Comprehensive recognition method and system for limestone reservoir fractures

Abstract

The invention discloses a comprehensive recognition method and system for limestone reservoir fractures, and belongs to the technical field of carbonate oil and gas reservoir development. The identification method comprises the following steps: identifying a crack based on rock core observation, obtaining crack development characteristics and a fraction dimension D, evaluating the crack development degree, and judging whether a target area is suitable for a crack identification model or not; obtaining the response characteristics of the crack on the logging curve based on the crack development characteristics and the rock core and imaging logging information; establishing a crack identification model by using a BP neural network method based on the response characteristics of the crack on the logging curve; when the target area is suitable for the crack recognition model, the crack development condition of the target area is recognized through the crack recognition model, and comprehensive recognition of cracks is completed.

Description

Technical Field

[0001] The invention belongs to the technical field of carbonate oil and gas reservoir development, and in particular relates to a limestone reservoir fracture comprehensive identification method and system. Background Art

[0002] Since the 1980s, during the exploration of Paleozoic clastic rocks in the basin, good gas-bearing shows have been found in Taiyuan limestone. It has always been believed that Taiyuan limestone is mainly composed of bioclastic micrite limestone, and the reservoir space types are mainly fractures and dissolved pores, and primary pores are not developed. However, due to the low level of research on Taiyuan limestone gas reservoirs, there are still factors that restrict further exploration, such as unclear reservoir formation mechanism and distribution law. First, for low-porosity and low-permeability unconventional limestone reservoirs, natural fractures are not only favorable reservoir spaces, but also effective seepage channels for natural gas migration. The degree of fracture development will directly affect the mining efficiency of limestone gas reservoirs; secondly, natural fractures are easily reactivated and opened during hydraulic fracturing, which directly affects the direction, angle and pressure of hydraulic fracturing, has a guiding effect on induced fractures, is conducive to the formation of network fractures, and thus affects the overall fracturing effect.

[0003] At present, the prediction of reservoir fractures is mainly divided into single well fracture identification and inter-well fracture prediction. Single well fracture identification is mainly evaluated through core and logging data. Core fracture observation and description is the most intuitive method to reflect the true situation of fractures; conventional logging data such as natural gamma logging, dual lateral resistivity logging, acoustic time difference logging, compensated neutron logging, compensated density logging, natural potential logging, resistivity logging, caliper logging, electromagnetic wave logging, formation dip logging, etc., the response characteristics of fracture development sections with different occurrences on conventional logging have relative numerical differences; the application of array sensing, acoustic and electrical imaging, and cross multi-pole array acoustic logging technology can more clearly and intuitively carry out the characteristics of single well fractures. Although core and imaging logging technology has high recognition accuracy, it is high in cost and small in number, making it difficult to objectively evaluate the development and distribution of reservoir fractures. Although most wells have conventional logging data, the recognition accuracy of conventional logging fractures is relatively low. The prediction of interwell fractures is mainly achieved based on seismic data. Fractured rocks in underground reservoirs will cause anisotropy in seismic wave propagation, which will be displayed to a certain extent in seismic attribute data. Seismic attributes contain information on fracture development zones. Selecting seismic attributes that are sensitive to fractures (such as coherence bodies, curvatures, ant bodies, etc.) can directly predict the characteristics of fracture distribution. However, seismic data can only identify large and medium-scale fractures, and cannot accurately predict small-scale fractures. In addition, seismic attributes are affected by lithology, fluids, etc., and there are multiple solutions to the identification results of fractures.

