Single well crustal stress profile analysis method and device combined with neural network algorithm

Abstract

The invention is suitable for the technical field of oil and gas field development, and discloses a single well ground stress profile analysis method, device and equipment combined with a neural network algorithm and a medium, and the method comprises the steps: obtaining first logging data of a single well with dipole acoustic logging data in a shale oil reservoir block; performing lithology classification on the first logging data to obtain second logging data; performing linear fitting on longitudinal wave velocity data and transverse wave velocity data in the second logging data based on lithology of a rock stratum in a single well to obtain a conversion relation of the longitudinal wave velocity data and the transverse wave velocity data; obtaining target logging data of a single well without dipole acoustic logging data in the shale oil reservoir block; and determining a ground stress profile corresponding to the target logging data based on the conversion relation of the longitudinal and transverse wave velocity data. The method can effectively solve the problem that the fitting degree of the linear longitudinal and transverse wave conversion relation is low in the calculation process of a conventional well logging interpretation crustal stress calculation method, and improves the accuracy of a crustal stress profile interpretation result.

Description

Technical Field

[0001] This invention relates to the field of oil and gas field development technology, and in particular to a method, apparatus, equipment and medium for single-well geostress profile analysis combined with neural network algorithms. Background Technology

[0002] Shale oil reservoirs are characterized by thin interlayers, well-developed bedding, and complex stress distribution, making them difficult to develop. Therefore, there is an urgent need for rock mechanical parameters.

[0003] There are many methods for obtaining in-situ stress in the field of petroleum engineering, which can be divided into four main categories: laboratory testing, field testing, well logging interpretation, and other methods. ① Laboratory testing mainly includes differential strain testing, Kaiser acoustic emission testing, wave velocity anisotropy testing, and paleomagnetic orientation testing. ② Field testing mainly includes ground fracturing experiments, hydraulic fracturing, stress relief methods, and stress recovery methods. ③ Well logging interpretation mainly includes acoustic logging interpretation, wellbore collapse method, and wellbore induced fracture interpretation. ④ Other methods include geological data analysis and seismic data prediction.

[0004] Currently, the commonly used method for single-well geostress profile analysis is sonic logging interpretation. However, most wells only undergo conventional sonic logging, which can only obtain P-wave velocity data. The S-wave velocity required for the interpretation of rock mechanics parameters needs to be calculated using empirical conversion formulas. For some strata in shale reservoirs, the S-wave data calculated by conventional linear / exponential empirical conversion formulas deviates greatly from the actual measurement results, leading to inaccurate geostress profile interpretation results. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a single-well geostress profile analysis method, apparatus, equipment, and medium that incorporates a neural network algorithm. This effectively solves the problem of low fitting degree of the linear P-wave and S-wave conversion relationship in conventional well logging interpretation geostress calculation methods, thereby improving the accuracy of geostress profile interpretation results.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a single-well geostress profile analysis method combining a neural network algorithm, comprising:

[0008] Acquire the first logging data of a single well with dipole sonic logging data within a shale oil reservoir block;

[0009] The first logging data is classified by lithology to obtain the second logging data;

[0010] Based on the lithology of the rock strata in a single well, linear fitting was performed on the P-wave velocity data and S-wave velocity data in the second logging data to obtain the conversion relationship between the P-wave velocity data and the S-wave velocity data.

[0011] To obtain target logging data for single wells in shale oil reservoir blocks that lack dipole sonic logging data;

[0012] Based on the conversion relationship between P-wave and S-wave velocity data, the geostress profile corresponding to the target well logging data is determined.

[0013] Furthermore, based on the lithology of the rock formations within a single well, linear fitting was performed on the P-wave velocity data and S-wave velocity data in the second logging data to obtain the conversion relationship between the P-wave and S-wave velocity data, including:

[0014] Based on the lithology of the rock formations within a single well, the P-wave velocity data and S-wave velocity data in the second logging data are grouped.

[0015] Linear fitting was performed on the P-wave velocity data and S-wave velocity data corresponding to each type of lithology using linear regression equations and the least squares method.

[0016] Based on the fitting results of linear fitting, the conversion relationship between P-wave and S-wave velocity data is determined.

[0017] Furthermore, based on the fitting results of the linear fitting, the transformation relationship of the P-wave and S-wave velocity data is determined, including:

[0018] The linear regression coefficient R corresponding to the fitting result of the linear fit 2 Lithologies with ≥80% lithology are identified as the first lithology set;

[0019] The linear fitting equations corresponding to each type of lithology in the first lithology set are retained as the first transformation relationship of the P-wave and S-wave velocity data corresponding to that type of lithology.

