A method and system for intelligently identifying complex lithofacies in river-facies tight sandstone reservoirs

Abstract

The present application relates to the field of lithofacies identification, and discloses a method and system for intelligently identifying complex lithofacies in river-facies tight sandstone reservoirs, including: establishing a lithofacies identification label library; analyzing lithofacies characteristics and their logging response characteristics to determine the initial limit value of lithofacies characteristics; constructing and training an intelligent identification model; the model includes a limit value module, a radial basis neural network, and a multi-layer perceptron neural network; the limit value module is used to divide the lithofacies identification label library into two sample data sets according to the initial limit value of the lithofacies characteristics; the first data set is input into the radial basis neural network for training, and the second data set is input into the multi-layer perceptron neural network for training; the lithofacies logging data to be identified and the lithofacies characteristics are input into the trained model to obtain a complex lithofacies identification result map. The method has high recognition accuracy and strong generalizability, and provides an effective way for fine anatomical research on reservoir structure and heterogeneity characteristics.

Description

Technical Field

[0001] The present invention relates to the field of lithofacies identification, and in particular to a method and system for intelligently identifying complex lithofacies in fluvial facies tight sandstone reservoirs. Background Art

[0002] Fluvial tight sandstone reservoirs are an important type of oil and natural gas reservoirs. For example, the main gas-bearing layers of the Shanxi Formation and Shihezi Formation in the Sulige Gas Field in the Ordos Basin and the Box 1-3 sections of the Lower Shihezi Formation in the Daniudi Gas Field are all fluvial tight sandstone reservoirs. However, the internal structure of fluvial tight sandstone reservoirs is complex and heterogeneous, and the sedimentary and reservoir characteristics vary greatly between different river types. As a result, oil and gas fields with such reservoirs face key technical problems such as the annual decline in single well production capacity and difficulty in selecting sweet spot reservoirs in the later stages of development. The key to solving these problems is to conduct detailed anatomical studies on sand body structure and its heterogeneity characteristics. As the basic sedimentary unit under specific hydrodynamic conditions, lithofacies is the basic unit for conducting studies on sand body structure and heterogeneity characteristics.

[0003] The concept of lithofacies was first proposed by Miall in 1978, and lithofacies were classified differently according to the lithological type and bedding characteristics of rocks. At present, the identification and characterization of lithofacies are based on the interpretation of single well lithofacies, which can be mainly summarized into three categories: the first is the conventional logging identification method, including the intersection plot method, structural characteristic parameter method, curve overlap method, etc., but these methods generally have a low compliance rate and rely more on the experience and knowledge of the interpreters; the second is the special logging identification method, including the use of formation dip logging data to establish the dip result map mode of different lithofacies for identification, and directly identify lithofacies through imaging logging image data, but the logging data of these two methods are expensive and costly, and cannot be promoted in dense well network blocks; the third is machine learning algorithms such as principal component analysis, cluster analysis and decision trees, but these algorithms lack consideration of geological characteristics, the fit between the logging data characteristics under the constraints of geological characteristics and the algorithm principles, which leads to the lack of rationality demonstration or mechanism understanding of the methods used. In summary, the precise identification and interpretation of complex lithofacies still have problems such as low identification accuracy, difficulty in promotion, and lack of mechanistic evidence.

[0004] Therefore, how to establish an identification method with high recognition accuracy, strong scalability and mechanistic evidence for single-well identification and interpretation of complex lithofacies that can effectively characterize the complex structural characteristics of fluvial facies tight sandstone reservoirs is a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a complex lithofacies intelligent identification method and system for river-facies tight sandstone reservoirs, which has high identification accuracy and strong applicability, and provides an effective way for fine anatomical research on reservoir structure and heterogeneity characteristics. The specific scheme is as follows:

[0006] An intelligent identification method for complex lithofacies of fluvial facies tight sandstone reservoirs, comprising:

[0007] Establish a petrographic identification label library;

[0008] Analyze the lithofacies characteristics and their logging response characteristics based on the lithofacies identification label library to determine the initial limit value of the lithofacies characteristics;

[0009] Constructing an intelligent recognition model; the intelligent recognition model includes a boundary value module, a radial basis neural network and a multi-layer perceptron neural network; the boundary value module is used to receive the initial boundary value of the lithofacies feature, and divide the lithofacies recognition label library into a first sample data set and a second sample data set according to the initial boundary value of the lithofacies feature;

[0010] Training the intelligent recognition model, inputting the first sample data set into the radial basis neural network for training, and inputting the second sample data set into the multilayer perceptron neural network for training;

[0011] The lithofacies logging data and lithofacies characteristics to be identified are input into the trained intelligent identification model to obtain a complex lithofacies identification result map.