[0004] In summary, how to accurately characterize the characteristics of fracture development is of great significance in the detailed description of the Taiyuan Formation limestone gas reservoir. However, due to its complex genesis and small scale, a single core, seismic or logging reservoir fracture identification method cannot meet the needs of exploration and development. A systematic, fast and effective optimization method has not been formed for the effective identification of reservoir fractures. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a comprehensive identification method and system for limestone reservoir fractures, so as to solve the problems of strong multi-solution and low accuracy of the existing fracture identification methods and further explore the distribution law of the Taiyuan Formation limestone reservoir.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] The present invention provides a comprehensive identification method for limestone reservoir fractures, comprising the following steps:

[0008] Identify fractures based on core observation, obtain fracture development characteristics and fractal dimension D, evaluate fracture development degree, and determine whether the target area is suitable for fracture identification model;

[0009] Based on the fracture development characteristics and core and imaging logging data, the response characteristics of fractures on the logging curve are obtained;

[0010] Based on the response characteristics of fractures on logging curves, a fracture identification model was established using the BP neural network method.

[0011] When the target area is suitable for the fracture identification model, the fracture identification model is used to identify the fracture development in the target area to complete the comprehensive identification of fractures.

[0012] In a specific implementation process, the fracture development characteristics include fracture types;

[0013] The fracture types include fracture types with different causes, fracture types with different angles between the fracture surface and the plane perpendicular to the core axis, and fracture types with different degrees of secondary mineral filling;

[0014] The types of cracks with different causes include structural cracks, dissolution cracks, sutures, weathering cracks and closed cracks;

[0015] The fracture types with different angles between the fracture surface and the plane perpendicular to the core axis include horizontal fractures, low-angle fractures, high-angle fractures and vertical fractures;

[0016] The types of fractures with different degrees of filling by secondary minerals include filled fractures, semi-filled fractures and unfilled fractures.

[0017] In a specific implementation process, the fracture development characteristics include fracture parameters;

[0018] The fracture parameters include fracture geometric parameters and fracture density and porosity; the fracture geometric parameters include fracture depth, width, length, inclination, orientation and properties.

[0019] In the specific implementation process, the porosity of the fracture is obtained based on the dual lateral resistivity curve, as shown in the following formula:

[0020]

[0021]

[0022] In the formula, R d —deep laterolog measurement value; Rs—shallow laterolog measurement value; R mf —Mud resistivity.

[0023] In a specific implementation process, the density of the cracks is represented by line density, surface density and volume density.

[0024] In the specific implementation process, the determination method of the fractal dimension D includes: scaling transformation method, covering method, density correlation function method, embedded space method and grid covering method.

[0025] In the specific implementation process, the method of measuring the fractal dimension D by the grid covering method is as follows:

[0026] Small squares with the same side length are used as units to cover the cracks on the core section, and then the number of small squares containing cracks is counted. The scale of the side length is gradually changed, and the number of objects with the specified characteristics is counted respectively;

[0027] The least square method is used to regress the number of objects with specified characteristics corresponding to the statistics to obtain the regression line. When the square of its linear correlation coefficient is greater than 0.98, it is considered that the distribution of core fractures has fractal characteristics, and the slope of the corresponding regression line is the value of the fractal dimension D of the core fracture distribution.

[0028] In a specific implementation process, the response characteristics of the fracture on the logging curve include natural gamma GR, compensated density DEN, compensated neutron CNL, acoustic time difference AC, caliper curve CAL, deep lateral logging measurement value Rd and shallow lateral logging measurement value Rs.

[0029] In the specific implementation process, the BP neural network includes an input layer, a hidden layer and an output layer;

[0030] The input layer is the response characteristics of the fracture on the well logging curve;

[0031] The output layer is a parameter of the crack development, wherein when the crack is not developed, the output is 0; when the crack is developed, the output is 1;

[0032] The hidden layer uses 10 nodes;

[0033] The hidden layer and the output layer respectively use the tanh function and the sigmoid function as activation functions.