[0020] The linear regression coefficient R corresponding to the fitting result of the linear fit 2 <80% of the lithology is identified as the second lithology set;

[0021] Based on the logging data and BP neural network model corresponding to the second lithology set, the second transformation relationship of the P-wave and S-wave velocity data corresponding to each type of lithology in the second lithology set;

[0022] Based on the first and second conversion relationships of the P-wave and S-wave velocity data, the conversion relationships of the P-wave and S-wave velocity data are determined.

[0023] Furthermore, based on the conversion relationship between P-wave and S-wave velocity data, the geostress profile corresponding to the target well logging data is determined, including:

[0024] Based on the conversion relationship between P-wave and S-wave velocity data, determine the corresponding S-wave velocity data for the target logging data;

[0025] Obtain the target layer density logging data for the single well corresponding to the target logging data;

[0026] Based on the target logging data and the target layer density logging data, combined with the combined spring model, the corresponding geostress value is calculated.

[0027] Based on the value of geostress, determine the geostress profile corresponding to the target well logging data.

[0028] Secondly, the present invention also provides a single-well geostress profile analysis device combining a neural network algorithm, comprising:

[0029] The acquisition module is used to acquire the first logging data of a single well with dipole sonic logging data in a shale oil reservoir block;

[0030] The classification module is used to classify the first logging data by lithology to obtain the second logging data.

[0031] The fitting module is used to perform linear fitting on the P-wave velocity data and S-wave velocity data in the second logging data based on the lithology of the rock strata in a single well, so as to obtain the conversion relationship between the P-wave velocity data and the S-wave velocity data.

[0032] The acquisition module is also used to acquire target logging data for single wells in shale oil reservoir blocks that do not have dipole sonic logging data;

[0033] The determination module is used to determine the geostress profile corresponding to the target well logging data based on the conversion relationship between P-wave and S-wave velocity data.

[0034] Furthermore, the fitting module is also used for:

[0035] Based on the lithology of the rock formations within a single well, the P-wave velocity data and S-wave velocity data in the second logging data are grouped.

[0036] Linear fitting was performed on the P-wave velocity data and S-wave velocity data corresponding to each type of lithology using linear regression equations and the least squares method.

[0037] Based on the fitting results of linear fitting, the conversion relationship between P-wave and S-wave velocity data is determined.

[0038] Furthermore, based on the fitting results of the linear fitting, the transformation relationship of the P-wave and S-wave velocity data is determined, including:

[0039] The linear regression coefficient R corresponding to the fitting result of the linear fit 2 Lithologies with ≥80% lithology are identified as the first lithology set;

[0040] The linear fitting equations corresponding to each type of lithology in the first lithology set are retained as the first transformation relationship of the P-wave and S-wave velocity data corresponding to that type of lithology.

[0041] The linear regression coefficient R corresponding to the fitting result of the linear fit 2 <80% of the lithology is identified as the second lithology set;

[0042] Based on the logging data and BP neural network model corresponding to the second lithology set, the second transformation relationship of the P-wave and S-wave velocity data corresponding to each type of lithology in the second lithology set;

[0043] Based on the first and second conversion relationships of the P-wave and S-wave velocity data, the conversion relationships of the P-wave and S-wave velocity data are determined.

[0044] Furthermore, the module is also used for:

[0045] Based on the conversion relationship between P-wave and S-wave velocity data, determine the corresponding S-wave velocity data for the target logging data;

[0046] Obtain the target layer density logging data for the single well corresponding to the target logging data;

[0047] Based on the target logging data and the target layer density logging data, combined with the combined spring model, the corresponding geostress value is calculated.

[0048] Based on the value of geostress, determine the geostress profile corresponding to the target well logging data.

[0049] Thirdly, the present invention also provides an electronic device, comprising: a processor and a memory;

[0050] The processor is coupled with the memory;

[0051] The processor is used to read and execute programs or instructions stored in the memory, causing the device to perform the method as described in the first aspect.

[0052] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method as described in the first aspect.

[0053] In summary, the technical solution provided in this application has at least the following technical effects or advantages:

[0054] The technical solution of this invention involves acquiring logging data from single wells within a shale oil reservoir block, classifying the logging data by lithology, performing linear fitting of P-wave and S-wave velocity data corresponding to each lithology, optimizing the P-wave and S-wave fitting relationships for some lithologies using a neural network algorithm, determining the conversion relationship between P-wave and S-wave velocity data, and finally predicting the geostress profile of a single well using a combined spring model. This method only requires collecting readily available field construction data, such as conventional logging data, during oil and gas field development, without requiring additional experimental testing. It effectively solves the problem of low fitting degree of the linear P-wave and S-wave conversion relationship in conventional logging interpretation geostress calculation methods, thus improving the accuracy of geostress profile interpretation results.

