Shale reservoir ground stress prediction method and device
By combining geophysical well logging and indoor experimental data, a three-dimensional geostress field model was established using linear regression and neural network method, the problems of fuzzy and large errors in the existing technology were solved, and the ground stress in shale reservoirs were improved Prediction accuracy and reliability.
Patent Information
- Application Number
- CN202310608953.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-29
- Publication Date
- 2024-11-29
Smart Images

Figure CN119047287A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of shale reservoir geostress prediction, and in particular to a shale reservoir geostress prediction method and device. Background Art
[0002] Shale has the characteristics of low porosity and low permeability. Horizontal drilling and staged fracturing are important ways to exploit shale oil and gas. With the increasing difficulty of conventional energy extraction, unconventional shale reservoirs have gradually become one of the important targets of exploration and development. The ground stress state is the main controlling factor in the shale oil and gas extraction process. In fracturing transformation, the ground stress state controls the shape, height, width and direction of hydraulic fractures, and can also affect the increase in production after fracturing. At the same time, ground stress is also an important basis for well network deployment, adjustment and selection of horizontal well section direction. Therefore, evaluating ground stress is crucial for shale reservoir exploitation.
[0003] In fact, stress activity can cause deformation or fracture of rocks, and it plays an important role in oil and gas exploration and development (Jenkins, 2017). Geostress is generally composed of tectonic stress, gravity stress, thermal stress and pore pressure, and its state is usually expressed by three normal stresses, namely, σ H , σ v , σ h , respectively the maximum horizontal principal stress, vertical stress and minimum horizontal principal stress (Lin et al., 2006; Kuuskraa and Ammer. 2004; Matsuki and Takeuchi. 1993). In the crust, the magnitude and direction of the stress in different regions and strata at different depths will change with space and time to form a stress field.
[0004] At present, there are many methods for evaluating geostress, such as hydraulic fracturing, acoustic emission and logging calculation. Fracturing refers to the method of calculating geostress through relevant formulas according to the change of pressure over time during fracturing. Hydraulic fracturing is reliable, direct and simple, but in actual production, not every well contains small-scale fracturing test data; Acoustic emission (Goodman. 1963) refers to when the rock is squeezed by external force, part of the energy stored in it will be released in the form of elastic waves. The direction of the geostress is obtained by analyzing the characteristics of these elastic wave sounds. The acoustic emission method can measure the stress in the deep reservoir, but due to the high cost of acoustic emission testing and the lack of continuity of the results, there are fewer indoor data sample points for testing, and for those who do not have The geostress state of the tested formation can only be roughly estimated; the formula calculation method based on logging curve data overcomes the shortcomings of high test price and discontinuous measuring points. It refers to the method of using logging data to calculate rock mechanical parameters such as formation Poisson's ratio, Young's modulus, shear modulus and bulk modulus, and then indirectly obtain geostress using formula expressions, but its calculation process cannot be directly calibrated with the results of indoor experimental measured geostress, and the final calculation results often have large errors with the measured data points; the neural network method can better make up for the drawback that the indoor experimental measured geostress results cannot be calibrated in the logging calculation process. Through the machine learning method, the connection between indoor experimental measured geostress and geophysical logging curves is sought, but a large number of indoor experimental measured geostress results are needed for support. Therefore, in view of the above shortcomings, a shale reservoir geostress prediction method and device are proposed. Summary of the invention
[0005] 1. Technical issues to be resolved
[0006] The present invention provides a method and device for predicting geostress in shale reservoirs, which avoids the defects in the prior art such as fuzziness, discontinuity and large errors in geostress evaluation results caused by the harsh use conditions of the commonly used geostress evaluation methods, resulting in too small sample size of experimental data.
[0007] (II) Technical solution
[0008] In order to solve the above problems, the present invention provides a shale reservoir ground stress prediction method, comprising:
[0009] Step S1: obtaining geophysical logging parameter data of multiple shale oil and gas wells, and obtaining X-mac cross-dipole acoustic logging data of one of the wells;
[0010] Step S2: performing environmental correction on the oil and gas wells described in step S1, and performing standardization processing on the acquired geophysical logging data;
[0011] Step S3: Acquire the indoor experimental data of all oil and gas wells and perform depth homing correction;
[0012] Step S4: According to the X-mac cross-dipole acoustic logging data obtained in step S1 and the geophysical logging data processed in step S2, the parameter data of the shear wave curve is predicted by the linear regression method;
[0013] Step S5: According to the geophysical logging data obtained in step S2, in combination with the shear wave time difference obtained in step S4, dynamic rock mechanical parameters are calculated jointly, and according to the experimental data obtained in step S3, dynamic and static corrections are performed on the dynamic rock mechanical parameters to obtain static rock mechanical parameters;
[0014] Step S6: fitting the poroelastic coefficient (biot coefficient) and pore pressure according to the experimental data of step S3, and jointly calculating the experimental stress data with the static rock mechanics parameters obtained in step S5, and performing error analysis;
[0015] Step S7: expanding the calculation result obtained in step S6 into a learning sample of geostress data, taking the geophysical logging parameters of a single well as a training sample, performing neural network calculation, and obtaining logging stress data;
[0016] Step S8: Use logging stress data to establish a geological grid model, perform spatial interpolation on the geophysical logging data of all oil and gas wells, calculate a three-dimensional geostress field model using a neural network method, and predict, analyze and evaluate the spatial geostress distribution of shale reservoirs based on the obtained calculation model.