[0034] The present invention also provides a comprehensive identification system for limestone reservoir fractures, based on any one of the comprehensive identification methods for limestone reservoir fractures, comprising:

[0035] Evaluation and judgment module, used to identify fractures based on core observation, obtain fracture development characteristics and fractional dimension D, evaluate fracture development degree, and judge whether the target area is suitable for fracture identification model;

[0036] Obtaining a corresponding feature module for obtaining the response characteristics of the fracture on the logging curve based on the fracture development characteristics and core and imaging logging data;

[0037] Establish a fracture identification model module, which is used to establish a fracture identification model based on the response characteristics of fractures on the logging curve using the BP neural network method;

[0038] The comprehensive identification module is used to identify the crack development in the target area using the crack identification model when the target area is suitable for the crack identification model, so as to complete the comprehensive identification of the cracks.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The present invention provides a comprehensive identification method for limestone reservoir fractures. First, the most accurate fracture identification data core is used for identification. The type and scale of fracture development are determined by fracture development characteristics. The fractal dimension D is further used to evaluate the degree of fracture development. The density and connectivity of fractures are reflected on the basis of considering the geometric relationship of fracture space. Due to the complexity of carbonate rock fracture distribution and the scarcity of core data, the true development degree and distribution of reservoir fractures cannot be accurately reflected only by core data. Further, based on fracture development characteristics and core and imaging logging data, the response characteristics of fractures on logging curves are obtained and a fracture identification model is established using a BP neural network method. The model is used to identify reservoir fractures, which can accurately characterize the fracture development characteristics. It is a systematic, fast and effective optimization method for reservoir fracture identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flow chart of a comprehensive identification method for limestone reservoir fractures provided by an embodiment of the present invention;

[0042] Figure 2 is a schematic diagram of the development characteristics of fractures on a rock core provided by an embodiment of the present invention;

[0043] Figure 3 It is the response characteristics of conventional and imaging logging of fractures in the study area provided by the embodiment of the present invention;

[0044] Figure 4 is a schematic diagram of the gradient descent method provided by an embodiment of the present invention;

[0045] Figure 5 It is a schematic diagram of the BP neural network method provided by an embodiment of the present invention;

[0046] Figure 6 This is a BP neural network crack recognition result diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0048] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0049] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0050] The present invention provides a method for comprehensive identification of limestone reservoir fractures, comprising the following steps:

[0051] S1: Identify fractures based on core observation, obtain fracture development characteristics and fractal dimension D, evaluate fracture development degree, and judge whether the target area is suitable for fracture identification model; that is, identify fractures based on core observation, obtain fracture development characteristics and fractal dimension D, evaluate fracture development degree, and judge whether fracture identification based on BP neural network can be carried out;

[0052] Fracture development characteristics include fracture types and fracture parameters.

[0053] Fracture types include fracture types with different causes such as tectonic fractures, dissolution fractures, suture lines, weathering fractures and closed fractures, fracture types with different angles between the fracture surface and the plane perpendicular to the core axis such as horizontal fractures, low-angle fractures, high-angle fractures and vertical fractures, and fracture types with different degrees of secondary mineral filling such as filled fractures, semi-filled fractures and unfilled fractures.

[0054] Fracture parameters include fracture geometry parameters, fracture density and porosity;

[0055] The geometric parameters of a fracture include its depth, width, length, inclination, orientation and properties.

[0056] The method of determining the fractal dimension D by the grid covering method is as follows:

[0057] Small squares with the same side length are used as units to cover the cracks on the core section, and then the number of small squares containing cracks is counted. The scale of the side length is gradually changed, and the number of objects with the specified characteristics is counted respectively;

[0058] The least square method is used to regress the number of objects with specified characteristics corresponding to the statistics to obtain the regression line. When the square of its linear correlation coefficient is greater than 0.98, it is considered that the distribution of core fractures has fractal characteristics, and the slope of the corresponding regression line is the value of the fractal dimension D of the core fracture distribution.