[0012] Preferably, in the above-mentioned complex lithofacies intelligent identification method of fluvial facies tight sandstone reservoir provided in the embodiment of the present invention, a lithofacies identification label library is established, including:

[0013] Observe and describe the core of the target layer of the coring well, identify the lithofacies type, and assign specific symbols to each lithofacies type;

[0014] Extract the maximum and minimum values ​​of each logging curve in the target layer, and normalize each logging curve in a min-max standardization manner to map the response values ​​contained in each logging curve to between 0 and 1;

[0015] For each identified lithofacies type, read the normalized logging response values ​​of all logging curves and draw a spider diagram;

[0016] According to the spider web diagram, the sensitivity of different logging curves to different lithofacies types is analyzed, and only a lithofacies identification label library is established.

[0017] Preferably, in the above-mentioned complex lithofacies intelligent identification method of fluvial facies tight sandstone reservoir provided by the embodiment of the present invention, drawing a spider web diagram includes:

[0018] Draw multiple straight lines outward from a point, each straight line represents a well logging curve;

[0019] For any lithofacies type, mark the logging response value point after normalization of the logging curve corresponding to the lithofacies type on each straight line;

[0020] Connect the points of the lithofacies type on each straight line with straight lines to draw a spider web diagram.

[0021] Preferably, in the above-mentioned complex lithofacies intelligent identification method of fluvial facies tight sandstone reservoir provided by the embodiment of the present invention, the sensitivity of different logging curves to different lithofacies types is analyzed according to the spider diagram, including:

[0022] On any straight line of the spider web diagram, if the numerical points corresponding to different lithofacies types on the straight line are closer, the logging curve represented by the straight line is less sensitive to distinguishing different lithofacies types; if the numerical points corresponding to different lithofacies types on the straight line are farther away, the logging curve represented by the straight line is more sensitive to distinguishing different lithofacies types.

[0023] Preferably, in the above-mentioned complex lithofacies intelligent identification method of fluvial facies tight sandstone reservoirs provided in an embodiment of the present invention, the lithofacies characteristics and their logging response characteristics are analyzed based on the lithofacies identification label library to determine the initial limit value of the lithofacies characteristics, including:

[0024] Calculating the lithofacies density and lithofacies frequency of the lithofacies units in the lithofacies identification tag library;

[0025] Calculate the mean density and frequency of similar lithofacies;

[0026] Draw the average lithofacies density distribution map and average lithofacies frequency distribution map of various lithofacies in different river phases;

[0027] Analyze the different logging response characteristics of various lithofacies in different fluvial facies and draw frequency distribution histograms of different logging data;

[0028] According to the average lithofacies density distribution diagram, average lithofacies frequency distribution diagram and different logging data frequency distribution histograms, the distribution of lithofacies characteristics, the influence of surrounding rocks on logging response and the distribution of logging data are analyzed. The lithofacies density limit value and lithofacies frequency limit value are obtained as the initial limit value of lithofacies characteristics based on the comprehensive analysis results.

[0029] Preferably, in the above-mentioned complex lithofacies intelligent identification method of fluvial facies tight sandstone reservoir provided by the embodiment of the present invention, the first sample data set includes sample data whose lithofacies density is greater than the lithofacies density limit value and whose lithofacies frequency is less than the lithofacies frequency limit value;

[0030] The second sample data set includes sample data whose lithofacies density is less than the lithofacies density limit value and whose lithofacies frequency is greater than the lithofacies frequency limit value.

[0031] Preferably, in the above-mentioned intelligent identification method for complex lithofacies of river-facies tight sandstone reservoirs provided in an embodiment of the present invention, the hidden layer activation function of the radial basis neural network is a Softmax function, the output layer activation function is an identity, and the error function is a sum of squares;

[0032] The hidden layer activation function of the multilayer perceptron neural network is a S-type function, the output layer activation function is a Softmax function, and the error function is cross entropy.

[0033] Preferably, in the above-mentioned complex lithofacies intelligent identification method of fluvial facies tight sandstone reservoir provided in the embodiment of the present invention, the process of training the intelligent identification model includes:

[0034] Adjusting the number of hidden layer neurons, the number of hidden layer layers, and the number of cycles of the radial basis neural network and the multilayer perceptron neural network;

[0035] Adjust the initial boundary value of the facies characteristic in the boundary value module.