[0055] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a schematic diagram of a single-well geostress profile analysis method combining a neural network algorithm in an embodiment of the present invention;

[0058] Figure 2 This is a further flowchart illustrating a single-well geostress profile analysis method combining a neural network algorithm in an embodiment of the present invention.

[0059] Figure 3 This is a fitting diagram of the longitudinal and transverse wave velocity lines of the dolomitic mudstone section of Well X (3610-3652m) in an embodiment of the present invention;

[0060] Figure 4 This is a fitting diagram of the longitudinal and transverse wave velocity lines of the dolomitic sandstone section of Well X (3610-3652m) in an embodiment of the present invention;

[0061] Figure 5 This is a fitting diagram of the longitudinal and transverse wave velocity lines of the silty sandstone section of Well X (3610-3652m) in an embodiment of the present invention;

[0062] Figure 6 This is a fitting diagram of the longitudinal and transverse wave velocity lines of the sandy dolomite section of Well X (3610-3652m) in an embodiment of the present invention;

[0063] Figure 7 This is a schematic diagram illustrating the explanation of the P-S wave velocity BP neural network fitting results of the sandy dolomite section of Well X (3610-3652m) in this embodiment of the invention, and the model training.

[0064] Figure 8 This is a schematic diagram illustrating the interpretation of the P-S wave velocity BP neural network fitting results of the sandy dolomite section of Well X (3610~3652m) in this embodiment of the invention, and a model test.

[0065] Figure 9 This is a schematic diagram illustrating the interpretation and model verification of the P-S wave velocity BP neural network fitting results of the sandy dolomite section of Well X (3610-3652m) in this embodiment of the invention.

[0066] Figure 10 This is a schematic diagram illustrating the interpretation and comprehensive evaluation of the P-S wave velocity BP neural network fitting results of the sandy dolomite section of Well X (3610~3652m) in this embodiment of the invention.

[0067] Figure 11 This is a schematic diagram of a single-well geostress profile analysis device combining a neural network algorithm provided in an embodiment of the present invention;

[0068] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0070] Shale oil reservoirs are characterized by thin, interbedded layers, well-developed bedding, and complex stress distribution, making development challenging and thus creating a strong demand for rock mechanics parameters. A typical shale oil reservoir study block contains multiple wells. Analysis of the geostress profiles corresponding to the strata within the shale oil reservoir within the study block can be achieved using logging data from multiple wells.

[0071] Figure 1 This is a schematic diagram of a single-well geostress profile analysis method combining a neural network algorithm according to an embodiment of the present invention. As shown in the figure, the method includes:

[0072] S101. Obtain the first logging data of a single well with dipole sonic logging data in a shale oil reservoir block.

[0073] For example, using sonic logging, logging data from a single well within a shale oil reservoir block that has dipole sonic logging data is obtained; this is known as the first logging data. The first logging data includes P-wave velocity data, S-wave velocity data, and lithological data of the rock formations corresponding to the well trajectory profile of the single well. The lithological data of the rock formations includes information on rock type, composition, structure, and physical properties.

[0074] S102. Classify the first logging data by lithology to obtain the second logging data;

[0075] For example, based on the lithology of the rock formations within a single well, the first logging data is classified. The classified logging data is the second logging data. The lithology corresponding to the classified second logging data can be represented as [L1, L2, ... L]. n ], n represents the lithology number, L n Indicates lithology.

[0076] S103. Based on the lithology of the rock strata in a single well, linear fitting is performed on the P-wave velocity data and S-wave velocity data in the second logging data to obtain the conversion relationship between the P-wave velocity data and the S-wave velocity data.

[0077] For example, based on the lithology of the rock formations within a single well, the P-wave velocity data and S-wave velocity data in the second logging data are grouped into n1 groups. Each group corresponding to a different lithology contains the corresponding P-wave velocity dataset and S-wave velocity dataset. The P-wave velocity dataset is represented as [X1, X2, ... X...]. m The transverse wave velocity dataset is represented as [Y1, Y2, ... Y]. m [m] represents the amount of data in the dataset. Using a linear regression equation and the least squares method, linear fitting is performed on the P-wave velocity dataset and S-wave velocity dataset corresponding to each type of lithology to obtain the corresponding linear regression equation and regression coefficients R0. 2 The expression for the linear regression equation is:

[0078] Y = a j X+b j (j=1...n1) (1)

[0079] In the formula, Y represents the transverse wave velocity; X represents the longitudinal wave velocity; j represents the data number; and a and b are the correlation coefficients of the linear regression equation.