[0059] The geological data obtained include core data and thin-section data, the logging data include conventional logging data and imaging logging data, and the seismic data include three-dimensional seismic data. Specifically, the core and thin-section data of the Taiyuan Group limestone reservoir are obtained for observation and description. In underground rock formations, cores are the most accurate fracture identification data, but the cost of obtaining cores is high. Therefore, it is necessary to carry out detailed fracture description for the few cores and thin sections with coring wells, including the identification of fracture types, the measurement of fracture geometric parameters, the calculation of fracture density and porosity, and the use of fractal dimension D value to evaluate fracture development.

[0060] Through core and thin section observations, the observed cracks are first classified. Carbonate rock cracks are generally divided into structural cracks, dissolution cracks, sutures, weathering fractures, closed cracks (cracks), etc. according to their causes; according to the different angles between the crack surface and the plane perpendicular to the core axis, they can be divided into: horizontal cracks, low-angle cracks, high-angle cracks and vertical cracks; according to the degree of secondary mineral filling, they can be divided into filled cracks, semi-filled cracks and unfilled cracks.

[0061] Among them, the types of cracks include geological cracks such as structural cracks, dissolution cracks, and suture lines. Structural cracks often have straight crack surfaces and regular shapes on the core, and early cracks are often filled with calcite. Dissolution cracks have uneven walls on the core, with different widths and short extension lengths, and often have pore-fracture structures. The width of the suture line is mostly less than 0.01mm, most of them are parallel to the limestone layer, often in the shape of a suture, and the cracks are mostly filled with calcite.

[0062] Among them, the geometric parameters of the crack include the depth, width, length, inclination and orientation of the crack. The measured geometric parameters of the crack are generally expressed in the form of a histogram. The crack width and length histograms can show the size variation range of the crack, and then determine the type and scale of the crack development; the crack inclination histogram can provide the stress state of the crack formation.

[0063] Fracture density and porosity can be calculated based on the fracture parameters obtained from the core description. Fracture density and fracture porosity are commonly used parameters to evaluate the degree of macroscopic fracture development in cores. Fracture density can be expressed by line density, surface density, and volume density. Among them, the porosity of fractures is mainly obtained based on the dual lateral resistivity curve:

[0064]

[0065]

[0066] Where R d ——deep laterolog measurement value; Rs——shallow laterolog measurement value; R mf ——Mud resistivity.

[0067] The fractal dimension D is used to evaluate the degree of crack development and assess whether the target area is suitable for crack identification based on neural networks.

[0068] Among them, the degree of fracture development is mainly evaluated based on the fractal dimension D value. Due to the complexity of carbonate fracture distribution, coupled with the limited diameter of the core and the influence of factors such as incomplete preservation, the fracture density cannot accurately reflect the actual development and distribution of reservoir fractures. French mathematician B. Mandelbort defined fractals as: its components are similar to the whole in some form, and used the fractal dimension D value to quantitatively describe the complexity of the fractal. A large number of studies have shown that the rock fracture process has good self-similarity, and fractal theory can be used to characterize the development of fractures. In order to quantitatively study fractal problems, a fractal set can be defined.

[0069]

[0070] Where N n ——the number of objects with the specified characteristics; r n ——linear dimension of characteristic object; C——proportional constant; D——fractional dimension.

[0071] Take the natural logarithm of both sides of the above equation

[0072] n n =lnC-Dlnr n

[0073] Visible ln N n with ln r n It is a linear relationship, and the fractal dimension D is the slope of the straight line.

[0074] There are several commonly used methods for measuring fractal dimension: scaling method, covering method, density correlation function method, embedded space method, grid covering method, etc. The grid covering method is used to measure the fractal dimension of Taiyuan Formation limestone reservoir fractures.

[0075] The specific statistical method of the grid covering method is as follows: Use small unit squares with a side length of r to cover the cracks on the core section, then count the number of small squares containing cracks, gradually change the scale of r, and count the corresponding Nn values. If the core crack distribution has fractal characteristics, then ln Nn and ln r n The least squares method is used to regress the statistical data. If the square of the linear correlation coefficient is greater than 0.98, the distribution of core fractures is considered to have fractal characteristics, and the slope of the corresponding regression line is the fractal dimension D value of the core fracture distribution.