[0017] Preferably, the geophysical logging parameter data of the oil and gas wells include natural gamma, natural potential, deep lateral resistivity, deep lateral resistivity, apparent resistivity, well diameter, density and compressional wave time difference.
[0018] Preferably, the oil and gas wells are all subjected to environmental correction using a standard layer statistical analysis method, and the geophysical logging data of all oil and gas wells are subjected to standardization processing using a histogram method.
[0019] Preferably, the parameters of the oil and gas wells measured by the indoor experiment include rock parameters and stress value parameters, and the depth reposition correction is to correspond and correct the depth data corresponding to the rock parameters and stress value parameters measured by the indoor experiment and the depth data of geophysical logging through the reposition diagram method.
[0020] Preferably, the dynamic rock mechanics parameters include dynamic Young's modulus, dynamic Poisson's ratio, bulk modulus and shear modulus, and the static rock mechanics parameters include static Young's modulus and static Poisson's ratio.
[0021] Preferably, the maximum horizontal geostress and the minimum horizontal geostress of the indoor experiment are calculated by combining the rock parameters, the biot coefficient and the pore pressure, and the calculation expressions are:
[0022]
[0023]
[0024] σ H is the maximum horizontal principal stress, MPa; σ h is the minimum horizontal principal stress, MPa; υ is Poisson's ratio, dimensionless; σ v is the overburden pressure, MPa; α is the biot coefficient, dimensionless; P p is the pore pressure, MPa; E is Young's modulus, GPa; ξ h is the strain in the direction of the minimum horizontal principal stress, dimensionless; ξ H is the strain in the direction of maximum horizontal principal stress, dimensionless.
[0025] Preferably, the biot coefficient is the ratio of the deformation of the pore space of the formation to the total volume change, and the expression of the biot coefficient is:
[0026] α=1-K dry / K m
[0027] Where: α is the biot coefficient; Kdry is the dry rock bulk modulus, GPa; Km is the mineral bulk modulus, GPa.
[0028] Preferably, the overburden stratum pressure is vertical geostress, and the expression of the vertical geostress is:
[0029]
[0030] Among them, σ v is the vertical geostress, MPa; h is the depth, m; ρ(h) is the density logging value, g / cm 3 ; g is the acceleration due to gravity, m / s 2 .
[0031] The present invention also provides a device based on the shale reservoir ground stress prediction method, comprising:
[0032] A data acquisition module, wherein the data acquisition module is respectively connected to a data correction and standard processing module and a measuring point value homing module, wherein the measuring point value homing module is respectively connected to a static rock mechanics obtaining module, a ground stress learning sample expansion module and a ground stress BP neural network method prediction module, wherein the data correction and standard processing module is connected to a shear wave time difference prediction module, wherein the shear wave time difference prediction module is connected to a static rock mechanics obtaining module, wherein the static rock mechanics obtaining module is respectively connected to a data correction and standard processing module and a ground stress learning sample expansion module, wherein the ground stress learning sample expansion module is connected to a ground stress BP neural network method prediction module, wherein the ground stress BP neural network method prediction module is connected to a ground stress evaluation module.
[0033] (III) Beneficial effects
[0034] The shale reservoir geostress prediction method and device provided by the present invention have strong practicability and accurate and reliable results. By acquiring a variety of information on oil and gas wells, the sample size of experimental data can be expanded. By establishing a three-dimensional geostress prediction model based on neural network learning, the geostress prediction accuracy of the study area can be effectively improved, the data error can be reduced, and the method has universal applicability. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 This is a flow chart of a method for predicting geostress in shale reservoirs according to an embodiment of the present invention;
[0036] Figure 2 This is a schematic diagram of the structure of a device for predicting geostress in shale reservoirs according to an embodiment of the present invention;
[0037] Figure 3 This is a correlation diagram of the P-wave time difference and S-wave time difference of the shale target layer of Well G2 of the present invention;
[0038] Figure 4 It is the dynamic and static correction diagram of the rock parameters of the shale reservoir of the present invention;
[0039] Figure 5 This is a single well cross-sectional diagram of reservoir rock mechanical parameters of Well G1 of the present invention;
[0040] Figure 6 The correlation analysis between the measured minimum horizontal stress sample points and the logging parameters of the present invention;
[0041] Figure 7 The correlation analysis between the measured maximum horizontal stress sample points and the logging parameters of the present invention;
[0042] Figure 8 This is a diagram of the ground stress prediction results of different G2 methods of the present invention;
[0043] Fig. 9 This is a diagram of the ground stress prediction results of different G4 methods of the present invention;
[0044] Fig.10 This is a diagram of the ground stress prediction results of G5 using different methods of the present invention.