[0080] Linear regression coefficient R 2 This indicates the degree of linear fit between the P-wave velocity dataset and the S-wave velocity dataset corresponding to each lithology type. Based on the degree of fit, the conversion relationship between the P-wave velocity data and the S-wave velocity data corresponding to the lithology of a single well is obtained.

[0081] The linear regression coefficient R 2 Each lithology type with ≥80% is identified as the first lithology set, denoted as [LR1, LR2, ... LR n1 The linear fitting equations corresponding to each type of lithology in the first lithology set are retained as the conversion formulas for the P-wave and S-wave velocity data corresponding to that type of lithology, thus obtaining the first conversion formula for the P-wave and S-wave velocity data corresponding to each type of lithology in the first lithology set.

[0082] The linear regression coefficient R 2 Each lithology type, representing less than 80% of the total, is identified as a second lithology set, denoted as [LF1, LF2, ... LF]. n2 The logging data corresponding to the second lithology set is the third logging data. Based on the third logging data and the BP neural network model, the second transformation relationship of the P-wave and S-wave velocity data corresponding to each type of lithology in the second lithology set is obtained.

[0083] Based on the second lithological set [LF1, LF2, ... LF] n2 The P-wave velocity and S-wave velocity data in the third logging data are grouped into n² groups. Each group corresponding to a different lithology contains the corresponding P-wave velocity and S-wave velocity datasets. The P-wave velocity dataset is represented as [X1, X2, ... X...]. m1 The transverse wave velocity dataset is represented as [Y1, Y2, ... Y]. m1 ], where m1 is the amount of data in the dataset.

[0084] From the P-wave velocity and S-wave velocity datasets corresponding to each lithology class, 70% of the data was randomly selected as sample data for the BP neural network model. Specifically, for the extracted P-wave velocity and S-wave velocity data, the P-wave velocity data was used as input data for the input layer of the BP neural network model, and the S-wave velocity data was used as sample data for the output layer. From the P-wave velocity and S-wave velocity datasets corresponding to each lithology class, 15% of the data was used for model testing, and 15% was used for model validation.

[0085] Let the hidden layer weight matrix be W. ij The output layer weight matrix is ​​W jk The output of each hidden layer unit is calculated using the following formula:

[0086] O j =f(∑w ij X i -θ j (2)

[0087] Y j =f(∑w jk O j-θ k (3) In the formula, O represents the output value of each hidden layer unit, θ represents the neuron threshold, f represents the nonlinear function, i represents the input layer data sample number, j represents the hidden layer data sample number, k represents the output layer sample number, Y represents the transverse wave velocity, and X represents the longitudinal wave velocity.

[0088] Then, according to the gradient descent criterion of the BP neural network, the error function e is defined. k Then, the weights are updated until the error converges to the expected value, at which point the training ends and the hidden layer weight matrix and the output layer weight matrix are obtained.

[0089]

[0090] W jk =W jk +ηH j e k (5) In the formula, W ij W is the hidden layer weight matrix. jk Let H be the output layer weight matrix, η be the learning rate, and H be the weight matrix. j For the output layer function, e k Let be the error function, i be the input layer data sample label, j be the hidden layer data sample label, and k be the output layer sample label.

[0091] Repeat the above steps until the second lithological set [LF1, LF2, ... LF] is processed. n2 The P-wave velocity and S-wave velocity data corresponding to all lithologies in the dataset are collected, and a hidden layer weight matrix and an output layer weight matrix are generated for each lithology type. This forms the corresponding neural network algorithm. Based on the neural network algorithm, the conversion relationship between the P-wave velocity and S-wave velocity data corresponding to each lithology type is obtained, which yields the second conversion relationship between the P-wave velocity and S-wave velocity data corresponding to each lithology type in the second lithology set.

[0092] The first conversion relationship of the above-mentioned P-wave and S-wave velocity data, combined with the second conversion relationship of the P-wave and S-wave velocity data, together constitute the conversion relationship of P-wave and S-wave velocity data corresponding to the lithology of the rock strata in a single well.

[0093] S104. Obtain the target logging data of a single well in a shale oil reservoir block that does not have dipole sonic logging data.

[0094] Using conventional sonic logging methods, logging data is obtained from single wells in shale oil reservoir blocks that lack dipole sonic logging data. This logging data contains only P-wave velocity measurements. Based on the lithology of the strata within the single well, the logging data is classified, and the classified logging data is the target logging data.

[0095] S105. Based on the conversion relationship between P-wave and S-wave velocity data, determine the geostress profile corresponding to the target well logging data.