[0076] The fractal dimension value D reflects the density and connectivity of the cracks on the basis of considering the spatial geometric relationship of the cracks. Generally speaking, the larger the fractal dimension value D is, the better the connectivity of the cracks in the area is, and a good fracture network system can be formed.

[0077] S2: Based on the fracture development characteristics and core and imaging logging data, the response characteristics of fractures on the logging curve are obtained; that is, the fracture results identified by core and imaging logging data are calibrated to obtain the response characteristics of fractures on the conventional logging curve, and the fracture sensitivity curve analysis and curve preprocessing are carried out; among them, the response characteristics include natural gamma GR, compensation density DEN, compensation neutron CNL, acoustic wave time difference AC, caliper curve CAL, deep lateral logging measurement value Rd and shallow lateral logging measurement value Rs. Preprocess the logging curves of the whole area, including standardization and normalization, to ensure the rationality of the logging curve data.

[0078] Due to the fact that the logging curves of some old wells have instrument calibration errors, abnormal operations, etc., the same logging curve has large differences between wells for the same formation, and the process of eliminating the above errors is usually called standardization. The essence of logging data standardization is to use the standard layer of the same formation system in the same oil field or region. The layer has stable deposition, wide distribution, the same deposition environment and similar geological and geophysical parameter distribution characteristics, so the logging response of the same logging curve is often consistent. The embodiments adopt the method histogram method, frequency cross plot method, three porosity (density logging, compensated neutron, acoustic time difference) cross plot method, etc.

[0079] Furthermore, due to various reasons, the logging curves often have burr interference. The logging curves should be filtered to filter out these fluctuations and burr interferences that are not related to the formation properties.

[0080] Furthermore, the numerical differences of logging data with different characteristics are also very large. Some characteristics change linearly, some change exponentially, and some change logarithmically. The difference in data will inevitably cause data with different characteristics to participate in the calculation with different weights. In addition, since the dimensions of various logging data are inconsistent, data with different dimensions cannot be processed in the same mathematical way. Therefore, the selected data must be normalized and scaled within a unified numerical range, such as [0, 1]. Otherwise, the experimental results will be seriously affected.

[0081] Among them, the linear normalization formula is as follows

[0082]

[0083] Logarithmic Normalization Formula

[0084]

[0085] X is the normalized logging curve value, X∈[0,1], X * is the original logging data, and are the maximum and minimum values ​​of the logging curve respectively.

[0086] Combine core and imaging logging data to obtain the response characteristics of fractures on conventional logging. Take the well section with core data and imaging logging as the standard well section, compare the fracture identification results obtained by core and imaging logging, that is, the development characteristics of fractures obtained by core observation and identification with the parameters on the conventional logging curve, analyze the response characteristics of fractures on conventional logging, and select the logging curve that is sensitive to fracture response.

[0087] S3: A fracture identification model is established based on the response characteristics of fractures on the logging curve using the BP neural network method; the BP neural network includes an input layer, a hidden layer and an output layer; the present invention mainly uses a single hidden layer BP neural network method, specifically: the BP neural network is a multi-layer feedforward network trained by error back propagation, which mainly consists of an input layer, an output layer and one or more hidden layers. In essence, the BP algorithm uses the square of the network error as the objective function and uses the gradient descent method to calculate the minimum value of the objective function.