[0045] Among them: 1-data acquisition module; 2-data correction and standard processing module; 3-measurement point value return module; 4-static rock mechanics calculation module; 5-ground stress learning sample expansion module; 6-ground stress BP neural network prediction module; 7-shear wave time difference prediction module; 8-ground stress evaluation module. DETAILED DESCRIPTION
[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0047] Figure 1 FIG. 1 is a flow chart of a method for predicting geostress in a shale reservoir according to an embodiment of the present invention. Figure 1 As shown, the present invention provides a shale reservoir ground stress prediction method, comprising:
[0048] Step S1: obtaining geophysical logging parameter data of multiple shale oil and gas wells, and obtaining X-mac cross-dipole acoustic logging data of one of the wells;
[0049] Step S2: performing environmental correction on the oil and gas wells described in step S1, and performing standardization processing on the acquired geophysical logging data;
[0050] Step S3: Acquire the indoor experimental data of all oil and gas wells and perform depth homing correction;
[0051] Step S4: According to the X-mac cross-dipole acoustic logging data obtained in step S1 and the geophysical logging data processed in step S2, the parameter data of the shear wave curve is predicted by the linear regression method;
[0052] Step S5: According to the geophysical logging data obtained in step S2, in combination with the shear wave time difference obtained in step S4, dynamic rock mechanical parameters are calculated jointly, and according to the experimental data obtained in step S3, dynamic and static corrections are performed on the dynamic rock mechanical parameters to obtain static rock mechanical parameters;
[0053] Step S6: fitting the poroelastic coefficient (biot coefficient) and pore pressure according to the experimental data of step S3, and jointly calculating the experimental stress data with the static rock mechanics parameters obtained in step S5, and performing error analysis;
[0054] Step S7: expanding the calculation result obtained in step S6 into a learning sample of geostress data, taking the geophysical logging parameters of a single well as a training sample, performing neural network calculation, and obtaining logging stress data;
[0055] Step S8: Use logging stress data to establish a geological grid model, perform spatial interpolation on the geophysical logging data of all oil and gas wells, calculate a three-dimensional geostress field model using a neural network method, and predict, analyze and evaluate the spatial geostress distribution of shale reservoirs based on the obtained calculation model.
[0056] In this prediction method, all oil and gas wells need to undergo geophysical logging. The geophysical logging parameter data include natural gamma, natural potential, deep lateral resistivity, deep lateral resistivity, apparent resistivity, well diameter, density and P-wave time difference. In addition, one of the oil and gas wells is selected to obtain the X-mac cross-dipole acoustic logging data of the well, including continuous curve data such as shear wave time difference and P-wave time difference.
[0057] It should be noted that all oil and gas wells need to be environmentally corrected using the standard layer statistical analysis method, and the geophysical logging data of all oil and gas wells are standardized using the histogram method.
[0058] The statistical analysis method of the standard layer requires statistical analysis of the relationship between the target curve and the reference curve of the standard layer. One or more logging curves that are less affected by the wellbore are used as the mother curve, also called the reference curve, and the curve to be corrected is the target curve. The correlation between the mother curve and the target curve is established in the non-expanded section, and the target curve to be corrected across the well section is predicted. The functional relationship is established using the multivariate regression analysis method, and the equation is solved to obtain the coefficients of each reference curve. Then, a mathematical model is established using the obtained coefficients and the reference curve. This method is to apply the mathematical model in the wellbore expansion section to obtain a new curve.
[0059] The histogram method first makes a frequency histogram of each logging curve of the standard layer of each well in the study area, and then uses the histogram of the core well or a specific well as the standard to compare the histograms of other wells with it. If the two have the same values and similar shapes, then their scale is accurate; if there is a significant difference between the two and the lithology of the standard layer is stable, then there is an error in its scale, and the difference between the response values of the standard layer and the core well standard layer is the correction amount.
[0060] In practical applications, the parameters of oil and gas wells measured by indoor experiments include rock parameters and stress value parameters. The depth reposition correction is to correspond and correct the depth data corresponding to the rock parameters and stress value parameters measured by indoor experiments and the depth data of geophysical logging through the reposition diagram method.
