full-frequency volcanic rock fractured rock physics model. Step 4: Calibrate the multi-scale dynamic equivalent medium model using experimental data. Step 5: Perform well-seismic velocity dispersion correction.
2. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 1, characterized in that: In Step 1, based on the characterization of fractures by full-borehole formation micro-resistivity imaging logging FMI data, the method for conventional logging curve fracture prediction based on support vector machines and calculating fracture characteristic parameters uses eight curves, namely dual laterolog resistivity RD, RS, micro-spherical focused resistivity MSFL, compressional wave velocity Vp, density DEN, natural gamma gr, neutron CNL, and caliper Cal, as inputs, and the corresponding depth high-conductivity fractures of FMI as outputs. The support vector machine SVM method is used to identify high-conductivity fractures in volcanic rocks in the study area, and then fracture characteristic parameters are calculated according to empirical formulas.
3. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 2, characterized in that: Step 1 includes: Step 11: Select wells in the work area containing FMI logging data, statistically analyze the fracture characteristics and the response relationships between non-fractures and conventional curves, and establish training samples. Step 12: Normalize the sample data. Step 13: Select the kernel function and penalty factor. Step 14: Conduct high-conductivity fracture discrimination. Step 15: Calculate fracture characteristic parameters.
4. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 3, characterized in that: In Step 11, it is necessary to statistically analyze no less than 60 conventional curve response values of high-conductivity fractures and no less than 20 conventional curve response values of non-high-conductivity fractures.
5. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 3, characterized in that: In Step 12, the sample data is normalized, and the normalization interval is [-1, 1]. The normalization calculation formula is as follows: Among them, x i 、x imax 、x imin They represent the original logging value of the ith conventional curve, the maximum value of the original logging of the ith conventional curve, and the minimum value of the original logging of the ith conventional curve, respectively. i Represents the normalized value of the i-th regular curve.
6. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 3, characterized in that: In Step 13, the Gaussian kernel is selected as the kernel function of the support vector machine method, as follows: Among them, m is the characteristic value of imaging logging, d is the order of the polynomial, σ is the width of the Gaussian distribution, the penalty factor C is 800, and y i Represents the normalized value of the i-th regular curve.
7. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 3, characterized in that: In Step 14, through the selection of samples, feature normalization, kernel function, and penalty factor, the support vector machine is trained using the training sample set to obtain a discrimination function for high-conductivity fracture identification based on the support vector machine method, as follows: n=sgn{∑α i n i k(y i ,m)+b} (3) where α i is the gradient coefficient, n, n i are the output and support vector samples for predicting high-conductivity crack samples, k(y i ,m) is the Gaussian kernel function, and b is the intercept.
8. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 3, characterized in that: In Step 15, after identifying high-conductivity fractures, the calculation of fracture angle, fracture width, and fracture density can be achieved according to the following formula: Y < 0, θ is less than 30 degrees; 0 < Y < 0.1, θ is between 30 - 60 degrees; Y > 0.1, θ is greater than 60 degrees (4) Among them, Y is the discriminant factor, θ is the crack inclination, and currently only qualitative discrimination can be made by region. d is the deep lateral resistivity, R s is the shallow lateral resistivity; Where a is the crack width, in meters, R mf is the resistivity of mud filtrate; where v c is the crack density, R b is the matrix resistivity.
9. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 1, characterized in that: In Step 2, select cores of volcanic rocks from the target formation to measure parameters such as porosity, permeability, density, and mineral composition, and then conduct low-frequency rock physics parameter measurements under formation conditions. The measurement frequency is 12 - 200 Hz, and the saturation states are dry, water-saturated, and oil-saturated. Finally, the porosity, permeability, density, mineral composition data of the cores, and the low-frequency compressional and shear wave velocity data of the cores under formation conditions are obtained, providing data support for subsequent verification of the rock physics model.
10. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 1, characterized in that: In Step 3, starting from the poroelastic theory, the poroelastic model is extended by introducing a set of large-scale fractures arranged in an oriented manner. Based on the fluid jet flow theory, a multi-scale dynamic equivalent medium model is proposed, which simultaneously considers fluid flow at mesoscopic and microscopic scales.
11. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 10, characterized in that: In step 3, the bulk modulus K of the equivalent medium of the multiscale dynamic equivalent medium model eff and shear modulus u eff It can be expressed as: Among them, ω is the angular frequency, K is the skeleton bulk modulus, μ is the shear modulus, ε f is the microcrack density, a is the crack radius, r is the microcrack radius, ε c is the crack density, θ is the inclination angle, is the porosity, λ is the Lame coefficient, Kc, Kp, A, and B are all intermediate parameters without specific physical meanings, A and B are Among them, K f is the bulk modulus of the fluid, τ is the time scale, which controls the frequency band range of the dispersion characteristics. It is proportional to the viscosity η of the fluid, inversely proportional to the permeability k, and related to the fracture radius a. Its reciprocal is called the characteristic frequency. For a smaller aspect ratio: is the particle size, ν is the Poisson's ratio of the solid particle; according to the formula proposed by Connell and Budiansky, the P-wave velocity V is obtained P and the speed of the S wave V S They are: Among them, Re is the real part and ρ is the density.
12. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 1, characterized in that: In step 4, the mineral composition analysis data is substituted into the multi-scale equivalent medium model to obtain the full-band rock physical parameters of the core, namely the P-wave velocity and S-wave velocity. By adjusting the fracture density parameters, the P-wave and S-wave velocities in the calculated logging frequency band are matched with the P-wave and S-wave velocities in the logging frequency band. At this time, the calculated low-frequency P-wave and S-wave velocity data are the data to be simulated.
13. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 1, characterized in that: In step 5, firstly, the lithologic component data is obtained through the logging interpretation data, and then the skeleton modulus K and μ are calculated based on the Voigt-Reuss-Hill model. Then, the porosity data and fracture characteristic parameters are obtained through the logging interpretation data and the aforementioned fracture prediction method, and then the dry rock elastic modulus is calculated based on the Berrymann model. Then, the fluid component data and permeability data are obtained through the logging interpretation data, and the P-wave velocity at the logging frequency is calculated based on the multi-scale equivalent medium model. If the absolute difference between the calculated P-wave velocity and the measured P-wave velocity is less than the error, then the P-wave and S-wave velocities in the seismic frequency band can be directly calculated. If the difference between the calculated P-wave velocity and the measured P-wave velocity is greater than the error, then the microcrack density is adjusted and the calculation is performed again according to the above process, and the cycle is repeated until the absolute difference between the calculated P-wave velocity and the measured P-wave velocity is less than the error.
14. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 13, characterized in that: Step 5 includes: Step 51, using the skeleton modulus calculation—Voigt-Reuss-Hill average theory; Step 52, calculating the elastic modulus of dry rock; Step 53, calculate the full-band longitudinal and transverse wave velocities of fluid-containing rocks.
15. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 14, characterized in that: In step 51, when the relative content and elastic modulus of each phase constituting the rock medium are known, the upper and lower limits of the effective elastic modulus of the rock medium can be calculated using the Voigt and Reuss formula; The upper limit Voigt formula represents the same strain state, that is, the ratio of the stress to the strain of the rock medium under the same strain for each phase constituting the rock medium; The Reuss lower limit represents the ratio of stress to strain of the rock medium under the same stress state and the various phases constituting the rock medium. The Hill formula is the arithmetic average of the Voigt upper limit and the Reuss lower limit: Where M V 、M R 、M VRH represents the values of bulk modulus K, shear modulus μ, and Young's modulus E obtained by the Voigt, Reuss, and Hill methods; f i is the volume content of the i-th component constituting the rock medium; M i is the elastic modulus of the ith component.
16. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 14, characterized in that: At step 52, the general form of the self-consistency approximation for the N-phase composite material is: In the ith material, x i is its volume fraction, P and Q represent geometric factors, and the superscript i on P and Q indicates that these factors are in the self-compatible effective modulus and The background medium contains material i; k i ,u i are the bulk modulus and shear modulus of the i-th mineral, respectively. The sum represents the consideration of all phases, including minerals and pores. Dry pores can be simulated by setting the inclusion modulus to zero; fluid-saturated pores can be simulated by setting the inclusion shear modulus to zero. At the same time, these equations are coupled and must be solved by simultaneous iteration.
17. The method for correcting well seismic dispersion suitable for large-scale fracture reservoirs in volcanic rocks according to claim 14, characterized in that: In step 53, after the dry rock elastic modulus is calculated, the full-band longitudinal and transverse wave velocities can be calculated based on the multi-scale equivalent medium model in combination with the parameters such as porosity, permeability, and fluid saturation of the logging interpretation results and the predicted fracture characteristic parameters.
Citation Information
Cited By
Ground stress azimuth earthquake prediction method based on transverse wave velocity change rate
CN120214895A
Volcanic rock reservoir gas saturation and permeability seismic inversion method and system
CN120559726A