[0088] In the construction of BP neural network, it is necessary to assume that x1, x2...x n is the input variable, y is the output variable, u is the output of the hidden layer neuron, and f is the mapping relationship of the activation function. The activation function refers to the function that the input variable x must pass through when passing through the neuron. Common activation functions include sigmod function, tanh function, ReLU function, etc. The present invention uses sigmod function and tanh function

[0089] Among them, the tanh function expression is as follows

[0090]

[0091] The Sigmod function expression is as follows

[0092]

[0093] BP neural network is mainly divided into two processes: forward propagation and reverse gradient calculation. In the forward propagation, ν ij is the weight of the i-th input variable and the j-th hidden layer neuron, θ u j is the bias of the jth neuron in the hidden layer u. Let ω j is the weight of the connection between the jth neuron and y, θ y is the bias of y, then we can get j The expression is:

[0094]

[0095]

[0096] Let J be the objective function, is the real output value. In order to minimize the error between the predicted value and the real value, assuming that the kth object is used as the target, the expression of the objective function J can be obtained:

[0097]

[0098] It should also be noted that the gradient descent method means that the error continues to decrease in the direction of the derivative until the gradient remains basically unchanged (such as Figure 4 ). In this process, we mainly focus on νij, θ j u ,ω j ,θ y These four parameters are optimized, so the optimization expression is obtained according to the gradient descent method (where μ is the learning rate):

[0099]

[0100]

[0101]

[0102]

[0103] After forward propagation, the next step is to carry out the reverse gradient process. Reverse gradient is to find the gradient of the objective function for each parameter, that is, the derivative. By finding the gradient of νij, θju, ωj, and θy, the reverse gradient descent formula is obtained by deduction:

[0104]

[0105]

[0106]

[0107]

[0108] The input layer is the response characteristics of the fracture on the well logging curve;

[0109] The output layer is a parameter of the crack development, wherein when the crack is not developed, the output is 0; when the crack is developed, the output is 1;

[0110] The hidden layer uses 10 nodes;

[0111] The hidden layer and the output layer respectively use the tanh function and the sigmoid function as activation functions.

[0112] like Figure 5As shown in the figure, specifically, the BP neural network method is used to establish a fracture recognition model. According to the conventional logging response of fractures, integrated core observation, imaging logging and other data, the typical fracture data are used as the sample set for BP neural network training. The seven conventional logging curve values ​​of natural gamma GR, compensated density DEN, compensated neutron CNL, acoustic time difference AC, caliper curve CAL, deep and shallow lateral RILD and RILS are selected as the characteristic value x as the output layer; the output is a parameter that characterizes the fracture recognition situation, 0 means that the fracture is not developed, and 1 means that the fracture is developed, that is, y; the hidden layer uses 10 nodes; the network hidden layer and output layer use tanh function and sigmod function as activation functions respectively.

[0113] A learning sample set consisting of identified fracture samples of the Taiyuan Formation limestone segment is selected, with part of the samples used as a validation set and part of the samples used as a prediction set. The maximum number of iterations and allowable error are set, and the neural network is trained to make the model meet the predetermined requirements.

[0114] S4: When the target area is suitable for the fracture identification model, the fracture identification model is used to identify the fracture development in the target area to complete the comprehensive identification of the fractures.

[0115] The present invention also provides a comprehensive identification system for limestone reservoir fractures, including an evaluation and judgment module, a corresponding feature acquisition module, a fracture identification model establishment module and a comprehensive identification module;

[0116] Evaluation and judgment module, used to identify fractures based on core observation, obtain fracture development characteristics and fractional dimension D, evaluate fracture development degree, and judge whether the target area is suitable for fracture identification model;

[0117] Obtaining a corresponding feature module for obtaining the response characteristics of the fracture on the logging curve based on the fracture development characteristics and core and imaging logging data;

[0118] Establish a fracture identification model module, which is used to establish a fracture identification model based on the response characteristics of fractures on the logging curve using the BP neural network method;

[0119] The comprehensive identification module is used to identify the crack development in the target area using the crack identification model when the target area is suitable for the crack identification model, so as to complete the comprehensive identification of the cracks.

[0120] Example

[0121] The reservoir fractures in this example are located in the limestone section of the Taiyuan Formation in the central and eastern part of the Ordos Basin. After gas testing, many wells in this area have obtained industrial gas flow, and the exploration potential is relatively large. Fractures, as an important reservoir space type, are frequently developed in this area, but the current identification and prediction methods for the Taiyuan Formation limestone are of low accuracy, which seriously restricts the exploration and development process.