[0061] The regression diagram method is to plot the data of indoor experiments into a rod diagram and compare it with the trend of the logging curve at the same depth. When the best agreement is achieved, the distance the core moves up and down is the core regression correction value; the core analysis density is plotted into a rod diagram and compared with the logging curve at the same depth. When a good agreement is achieved, the core movement amount is used as the correction regression value. The regression diagram method can display the regression depth difference more intuitively.
[0062] In this prediction method, dynamic rock mechanics parameters include dynamic Young's modulus, dynamic Poisson's ratio, bulk modulus and shear modulus, and static rock mechanics parameters include static Young's modulus and static Poisson's ratio.
[0063] In practical applications, according to different methods of obtaining rock mechanical parameters, rock mechanical parameters are divided into two types, namely static parameters and dynamic parameters. Among them, static parameters are rock parameters obtained by uniaxial or triaxial loading tests of rock samples in the laboratory, and dynamic rock parameters are mechanical properties parameters exhibited under various dynamic loads or cyclically changing loads, such as sound waves, impact and vibration, etc. The rock parameters calculated using logging curve data are dynamic rock parameters.
[0064] In this prediction method, the theoretical relationship between P-wave and S-wave velocities and rock dynamic parameters is given according to the P-wave and S-wave propagation equations. The P-wave time difference Δtp and S-wave time difference Δts are fitted by the logging data through the linear regression prediction method. The bulk density ρ is obtained by density logging, and various rock mechanical parameters can be calculated, including dynamic Young's modulus E, bulk modulus K, shear modulus G, and dynamic Poisson's ratio μ.
[0065] It should be noted that the elastic modulus calculated directly through sound waves and density data is dynamic, and there is a certain gap between it and the static mechanical properties of rock. It is necessary to use laboratory data analysis to obtain a conversion formula to convert the dynamic elastic modulus into a static modulus, which is generally a linear conversion.
[0066] In practical applications, the vertical geostress, horizontal maximum stress and horizontal minimum stress of multiple wells in the study area are calculated based on density logging curve data and static rock parameters. Since the experimental method is used to determine geostress, the number of measurements is limited and a continuous geostress profile cannot be obtained. However, logging data has the characteristics of good continuity and high resolution, so this method is easy to promote.
[0067] First, using the density logging curve data, the vertical geostress is obtained by calculating the integral of formation thickness and density. The calculation expression is:
[0068]
[0069] Among them, σ vis the vertical geostress, MPa; h is the depth, m; ρ(h) is the density logging value, g / cm 3 ; g is the acceleration due to gravity, m / s 2 .
[0070] Then, the formation pore pressure is obtained, which is obtained by fitting the relationship between pressure and depth based on the indoor experimental measurement data.
[0071] Next, we obtain the biot coefficient, also known as the porous elastic coefficient, which is obtained by fitting the data measured by indoor experiments. First, we use the volume content and porosity of each mineral measured by indoor experiments as input conditions to solve the biot coefficient, which is the ratio of the deformation of the pore space of the formation to the total volume change. The expression of the biot coefficient is:
[0072] α=1-K dry / K m
[0073] Where: α is the biot coefficient; Kdry is the dry rock bulk modulus, GPa; Km is the mineral bulk modulus, GPa.
[0074] The maximum horizontal stress and the minimum horizontal stress are calculated by combining rock parameters, biot coefficient and pore pressure. The calculation expressions are:
[0075]
[0076]
[0077] In the formula, σ H is the maximum horizontal principal stress, MPa; σ h is the minimum horizontal principal stress, MPa; υ is Poisson's ratio, dimensionless; σ v is the overburden pressure, MPa; α is the biot coefficient, dimensionless; P p is the pore pressure, MPa; E is Young's modulus, GPa; ξ h is the strain in the direction of the minimum horizontal principal stress, dimensionless; ξ H is the strain in the direction of the maximum horizontal principal stress, dimensionless, obtained by fitting the experimental data.
[0078] It should be noted that the overburden pressure is vertical ground stress.
[0079] It should be noted that the calculation of geostress is a complex technical task. Currently, there is no calculation formula suitable for all wells in the same area. These calculation formulas are usually called geostress calculation models. This indirect calculation method has a certain deviation from the actual geostress value, and other methods are needed to correct its results.
[0080] In practical applications, in order to avoid the problem of deviation in calculating the maximum horizontal stress and the minimum horizontal stress using the geostress calculation formula, the calibration and prediction advantages of the BP neural network are used to establish an implicit relationship between the reservoir geostress and conventional logging data, and finally a new geostress prediction model for the region is formed.
[0081] In this prediction method, the learning samples obtained through step S7 are relatively small, the data is discontinuous, and there will be strong multi-solutions. The data rules between the measuring points have a certain randomness. To solve this problem, the neural network method and the calculation formula method are organically combined, that is, the continuous data results of the wells with smaller errors in the ground stress results calculated by the calculation formula method are used as sample data for machine learning, thereby realizing the expansion of sample data from discontinuous small core samples to continuous large logging samples, further reducing the multi-solutions of the prediction, and maintaining a good correlation between the prediction results and the logging curves.