[0096] Based on the conversion relationship between the P-wave and S-wave velocity data, the corresponding S-wave velocity data for the target logging data is determined.

[0097] As described above, the conversion relationship between P-wave and S-wave velocity data includes a first conversion relationship between S-wave velocity data and a second conversion relationship between P-wave and S-wave velocity data. The first conversion relationship between S-wave velocity data corresponds to the first lithology set, and the second conversion relationship between P-wave and S-wave velocity data corresponds to the second lithology set.

[0098] Based on the lithology of the rock formations within a single well, the target logging data are classified. The lithology corresponding to the classified logging data can also be represented as [L1, L2, ... L]. n The lithology corresponding to the classified logging data is screened. If the screened lithology is the same as a certain type of lithology in the first lithology set, the corresponding shear wave velocity data is calculated using the first transformation relationship of the P-wave and S-wave velocity data corresponding to that lithology. That is, based on the P-wave velocity data corresponding to that lithology, the corresponding shear wave velocity data is calculated using the linear fitting equation corresponding to that lithology. If the screened lithology is the same as a certain type of lithology in the second lithology set, the corresponding shear wave velocity data is calculated using the second transformation relationship of the shear wave velocity data corresponding to that lithology. That is, based on the P-wave velocity data corresponding to that lithology, the corresponding shear wave velocity data is calculated using a determined neural network algorithm. The shear wave velocity data obtained above is used as part of the target logging data and participates in subsequent calculations.

[0099] Obtain the target layer density logging data for the corresponding single well. Substitute the target logging data and the target layer density logging data into the wave equation to calculate the dynamic Young's modulus E. 动 With dynamic Poisson ratio μ 动 Then, substituting the dynamic elastic parameters into the relevant empirical formulas for dynamic-static transformation of the relevant strata, the static elastic parameter E is obtained. 静 With μ 静 The wave equation is:

[0100]

[0101] In the formula: V P V represents the longitudinal wave velocity; S E represents the transverse wave velocity. 动 μ represents the dynamic Young's modulus. 动 ρ represents the dynamic Poisson's ratio; ρ represents the rock density.

[0102] We can obtain:

[0103]

[0104] The vertical in-situ stress is calculated by integrating the density logging data ρ(h) (ρ(h) is a function of formation density as a function of formation depth).

[0105]

[0106] Calculate the pore pressure value P based on the block pore pressure coefficient. p .

[0107] Calculate the maximum and minimum horizontal principal stresses σ using the formula for the combined spring model. H σ h ,

[0108]

[0109]

[0110] In the formula, σ H For the maximum horizontal principal stress, σ h The minimum principal stress is α, the effective stress coefficient is α, and the structural coefficients are Ex and Ey. 静 μ is the static Young's modulus. 静 This is the static Poisson's ratio.

[0111] By calculating the maximum and minimum principal stresses at the horizontal level, the geostress profile corresponding to the target well logging data can be determined.

[0112] Figure 2 This is a further flowchart illustrating a single-well geostress profile analysis method combining a neural network algorithm, as described in an embodiment of the present invention.

[0113] In summary, the technical solutions in the embodiments of this application have at least the following technical effects or advantages:

[0114] The technical solution of this invention involves acquiring logging data from single wells within a shale oil reservoir block, classifying the logging data by lithology, performing linear fitting of P-wave and S-wave velocity data corresponding to each lithology, optimizing the P-wave and S-wave fitting relationships for some lithologies using a neural network algorithm, determining the conversion relationship between P-wave and S-wave velocity data, and finally predicting the geostress profile of a single well using a combined spring model. This method only requires collecting readily available field construction data, such as conventional logging data, during oil and gas field development, without requiring additional experimental testing. It effectively solves the problem of low fitting degree of the linear P-wave and S-wave conversion relationship in conventional logging interpretation geostress calculation methods, thus improving the accuracy of geostress profile interpretation results.

[0115] The technical solution of the present invention will be further explained below in conjunction with actual application scenarios.

[0116] For the Lucaogou Formation reservoir in a certain block of shale oil reservoir in Jimsar area, the well with dipole sonic logging data is well X, with a depth of 3652.0m. Its adjacent well Y only has conventional logging data and no dipole sonic logging data, with a depth of 3638m.

[0117] (1) Based on well logging data, the lithology of the target formation 3610–3652 m in Well X was classified into four categories, as shown in Table 1. Here, AC represents the longitudinal wave velocity, CNL represents the clay content, Den represents the density, and RT represents the resistivity. Table 1 shows the lithological classification standards for Jimsar shale oil.