[0122] Therefore, in view of the problems existing in the prior art, Figure 1 The present invention is described in detail.

[0123] Core and thin-section data of the Taiyuan Formation limestone reservoir were obtained for observation and description. In underground rock formations, cores are the most accurate data for fracture identification, but the cost of obtaining cores is high. Therefore, detailed fracture descriptions should be carried out for the few cores and thin sections with coring wells, including the identification of fracture types, measurement of fracture geometric parameters, calculation of fracture density and porosity, and the use of fractal dimension D value to evaluate fracture development.

[0124] like Figure 2 As shown, the development characteristics of cracks on the core are obtained. The effective cracks in the Taiyuan Formation limestone in this embodiment are mainly structural cracks and dissolution cracks. The crack density is mainly distributed between 0-15 / m, with an average of 8 / m; most of the crack widths are less than 0.5 mm, accounting for 59.3%; the crack inclination is 59-85°, with an average of 72°, and the crack azimuth is 311-375°, with an average of 340°; the cracks are mainly high-angle cracks, which are completely filled with calcite or mud, accounting for 74.9%, and semi-filled account for 25.1%.

[0125] By measuring the fractal dimension D values ​​of 61 cores, it was found that the average fractal dimension D value in the study area was 1.2, and the fractures were relatively densely developed, which is suitable for fracture identification based on neural networks.

[0126] like Figure 3 As shown in the figure, the response characteristics of conventional and imaging logging of fractures in the study area are obtained. By comparing the fracture identification results obtained by core and imaging logging with the parameters on the conventional logging curve, it is found that natural gamma GR, compensated density DEN, compensated neutron CNL, acoustic wave time difference AC, caliper curve CAL, deep and shallow lateral RILD and RILS have obvious response characteristics in the well section with developed fractures in the Taiyuan Formation limestone.

[0127] Combined with the above crack evaluation work, a crack identification model was established using the BP neural network method.

[0128] In this embodiment, the BP neural network is designed as shown in the figure, and a three-layer BP neural network is used, as shown in the figure, namely, 1 input layer, 1 hidden layer and 1 output layer. Natural gamma GR, compensation density DEN, compensation neutron CNL, acoustic wave time difference AC, caliper curve CAL, deep and shallow lateral RILD and RILS. These seven conventional well logging extracted fracture characteristic parameters are used as the input of the network.

[0129] Furthermore, the output is a parameter that characterizes the development of cracks, that is, 1 node is selected, 0 indicates that cracks are not developed, 1 indicates that cracks are developed, and 10 nodes are selected for the hidden layer. The tanh function and sigmoid function are used as activation functions for the hidden layer and output layer of the network, respectively.

[0130] The fracture identification model is used to evaluate the fracture development of the well section to be identified.

[0131] A learning sample set consisting of the identified fracture samples of the Taiyuan Formation limestone layer was selected, with part of the samples used as the validation set and part of the samples used as the prediction set. The maximum number of iterations and the allowable error were set to train the neural network so that the model could meet the predetermined requirements. Figure 6 As shown in the figure, based on the trained fracture recognition model, 18 samples were selected from the limestone layer in the study area for fracture identification. Compared with the imaging display, the coincidence rate reached 83.3%.

[0132] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A comprehensive identification method for limestone reservoir fractures, characterized in that: The following steps are involved: Identify fractures based on core observation, obtain fracture development characteristics and fractal dimension D, evaluate fracture development degree, and determine whether the target area is suitable for fracture identification model; Based on the fracture development characteristics and core and imaging logging data, the response characteristics of fractures on the logging curve are obtained; Based on the response characteristics of fractures on logging curves, a fracture identification model was established using the BP neural network method. When the target area is suitable for the fracture identification model, the fracture identification model is used to identify the fracture development in the target area to complete the comprehensive identification of fractures.