[0082] Figure 2 FIG. 1 is a schematic diagram of the structure of a device for predicting shale reservoir in-situ stress according to an embodiment of the present invention. Figure 2 As shown, based on this shale reservoir geostress prediction method, a device for coordinated application is also provided, specifically comprising: a data acquisition module 1, wherein the data acquisition module 1 is respectively connected to a data correction and standard processing module 2 and a measuring point value return module 3, wherein the measuring point value return module 3 is respectively connected to a static rock mechanics obtaining module 4, a geostress learning sample expansion module 5 and a geostress BP neural network method prediction module 6, wherein the data correction and standard processing module 2 is connected to a shear wave time difference prediction module 7, wherein the shear wave time difference prediction module 7 is connected to a static rock mechanics obtaining module 4, wherein the static rock mechanics obtaining module 4 is respectively connected to a data correction and standard processing module 2 and a geostress learning sample expansion module 5, wherein the geostress learning sample expansion module 5 is connected to a geostress BP neural network method prediction module 6, wherein the geostress BP neural network method prediction module 6 is connected to a geostress evaluation module 8.
[0083] The prediction method is particularly provided with an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the shale reservoir geostress prediction method is implemented. At the same time, the device also provides a computer-readable storage medium storing the computer program. When the computer program is executed by the processor, the steps of the shale reservoir geostress prediction method are implemented.
[0084] Compared with the prior art, the present invention has the following beneficial effects:
[0085] (1) The present invention comprehensively utilizes various data such as core analysis data, indoor mechanical experiments, conventional logging, XMAC logging, etc., and innovatively proposes a point, line, and volume geostress prediction method combining artificial intelligence with traditional geomechanics on the basis of research on structure, sedimentary phase, and reservoir properties, thereby realizing the expansion from small core samples to large logging samples;
[0086] (2) The present invention predicts and analyzes the spatial geostress distribution of shale reservoirs in the test area by establishing a three-dimensional geostress prediction model based on neural network learning. The relative error between the geostress prediction result and the measured data is within 3%, which improves the geostress prediction accuracy of the study area. The research results provide an important basis for the design of geological engineering integration schemes.
[0087] In summary, the present invention has the advantages of simple logic, accuracy and reliability, and has high practical value and promotion value in the technical field of shale reservoir stress prediction.
[0088] A shale oil reservoir in the northern part of the Songliao Basin is a typical large-scale continental freshwater fault-depression lake basin. The Qingshankou Formation deposition period was the heyday of the basin depression development, forming a semi-deep lake-deep lake phase organic-rich fine-grained sedimentary rock, mainly mud-grade shale with a particle size of less than 0.004mm, and the cumulative thickness of other lithologies such as siltstone and carbonate rock accounts for less than 10% of the total thickness. Among them, the depression area with a vitrinite reflectance greater than 1.2% and a formation pressure coefficient greater than 1.2 has formed a typical continental shale-type shale oil and gas belt with rich organic matter, high maturity, high formation pressure and low density. Taking this shale oil reservoir as an example, the following describes in detail the practical application of this shale reservoir ground stress prediction method and device.
[0089] Step S1: obtaining geophysical logging parameter data of multiple shale oil and gas wells, and obtaining X-mac cross-dipole acoustic logging data of one of the wells.
[0090] In this embodiment, the data used are logging data of 5 wells in the oil reservoir block. The 5 wells are specifically G1, G2, G3, G4 and G5. The drilling data has a total of 8 logging curves, which are geophysical logging parameters such as natural gamma, natural potential, deep lateral resistivity, deep lateral resistivity, apparent resistivity, well diameter, density and P-wave time difference. The sampling interval is 0.125m, and the coverage range is more than 2550 meters from the surface to underground. In addition, G2 well has X-mac cross-dipole acoustic logging data, including continuous curve data such as shear wave time difference and P-wave time difference.
[0091] Step S2: Perform environmental correction on the oil and gas wells described in step S1, and perform standardization on the acquired geophysical logging data.
[0092] In this embodiment, a standard layer statistical analysis method is used to perform environmental correction, which is specifically divided into the following steps:
[0093] First, eliminate the depth data outside the target layer, which is 30 meters above and below the oil and gas well, and only retain the shale reservoir target layer data to facilitate subsequent data processing; secondly, draw the transverse profile of the geophysical logging curves such as wellbore diameter, gamma, longitudinal wave time difference and density of multiple wells; thirdly, through transverse comparison, find out the wells or depth sections with abnormal well diameters, and analyze the degree to which different curves are affected by the well diameter; fourthly, use one or more logging curves that are less affected by the wellbore as the mother curve, and reconstruct other curves of the depth section with larger errors at the expansion point through empirical formulas or linear relationships.