[0118] Table 1

[0119] Lithology RT / AC DEN / CNL Dolomitic mudstone 0.08～0.21 ≤0.11 Sandy dolomite 0.21～0.55 ＞0.11 Dolomitic sandstone 0.55～1.3 / fine sandstone 0.21～0.55 ≤0.11

[0120] (2) The dipole sonic logging data of well X were classified according to the four types of lithology.

[0121] (3) Linear fitting was performed on the P-wave and S-wave velocity data corresponding to the four types of lithology, and the results are as follows: Figures 3-6 As shown.

[0122] The fitting formulas for P-wave and S-wave velocities of each lithology are as follows:

[0123] Fine-grained sandstone: y = 0.616x - 0.292 (km / s)R 2 =0.801 (10)

[0124] Sandy dolomite: y = 0.523x + 0.088 (km / s)R 2 =0.633 (11)

[0125] Dolomitic mudstone: y = 0.604x - 0.272 (km / s)R 2 =0.894 (12)

[0126] Dolomitic sandstone: y = 0.578x - 0.203 (Km / s)R 2 =0.877 (13)

[0127] Among them, the correlation coefficients of the linear fitting formulas for fine sandstone, dolomitic mudstone, and dolomitic sandstone are all greater than 0.8, and the relevant formulas can be used in the subsequent calculation of well B; however, the correlation coefficient of the linear fitting formula for sandy dolomite is 0.633, which is less than 0.8, so it is necessary to analyze the conversion relationship between P-wave and S-wave velocities of this type of lithology through a BP neural network model.

[0128] (4) 70% of the P- and S-wave velocity measurement data points from the sandstone and dolomite section of Well X were randomly selected as training data for the BP neural network model, 15% of the data were used for model testing, and 15% of the data were used for model validation. The interpretation graph of the neural network fitting results shows that after training, well testing, and validation, the overall correlation coefficient of the model reached 0.783, which is 23.7% higher than that of the linear fitting formula.

[0129] Figure 7 This is a schematic diagram illustrating the explanation of the P-S wave velocity BP neural network fitting results of the sandy dolomite section of Well X (3610-3652m) in this embodiment of the invention, and the model training.

[0130] Figure 8 This is a schematic diagram illustrating the interpretation of the P-S wave velocity BP neural network fitting results of the sandy dolomite section of Well X (3610~3652m) in this embodiment of the invention, and a model test.

[0131] Figure 9 This is a schematic diagram illustrating the interpretation and model verification of the P-S wave velocity BP neural network fitting results of the sandy dolomite section of Well X (3610-3652m) in this embodiment of the invention.

[0132] (5) For well Y, repeat step one to divide the target layer (3581-3627m) into four lithologies: fine sandstone, dolomitic mudstone, dolomitic sandstone, and sandy dolomite. Substitute the P-wave velocity data corresponding to the well sections with fine sandstone, dolomitic mudstone, and dolomitic sandstone lithologies into formulas (10), (12), and (13) to calculate the corresponding S-wave velocity data.

[0133] (6) Substitute the longitudinal wave velocity data corresponding to the sand and shale section of well Y into the BP neural network model determined in step four, and calculate the corresponding transverse wave velocity data through the neural network algorithm.

[0134] (7) Substitute the target layer density logging data and the sonic logging data calculated above into the wave equation formula (6) to obtain the dynamic Young's modulus E. 动 With dynamic Poisson ratio μ 动 Then, substituting the dynamic elastic parameters into the relevant empirical formulas for dynamic-static transformation of the relevant strata, the static elastic parameter E is obtained. 静 With μ 静 .

[0135] (8) The vertical stress is calculated by integrating the density logging data ρ(h) (ρ(h) is a function of formation density as a function of formation depth).

[0136] (9) Calculate the pore pressure value based on the block pore pressure coefficient (α=1.42).

[0137] (10) Calculate the maximum and minimum principal stresses in the horizontal direction according to the calculation formula of the combined spring model.

[0138] Core samples (core lithology: sandstone dolomite) were taken from Well Y at depths of 3591.5m and 3612.4m and Kaiser experiments were performed. Compared with the in-situ stress data obtained by conventional well logging interpretation methods using only linear fitting formulas, the in-situ stress calculation results obtained by the algorithm of this invention are closer to the corresponding Kaiser core experiment results (as shown in Table 2). Therefore, the prediction results of this embodiment have high reliability, and the prediction method is reliable. Table 2 is a comparison table of interpretation results of different rock mechanics analysis methods for Well Y.