2. The comprehensive identification method of limestone reservoir fractures according to claim 1, characterized in that: The fracture development characteristics include fracture types; The fracture types include fracture types with different causes, fracture types with different angles between the fracture surface and the plane perpendicular to the core axis, and fracture types with different degrees of secondary mineral filling; The types of cracks with different causes include structural cracks, dissolution cracks, sutures, weathering cracks and closed cracks; The fracture types with different angles between the fracture surface and the plane perpendicular to the core axis include horizontal fractures, low-angle fractures, high-angle fractures and vertical fractures; The types of fractures with different degrees of filling by secondary minerals include filled fractures, semi-filled fractures and unfilled fractures.

3. The comprehensive identification method of limestone reservoir fractures according to claim 1, characterized in that: The fracture development characteristics include fracture parameters; The fracture parameters include fracture geometric parameters and fracture density and porosity; the fracture geometric parameters include fracture depth, width, length, inclination, orientation and properties.

4. The comprehensive identification method of limestone reservoir fractures according to claim 3 is characterized in that: The porosity of the fracture is obtained based on the dual lateral resistivity curve, as shown in the following formula: In the formula, R d —deep laterolog measurement value; Rs—shallow laterolog measurement value; R mf —Mud resistivity.

5. The comprehensive identification method of limestone reservoir fractures according to claim 3 is characterized in that: The density of the cracks is represented by line density, surface density and volume density.

6. The comprehensive identification method of limestone reservoir fractures according to claim 1, characterized in that: The methods for determining the fractal dimension D include: a scaling transformation method, a covering method, a density correlation function method, an embedded space method, and a grid covering method.

7. The comprehensive identification method of limestone reservoir fractures according to claim 6, characterized in that: The method for determining the fractal dimension D by the grid covering method is as follows: Small squares with the same side length are used as units to cover the cracks on the core section, and then the number of small squares containing cracks is counted. The scale of the side length is gradually changed, and the number of objects with the specified characteristics is counted respectively; The least square method is used to regress the number of objects with specified characteristics corresponding to the statistics to obtain the regression line. When the square of its linear correlation coefficient is greater than 0.98, it is considered that the distribution of core fractures has fractal characteristics, and the slope of the corresponding regression line is the value of the fractal dimension D of the core fracture distribution.

8. The comprehensive identification method of limestone reservoir fractures according to claim 1, characterized in that: The response characteristics of the fracture on the logging curve include natural gamma GR, compensated density DEN, compensated neutron CNL, acoustic time difference AC, caliper curve CAL, deep lateral logging measurement value Rd and shallow lateral logging measurement value Rs.

9. The comprehensive identification method of limestone reservoir fractures according to claim 1, characterized in that: The BP neural network includes an input layer, a hidden layer and an output layer; The input layer is the response characteristics of the fracture on the well logging curve; The output layer is a parameter of the crack development, wherein when the crack is not developed, the output is 0; when the crack is developed, the output is 1; The hidden layer uses 10 nodes; The hidden layer and the output layer respectively use the tanh function and the sigmoid function as activation functions.

10. A comprehensive identification system for limestone reservoir fractures, characterized in that: The comprehensive identification method of limestone reservoir fractures according to any one of claims 1 to 9 comprises: Evaluation and judgment module, used to identify fractures based on core observation, obtain fracture development characteristics and fractional dimension D, evaluate fracture development degree, and judge whether the target area is suitable for fracture identification model; Obtaining a corresponding feature module for obtaining the response characteristics of the fracture on the logging curve based on the fracture development characteristics and core and imaging logging data; Establish a fracture identification model module, which is used to establish a fracture identification model based on the response characteristics of fractures on the logging curve using the BP neural network method; The comprehensive identification module is used to identify the crack development in the target area using the crack identification model when the target area is suitable for the crack identification model, so as to complete the comprehensive identification of the cracks.

Images