[0094] Meanwhile, in this embodiment, Figure 5 This is a single well profile diagram of the reservoir rock mechanical parameters of the G1 well of the present invention, such as Figure 5 As shown in the figure, by using the linear correlation analysis of the data from wells other than G5, the linear relationship between the logarithm of the P-wave time difference and the logarithm of the density is poor, while the linear relationship between the logarithm of the P-wave time difference and the logarithm of the natural gamma is good. The P-wave time difference curve of the expanded section of Well G5 was corrected using the established linear relationship.
[0095] Step S3: Acquire the indoor experimental data of all oil and gas wells and perform depth homeostasis correction.
[0096] In this embodiment, uniaxial compression and triaxial rock mechanics experimental tests were carried out in 5 wells in the work area, and the depth of the indoor experimental core depths of the 5 coring wells in the work area was corrected, with an adjustment rate of 2.33%. The depth correction amount was not large, with an average of 0.5m and a maximum of 1.2m, which was within the allowable correction error range.
[0097] Step S4: Based on the X-mac cross-dipole acoustic logging data obtained in step S1 and the geophysical logging data processed in step S2, the parameter data of the shear wave curve is predicted by the linear regression method.
[0098] In this embodiment, Figure 3 This is a correlation diagram of the longitudinal wave time difference and the shear wave time difference of the shale target layer in the G2 well of the present invention, as shown in Figure 3 As shown in the figure, based on the linear regression of the shear wave time difference curve data and the longitudinal wave time difference curve data of the wells belonging to the Q1-Q7 layer of the G2 well, the shear wave data calculation expression of the shale in this work area is established. The calculation error of the regression equation is R 2 =0.897, with a high correlation, which can be used as a reference for calculating the shear wave time difference curve data of other wells without X-MAC cross-dipole acoustic logging data. The expression of the shear wave time difference is:
[0099]
[0100] Among them, △t s and △t p It is the time difference between shear wave and longitudinal wave of the formation.
[0101] In this embodiment, the shear wave time difference curve data of the X-MAC cross-dipole acoustic logging in the Q8-Q9 layer of the G2 well is used as the posterior data for error analysis. The results show that the shear wave predicted by linear regression is highly consistent with the measured shear wave, and the relative error of the shear wave time difference prediction is within 2%, indicating that the effect of linear regression in predicting the shear wave time difference curve data is relatively reliable, which provides a basis for the next step of shale reservoir rock parameter calculation.
[0102] Step S5: Based on the geophysical logging data obtained in step S2 and the shear wave time difference obtained in step S4, dynamic rock mechanics parameters are calculated jointly. Based on the experimental data obtained in step S3, dynamic and static corrections are performed on the dynamic rock mechanics parameters to obtain static rock mechanics parameters.
[0103] (1) The expression of dynamic Young's modulus E is:
[0104]
[0105] Here, ρ represents the bulk density.
[0106] (2) The expression of dynamic Poisson's ratio μ is:
[0107]
[0108] (3) The expression of shear modulus G is:
[0109]
[0110] (4) The expression of bulk modulus K is:
[0111]
[0112] The deformation and failure of rocks are relatively slow static processes. All rock mechanics analyses require the determination of static parameters to predict the deformation and failure of rocks.
[0113] Figure 4 The dynamic and static correction diagram of the shale reservoir rock parameters of the present invention is as follows: Figure 4As shown in the figure, dynamic and static correction of dynamic rock mechanical parameters calculated by logging is realized. According to the corrected static rock parameter results, the rock parameter profile of a single well is drawn. The calculation results of static Young's modulus of well G1 have certain similarities with the morphology of acoustic logging curve and density logging curve. The static Poisson's ratio curve has a negative correlation with the morphology of acoustic logging curve and density logging curve, which is consistent with the trend of indoor experimental data, indicating that the calculation results reflect the rock mechanical properties of the reservoir well. According to the calculation results, the rock parameters of each single well in the study area can be evaluated.
[0114] Step S6: According to the experimental data of step S3, the porous elastic coefficient (biot coefficient) and pore pressure are fitted and obtained, and the static rock mechanics parameters obtained in step S5 are used to jointly calculate the experimental stress data and perform error analysis.
[0115] In this embodiment, the biot coefficient and the effective porosity are linearly fitted to establish the biot coefficient expression of the shale reservoir:
[0116] α=0.386·ln(φ eff )+1.743
[0117] Among them: α is the biot coefficient, dimensionless; Ф eff is the effective porosity of shale, %.
[0118] In practical applications, effective porosity is calculated based on logging curves, and then the biot coefficient of the single well and modeling area in the whole area is obtained. In this embodiment, the error analysis of the ground stress results calculated by logging is carried out according to the measured data of the ground stress of the shale oil reservoir in the northern Songliao Basin. Table 1 is a formula method based on logging curve parameters to calculate the ground stress and error statistics. As shown in Table 1, the difference between the calculated stress value and the measured stress value in the indoor experiment is divided by the percentage of the measured stress value in the indoor experiment.