[0036] Preferably, in the above-mentioned complex lithofacies intelligent identification method of fluvial facies tight sandstone reservoir provided in the embodiment of the present invention, before inputting the lithofacies logging data and lithofacies characteristics to be identified into the trained intelligent identification model, it also includes:

[0037] Create a module for establishing the lithofacies layer to be identified;

[0038] The well logging curve data of the lithofacies layer to be identified is input into the lithofacies layer establishment module to obtain the well logging data and lithofacies characteristics of the lithofacies layer to be identified; the lithofacies characteristics include lithofacies density and lithofacies frequency.

[0039] The embodiment of the present invention also provides an intelligent identification system for complex lithofacies of fluvial facies tight sandstone reservoirs, comprising:

[0040] A database building module is used to build a lithofacies identification label library;

[0041] A limit value determination module is used to analyze the lithofacies characteristics and their logging response characteristics based on the lithofacies identification label library to determine the initial limit value of the lithofacies characteristics;

[0042] A model building module is used to build an intelligent recognition model; the intelligent recognition model includes a boundary value module, a radial basis neural network and a multi-layer perceptron neural network; the boundary value module is used to receive the initial boundary value of the lithofacies feature, and divide the lithofacies recognition label library into a first sample data set and a second sample data set according to the initial boundary value of the lithofacies feature;

[0043] A model training module, used for training the intelligent recognition model, inputting the first sample data set into the radial basis neural network for training, and inputting the second sample data set into the multilayer perceptron neural network for training;

[0044] The model reasoning module is used to input the lithofacies logging data and lithofacies characteristics to be identified into the trained intelligent identification model to obtain a complex lithofacies identification result map.

[0045] It can be seen from the above technical scheme that the present invention provides a complex lithofacies intelligent identification method for river-facies tight sandstone reservoirs, including: establishing a lithofacies identification label library; analyzing lithofacies characteristics and their logging response characteristics based on the lithofacies identification label library to determine the initial boundary value of the lithofacies characteristics; constructing an intelligent identification model; the intelligent identification model includes a boundary value module, a radial basis neural network and a multi-layer perceptron neural network; the boundary value module is used to receive the initial boundary value of the lithofacies characteristics, and divide the lithofacies identification label library into a first sample data set and a second sample data set according to the initial boundary value of the lithofacies characteristics; training the intelligent identification model, inputting the first sample data set into the radial basis neural network for training, and at the same time inputting the second sample data set into the multi-layer perceptron neural network for training; inputting the lithofacies logging data and lithofacies characteristics to be identified into the trained intelligent identification model to obtain a complex lithofacies identification result map.

[0046] Through the above identification method provided by the present invention, the complex lithofacies of river-facies tight sandstone reservoirs can be effectively identified, with high identification accuracy and strong applicability, providing an effective way for fine anatomical research on reservoir structure and heterogeneity characteristics, which is of great significance for the efficient development of river-facies tight sandstone reservoirs. In addition, the present invention also provides a corresponding system for the intelligent identification method of complex lithofacies of river-facies tight sandstone reservoirs, further making the above method more practical, and the system has corresponding advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0048] Figure 1 A flow chart of a method for intelligently identifying complex lithofacies in fluvial facies tight sandstone reservoirs provided in an embodiment of the present invention;

[0049] Figure 2 A schematic diagram of the structure of a radial basis neural network provided by an embodiment of the present invention;

[0050] Figure 3 A schematic diagram of the structure of a multilayer perceptron neural network provided by an embodiment of the present invention;

[0051] Figure 4 A spider diagram of the sensitivity of different lithofacies logging responses provided by an embodiment of the present invention;

[0052] Figure 5 A statistical diagram of the lithofacies density of the Quhe facies reservoir of the target layer provided by the embodiment of the present invention;

[0053] Figure 6 A statistical diagram of the lithofacies density of the braided river facies reservoir of the target layer provided by an embodiment of the present invention;

[0054] Figure 7 A lithology frequency statistical diagram of the Quhe facies reservoir of the target layer provided by the embodiment of the present invention;

[0055] Figure 8 A lithology frequency statistical diagram of a braided river facies reservoir of a target layer provided by an embodiment of the present invention;

[0056] Figure 9a to Figure 9h They are respectively the frequency distribution histograms of various lithofacies logging data of the Quhe facies reservoir of the target layer provided by the embodiments of the present invention;

[0057] Figure 10a to Figure 10h They are respectively the frequency distribution histograms of various lithofacies logging data of the braided river facies reservoir of the target layer provided by the embodiments of the present invention;