[0139] Table 2

[0140]

[0141] Figure 11 This is a schematic diagram of a single-well geostress profile analysis device combining a neural network algorithm provided in an embodiment of the present invention. As shown in the figure, the device includes:

[0142] The acquisition module is used to acquire the first logging data of a single well with dipole sonic logging data in a shale oil reservoir block;

[0143] The classification module is used to classify the first logging data by lithology to obtain the second logging data.

[0144] The fitting module is used to perform linear fitting on the P-wave velocity data and S-wave velocity data in the second logging data based on the lithology of the rock strata in a single well, so as to obtain the conversion relationship between the P-wave velocity data and the S-wave velocity data.

[0145] The acquisition module is also used to acquire target logging data for single wells in shale oil reservoir blocks that do not have dipole sonic logging data;

[0146] The determination module is used to determine the geostress profile corresponding to the target well logging data based on the conversion relationship between P-wave and S-wave velocity data.

[0147] For example, the fitting module is also used for:

[0148] Based on the lithology of the rock formations within a single well, the P-wave velocity data and S-wave velocity data in the second logging data are grouped.

[0149] Linear fitting was performed on the P-wave velocity data and S-wave velocity data corresponding to each type of lithology using linear regression equations and the least squares method.

[0150] Based on the fitting results of linear fitting, the conversion relationship between P-wave and S-wave velocity data is determined.

[0151] For example, based on the fitting results of linear fitting, the transformation relationship of P-wave and S-wave velocity data is determined, including:

[0152] The linear regression coefficient R corresponding to the fitting result of the linear fit 2 Lithologies with ≥80% lithology are identified as the first lithology set;

[0153] The linear fitting equations corresponding to each type of lithology in the first lithology set are retained as the first transformation relationship of the P-wave and S-wave velocity data corresponding to that type of lithology.

[0154] The linear regression coefficient R corresponding to the fitting result of the linear fit 2 <80% of the lithology is identified as the second lithology set;

[0155] Based on the logging data and BP neural network model corresponding to the second lithology set, the second transformation relationship of the P-wave and S-wave velocity data corresponding to each type of lithology in the second lithology set;

[0156] Based on the first and second conversion relationships of the P-wave and S-wave velocity data, the conversion relationships of the P-wave and S-wave velocity data are determined.

[0157] For example, the determining module is also used for:

[0158] Based on the conversion relationship between P-wave and S-wave velocity data, determine the corresponding S-wave velocity data for the target logging data;

[0159] Obtain the target layer density logging data for the single well corresponding to the target logging data;

[0160] Based on the target logging data and the target layer density logging data, combined with the combined spring model, the corresponding geostress value is calculated.

[0161] Based on the value of geostress, determine the geostress profile corresponding to the target well logging data.

[0162] It should be noted that, for ease of explanation, Figure 11 For example, only the main modules of the device structure are shown. In practical applications, the device may also include modules or components not shown in the figures; the device is not limited to the above-described module structure, but may also be other module structures that implement the above method embodiments.

[0163] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0164] like Figure 12 As shown, the electronic device includes: a processor and a memory;

[0165] The processor is used to read and execute programs and instructions stored in the memory, causing the electronic device to perform the above-described method embodiments.

[0166] It should be noted that, for ease of explanation, Figure 12 Only the main components of the electronic device are shown. In actual applications, the electronic device may also include components or assemblies not shown in the figure.

[0167] The present invention also provides a computer-readable storage medium storing a program or instructions, which, when read and executed by a computer, causes the computer to perform the above-described method embodiments.

[0168] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A single well geostress profile analysis method combined with a neural network algorithm, characterized in that, include: Acquire the first logging data of a single well with dipole sonic logging data within a shale oil reservoir block; The first logging data is classified by lithology to obtain the second logging data; Based on the lithology of the rock strata in a single well, linear fitting is performed on the P-wave velocity data and S-wave velocity data in the second logging data to obtain the conversion relationship between the P-wave velocity data and the S-wave velocity data. To obtain target logging data for single wells in shale oil reservoir blocks that lack dipole sonic logging data; Based on the conversion relationship of the P-wave and S-wave velocity data, the geostress profile corresponding to the target well logging data is determined.

2. The method of claim 1, wherein the neural network algorithm is trained using a plurality of wellbore stress profile data sets. Based on the lithology of the rock formations within a single well, linear fitting is performed on the P-wave velocity data and S-wave velocity data in the second logging data to obtain the conversion relationship between the P-wave and S-wave velocity data, including: Based on the lithology of the rock formations within a single well, the P-wave velocity data and S-wave velocity data in the second logging data are grouped. Linear fitting was performed on the P-wave velocity data and S-wave velocity data corresponding to each type of lithology using linear regression equations and the least squares method. Based on the fitting results of the linear fitting, the conversion relationship between the P-wave and S-wave velocity data is determined.