[0119] Table 1 Calculation of ground stress and error statistics based on the formula method of logging curve parameters
[0120]
[0121] Through error analysis, the overall error of vertical stress is within 5%, which meets the error requirement. However, the errors of the calculated minimum horizontal stress of G2, G4, and G5 and the maximum horizontal stress of G5 are large, greater than 5%.
[0122] Step S7: Expand the calculation result obtained in step S6 into a learning sample of geostress data, use the geophysical logging parameters of a single well as a training sample, perform neural network calculation, and obtain logging stress data.
[0123] In this embodiment, Figure 6This is the correlation analysis between the measured minimum horizontal stress sample points and the logging parameters of the present invention, such as Figure 6 As shown in the figure, there are only 48 measured geostress data points in the work area, the learning samples are relatively small, and the data are discontinuous. The neural network method is organically combined with the calculation formula method. The specific approach is: select the continuous vertical stress, maximum horizontal stress and minimum horizontal stress calculation result data of the two wells G1 and G3 representing different depths, and the stress value errors calculated by the formula method are all within 3%. The 48 geostress parameter data rows measured indoors are expanded to form a large sample learning sample data.
[0124] In practical applications, Figure 7 This is the correlation analysis between the measured maximum horizontal stress sample points and the logging parameters of the present invention, such as Figure 7 As shown in the figure, after determining the training input parameters of the above machine learning, a BP neural network model consisting of 1 input layer, 2 hidden layers and 1 output layer built by Matlab is used, and the Levenberg-Marquardt backpropagation training method is selected for training. Figure 8 This is a diagram of the ground stress prediction results of different G2 methods of the present invention. Fig. 9 This is a diagram of the ground stress prediction results of different G4 methods of the present invention. Fig.10 The G5 in-situ stress prediction results of different methods of the present invention are shown in FIG. Figures 8 to 10 As shown in the figure, the horizontal maximum principal stress and the horizontal minimum principal stress of wells G2, G4 and G5 are predicted, and the ground stress prediction results of the three wells are obtained.
[0125] Table 2 Comparison of ground stress prediction errors of different learning sample neural network models
[0126]
[0127] After error statistics, Table 2 is obtained. Table 2 is a comparison table of errors in predicting ground stress by neural network models with different learning samples. Error analysis shows that, as shown in Table 2, the average error of the maximum horizontal principal stress is 4.73% and the average error of the minimum horizontal principal stress is 10.28% when the BP neural network is predicted only by indoor experimental data. Compared with the error of calculating ground stress by the formula method based on logging curve parameters, the error has not decreased. The average error of the maximum horizontal principal stress predicted by the BP neural network using the expanded large sample is 1.54%, and the average error of the minimum horizontal principal stress is 1.45%.
[0128] Step S8: Use logging stress data to establish a geological grid model, perform spatial interpolation on the geophysical logging data of all oil and gas wells, calculate a three-dimensional geostress field model using a neural network method, and predict, analyze and evaluate the spatial geostress distribution of shale reservoirs based on the obtained calculation model.
[0129] In this embodiment, the prediction error of the BP neural network using the expanded large sample is significantly smaller than the prediction error using the small sample. Therefore, this method can be used to expand the learning sample of the indoor experimental data using the in-situ stress calculation data of the small error well, convert the small sample into a large sample, and then perform the BP neural network prediction. The prediction result obtained in this way is more accurate, and this method can be used to predict and comprehensively evaluate the in-situ stress size of other wells in the area.
[0130] The shale reservoir geostress prediction method and device provided by the present invention have strong practicability and accurate and reliable results. By acquiring a variety of information on oil and gas wells, the sample size of experimental data can be expanded. By establishing a three-dimensional geostress prediction model based on neural network learning, the geostress prediction accuracy of the study area can be effectively improved, the data error can be reduced, and the method has universal applicability.
[0131] The above implementation modes are only used to illustrate the present invention, but not to limit the present invention. Ordinary technicians in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, all equivalent technical solutions also belong to the scope of the present invention. The patent protection scope of the present invention should be defined by the claims.