[0058] Fig.11 A result map of the lithofacies layer to be identified in the fluvial facies tight sandstone reservoir provided in an embodiment of the present invention;

[0059] Fig.12 A result diagram of complex lithofacies identification of fluvial facies tight sandstone reservoirs provided in an embodiment of the present invention;

[0060] Fig.13 A schematic diagram of the structure of a complex lithofacies intelligent identification system for fluvial facies tight sandstone reservoirs provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] The present invention provides a method for intelligently identifying complex lithofacies in river-facies tight sandstone reservoirs. Figure 1 As shown, the following steps are included:

[0063] S101, establishing a lithofacies identification label library;

[0064] S102, analyzing the lithofacies characteristics and their logging response characteristics based on the lithofacies identification label library to determine the initial limit value of the lithofacies characteristics;

[0065] In practical applications, the lithofacies characteristics include lithofacies density and lithofacies frequency; the initial limit values ​​of lithofacies characteristics include lithofacies density limit values ​​and lithofacies frequency limit values.

[0066] S103, constructing an intelligent recognition model; the intelligent recognition model includes a boundary value module, a radial basis neural network and a multilayer perceptron neural network; the boundary value module is used to receive an initial boundary value of a lithofacies feature, and divide the lithofacies recognition label library into a first sample data set and a second sample data set according to the initial boundary value of the lithofacies feature;

[0067] It should be noted that if Figure 2 As shown in the figure, the radial basis neural network (RBFNN) is a three-layer neural network based on the local approximation of the Gaussian kernel, which includes an input layer, a hidden layer, and an output layer. The basic idea of ​​RBFNN is to use the radial basis as the "basis" of the hidden unit to form the hidden layer space. The hidden layer transforms the input vector and transforms the low-dimensional pattern input data into the high-dimensional space, so that the linear inseparable problem in the low-dimensional space can be linearly separable in the high-dimensional space. Its local approximation based on the Gaussian kernel makes its network structure simple and its learning ability fast, but it also makes it have a strong learning ability for samples with concentrated data distribution and a high prediction accuracy. On the contrary, if the data distribution of the samples is relatively scattered, the learning rate and prediction accuracy of RBFNN will decrease.

[0068] like Figure 3 As shown in the figure, a multilayer perceptron neural network (MLPNN) is a globally approximated neural network composed of a group of parallel multiple perceptrons, which contains multiple layers of nodes, and each layer of nodes is fully connected to the nodes of the next layer of the network. The basic idea of ​​MLPNN is based on global approximation, using a single-layer perceptron to solve linear separable problems, and constructing a multilayer perceptron to combine linear boundaries into nonlinear boundaries to solve nonlinear separable problems, which has a high degree of nonlinear global effect. Based on the characteristics of multiple linear boundaries forming nonlinear boundaries and global approximation, it has a higher learning rate and prediction accuracy when facing samples with scattered data distribution. On the contrary, when the sample data distribution is relatively concentrated, it is easy to fall into local extreme values, resulting in a decrease in learning rate and prediction accuracy.

[0069] By combining the above-mentioned boundary value module, RBFNN and MLPNN, the intelligent recognition model can be built.

[0070] S104, training the intelligent recognition model, inputting the first sample data set into a radial basis neural network for training, and inputting the second sample data set into a multilayer perceptron neural network for training;

[0071] S105, inputting the lithofacies logging data and lithofacies characteristics to be identified into the trained intelligent identification model to obtain a complex lithofacies identification result map.

[0072] In the above-mentioned intelligent identification method of complex lithofacies of fluvial tight sandstone reservoirs provided in an embodiment of the present invention, the identification of complex lithofacies of fluvial tight sandstone reservoirs can be effectively realized, with high identification accuracy and strong generalizability, providing an effective way for the detailed anatomical study of reservoir structure and heterogeneity characteristics, which is of great significance to the efficient development of fluvial tight sandstone reservoirs.

[0073] Further, in specific implementation, in the above-mentioned intelligent identification method for complex lithofacies in river-facies tight sandstone reservoirs provided in an embodiment of the present invention, step S101 establishes a lithofacies identification label library, which may specifically include: first, observing and describing the core of the target layer section of the coring well, clarifying the lithology and bedding characteristics of the core, identifying the lithofacies type, and assigning a specific symbol to each lithofacies type; then, extracting the maximum and minimum values ​​of each logging curve in the target layer section, and then normalizing each logging curve in accordance with the min-max normalization method, so that the response values ​​contained in each logging curve are mapped to between [0,1]; thereafter, for each identified lithofacies type, reading the normalized logging response values ​​of all logging curves, and drawing a spider diagram; according to the spider diagram, analyzing the sensitivity of different logging curves to different lithofacies types, and only establishing a lithofacies identification label library.