3. The method of claim 2, wherein the neural network algorithm is trained using a plurality of wellbore stress profile data sets. The determination of the transformation relationship between P-wave and S-wave velocity data based on the fitting results of the linear fitting includes: The linear regression coefficient R corresponding to the fitting result of the linear fitting 2 ≥ 80% of the lithology is determined as the first lithology set; The linear fitting equations corresponding to each type of lithology in the first lithology set are retained as the first transformation relationship of the P-wave and S-wave velocity data corresponding to that type of lithology. The linear regression coefficient R corresponding to the fitting result of the linear fit is... 2 <80% of the lithology is identified as the second lithology set; Based on the logging data and BP neural network model corresponding to the second lithology set, the second transformation relationship of the P-wave and S-wave velocity data corresponding to each type of lithology in the second lithology set; Based on the first conversion relationship of the P-wave and S-wave velocity data and the second conversion relationship of the P-wave and S-wave velocity data, the conversion relationship of the P-wave and S-wave velocity data is determined.

4. The single well geostress profile analysis method incorporating neural network algorithm according to any one of claims 1-3, characterized in that, The process of determining the geostress profile corresponding to the target well logging data based on the conversion relationship of the P-wave and S-wave velocity data includes: Based on the conversion relationship of the P-wave and S-wave velocity data, determine the S-wave velocity data corresponding to the target logging data; Obtain the target layer density logging data of the single well corresponding to the target logging data; Based on the target logging data and the target layer density logging data, combined with the combined spring model, the corresponding geostress value is calculated. Based on the value of the geostress, the geostress profile corresponding to the target well logging data is determined.

5. A single-well geostress profile analysis device combining neural network algorithms, characterized in that, include: The acquisition module is used to acquire the first logging data of a single well with dipole sonic logging data in a shale oil reservoir block; The classification module is used to classify the first logging data into lithology to obtain the second logging data; The fitting module is used to perform linear fitting on the P-wave velocity data and S-wave velocity data in the second logging data based on the lithology of the rock strata in a single well, so as to obtain the conversion relationship between the P-wave velocity data and the S-wave velocity data. The acquisition module is also used to acquire target logging data for single wells in shale oil reservoir blocks that do not have dipole sonic logging data; The determination module is used to determine the geostress profile corresponding to the target well logging data based on the conversion relationship of the P-wave and S-wave velocity data.

6. The single-well geostress profile analysis device combining neural network algorithm according to claim 5, characterized in that, The fitting module is also used for: Based on the lithology of the rock formations within a single well, the P-wave velocity data and S-wave velocity data in the second logging data are grouped. Linear fitting was performed on the P-wave velocity data and S-wave velocity data corresponding to each type of lithology using linear regression equations and the least squares method. Based on the fitting results of the linear fitting, the conversion relationship between the P-wave and S-wave velocity data is determined.

7. The single well geostress profile analysis device combined with neural network algorithm according to claim 6, characterized in that, The determination of the transformation relationship between P-wave and S-wave velocity data based on the fitting results of the linear fitting includes: The linear regression coefficient R corresponding to the fitting result of the linear fitting 2 ≥ 80% of the lithology is determined as the first lithology set; The linear fitting equations corresponding to each type of lithology in the first lithology set are retained as the first transformation relationship of the P-wave and S-wave velocity data corresponding to that type of lithology. The linear regression coefficient R corresponding to the fitting result of the linear fit is... 2 <80% of the lithology is identified as the second lithology set; Based on the logging data and BP neural network model corresponding to the second lithology set, the second transformation relationship of the P-wave and S-wave velocity data corresponding to each type of lithology in the second lithology set; Based on the first conversion relationship of the P-wave and S-wave velocity data and the second conversion relationship of the P-wave and S-wave velocity data, the conversion relationship of the P-wave and S-wave velocity data is determined.

8. The single well geostress profile analysis device combined with neural network algorithm according to any one of claims 5-7, characterized in that, The determining module is further configured to: Based on the conversion relationship of the P-wave and S-wave velocity data, determine the S-wave velocity data corresponding to the target logging data; Obtain the target layer density logging data of the single well corresponding to the target logging data; Based on the target logging data and the target layer density logging data, combined with the combined spring model, the corresponding geostress value is calculated. Based on the value of the geostress, the geostress profile corresponding to the target well logging data is determined.

9. An electronic device, comprising: include: Processor and memory; The processor is coupled to the memory; The processor is configured to read and execute the program or instructions stored in the memory, causing the device to perform the method as described in any one of claims 1-4.

10. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the method as described in any one of claims 1-4.

Images