Claims
1. A method for predicting geostress in shale reservoirs, characterized in that: include: Step S1: obtaining geophysical logging parameter data of multiple shale oil and gas wells, and obtaining X-mac cross-dipole acoustic logging data of one of the wells; Step S2: performing environmental correction on the oil and gas wells described in step S1, and performing standardization processing on the acquired geophysical logging data; Step S3: Acquire the indoor experimental data of all oil and gas wells and perform depth homing correction; Step S4: According to the X-mac cross-dipole acoustic logging data obtained in step S1 and the geophysical logging data processed in step S2, the parameter data of the shear wave curve is predicted by the linear regression method; Step S5: According to the geophysical logging data obtained in step S2, in combination with the shear wave time difference obtained in step S4, dynamic rock mechanical parameters are calculated jointly, and according to the experimental data obtained in step S3, dynamic and static corrections are performed on the dynamic rock mechanical parameters to obtain static rock mechanical parameters; Step S6: fitting the poroelastic coefficient (biot coefficient) and pore pressure according to the experimental data of step S3, and jointly calculating the experimental stress data with the static rock mechanics parameters obtained in step S5, and performing error analysis; Step S7: expanding the calculation result obtained in step S6 into a learning sample of geostress data, taking the geophysical logging parameters of a single well as a training sample, performing neural network calculation, and obtaining logging stress data; Step S8: Use logging stress data to establish a geological grid model, perform spatial interpolation on the geophysical logging data of all oil and gas wells, calculate a three-dimensional geostress field model using a neural network method, and predict, analyze and evaluate the spatial geostress distribution of shale reservoirs based on the obtained calculation model.
2. The shale reservoir in-situ stress prediction method according to claim 1, characterized in that: The geophysical logging parameter data of the oil and gas wells include natural gamma, natural potential, deep lateral resistivity, deep lateral resistivity, apparent resistivity, well diameter, density and compressional wave time difference.
3. The method for predicting shale reservoir in-situ stress according to claim 1, characterized in that: All the oil and gas wells are subjected to environmental correction using the standard layer statistical analysis method, and the geophysical logging data of all the oil and gas wells are subjected to standardization processing using the histogram method.
4. The method for predicting shale reservoir in-situ stress according to claim 1, characterized in that: The parameters of the oil and gas wells measured by the indoor experiment include rock parameters and stress value parameters. The depth reposition correction is to correspond and correct the depth data corresponding to the rock parameters and stress value parameters measured by the indoor experiment and the depth data of geophysical logging through the reposition diagram method.
5. The method for predicting shale reservoir in-situ stress according to claim 1, characterized in that: The dynamic rock mechanics parameters include dynamic Young's modulus, dynamic Poisson's ratio, bulk modulus and shear modulus, and the static rock mechanics parameters include static Young's modulus and static Poisson's ratio.
6. The method for predicting shale reservoir in-situ stress according to claim 1, characterized in that: The maximum horizontal ground stress and the minimum horizontal ground stress of the indoor experiment are calculated by combining rock parameters, biot coefficient and pore pressure. The calculation expressions are: σ H is the maximum horizontal principal stress, MPa; σ h is the minimum horizontal principal stress, MPa; υ is Poisson's ratio, dimensionless; σ v is the overburden pressure, MPa; α is the biot coefficient, dimensionless; P p is the pore pressure, MPa; E is Young's modulus, GPa; ξ h is the strain in the direction of the minimum horizontal principal stress, dimensionless; ξ H is the strain in the direction of maximum horizontal principal stress, dimensionless.
7. The method for predicting shale reservoir in-situ stress according to claim 6, characterized in that: The biot coefficient is the ratio of the pore space deformation of the formation to the total volume change. The expression of the biot coefficient is: α=1-K dry / K m Where: α is the biot coefficient; Kdry is the dry rock bulk modulus, GPa; Km is the mineral bulk modulus, GPa.
8. The method for predicting shale reservoir in-situ stress according to claim 6, characterized in that: The overburden stratum pressure is the vertical geostress, and the expression of the vertical geostress is: Among them, σ v is the vertical geostress, MPa; h is the depth, m; ρ(h) is the density logging value, g / cm 3 ; g is the acceleration due to gravity, m / s 2 .
9. A device based on the shale reservoir in-situ stress prediction method according to any one of claims 1 to 8, characterized in that: include: A data acquisition module (1), wherein the data acquisition module (1) is respectively connected to a data correction and standard processing module (2) and a measuring point value return module (3), wherein the measuring point value return module (3) is respectively connected to a static rock mechanics obtaining module (4), a geostress learning sample expansion module (5) and a geostress BP neural network method prediction module (6), wherein the data correction and standard processing module (2) is connected to a shear wave time difference prediction module (7), wherein the shear wave time difference prediction module (7) is connected to the static rock mechanics obtaining module (4), wherein the static rock mechanics obtaining module (4) is respectively connected to the data correction and standard processing module (2) and the geostress learning sample expansion module (5), wherein the geostress learning sample expansion module (5) is connected to the geostress BP neural network method prediction module (6), wherein the geostress BP neural network method prediction module (6) is connected to a geostress evaluation module (8).
Citation Information
Cited By
Method and system for quantitatively representing connectivity between injection and production wells by using machine learning
CN119712063A
Deep coal reservoir top and bottom plate fracturing packer well logging evaluation method
CN120145238A