[0074] In specific implementation, drawing a spider web diagram in the above steps may specifically include: drawing a point, and drawing multiple straight lines outward from the point as a starting point, each straight line representing a well logging curve; for any lithofacies type, marking the well logging response value point after normalization of the well logging curve corresponding to the lithofacies type on each straight line; and connecting the points of the lithofacies type on each straight line with straight lines to draw a spider web diagram.

[0075] In specific implementation, the above steps analyze the sensitivity of different logging curves to different lithofacies types based on the spider web diagram, which may specifically include: on any straight line of the spider web diagram, if the numerical points corresponding to different lithofacies types on the straight line are closer, the logging curve represented by the straight line is less sensitive to distinguishing different lithofacies types; if the numerical points corresponding to different lithofacies types on the straight line are farther away, the logging curve represented by the straight line is more sensitive to distinguishing different lithofacies types.

[0076] In specific implementation, in the above-mentioned complex lithofacies intelligent identification method of river-facies tight sandstone reservoirs provided in the embodiment of the present invention, step S102 analyzes the lithofacies characteristics and their logging response characteristics based on the lithofacies identification label library to determine the initial limit value of the lithofacies characteristics, which can specifically include: for lithofacies density, first calculate the lithofacies density of the lithofacies unit in the lithofacies identification label library, and the formula for calculating the lithofacies density is: (single lithofacies thickness) / (lithofacies combination thickness), then calculate the mean of the density of the same lithofacies, and finally draw the average lithofacies density distribution map of various lithofacies of different river facies. Next, for lithofacies frequency, first calculate the lithofacies frequency of the lithofacies unit in the lithofacies identification label library, and the formula for calculating the lithofacies frequency is: the number of occurrences of a certain type of lithofacies within a single lithofacies combination, then calculate the mean of the frequency of the same lithofacies, and finally draw the average lithofacies frequency distribution map of various lithofacies of different river facies. Next, we analyze the characteristics of lithofacies logging response. First, we analyze the influence of surrounding rocks on the logging response of various lithofacies in different river phases, and then draw the frequency distribution histogram of different logging data of various lithofacies in the lithofacies identification label library. Finally, based on the average lithofacies density distribution map, average lithofacies frequency distribution map and different logging data frequency distribution histograms of the above-mentioned lithofacies, we analyze the distribution of lithofacies characteristics, the influence of surrounding rocks on logging response and the distribution of logging data. The comprehensive analysis results give the lithofacies density limit value and lithofacies frequency limit value as the initial limit value of lithofacies characteristics.

[0077] It should be noted that, according to the sedimentary characteristics of fluvial reservoirs, the thickness of a single lithofacies in meandering river sedimentary reservoirs is relatively small, and lithofacies change frequently, that is, most lithofacies types are characterized by low lithofacies density and high lithofacies frequency; the thickness of a single lithofacies in braided river sedimentary reservoirs is relatively large, and lithofacies change infrequently, that is, most lithofacies are characterized by high lithofacies density and low lithofacies frequency.

[0078] In the process of analyzing the logging response characteristics, different logging data response characteristics of each lithofacies label in the lithofacies identification label library are statistically analyzed for different river facies types based on lithofacies type, and frequency distribution histograms are made. For different lithofacies units, due to the differences in lithofacies density and frequency, the degree to which their logging responses are affected by the surrounding rocks is different. When the lithofacies density is large and the lithofacies frequency is low, the surrounding rock has a relatively small impact, and the logging data distribution is relatively concentrated; when the lithofacies density is small and the lithofacies frequency is high, the surrounding rock has a relatively large impact, and the logging data distribution is relatively scattered. A comprehensive analysis of the lithofacies distribution law and the logging response characteristics under its constraints is conducted, and the initial limit values ​​of the lithofacies characteristics (lithofacies density and lithofacies frequency) with different logging data distribution characteristics are given. Different study areas and different target layers may have different lithofacies distribution characteristics and logging response characteristics, and thus different initial limit values ​​of lithofacies characteristics.

[0079] In specific implementation, in the above-mentioned intelligent identification method of complex lithofacies in river-facies tight sandstone reservoirs provided in an embodiment of the present invention, the first sample data set in step S103 includes sample data whose lithofacies density is greater than the lithofacies density limit value and whose lithofacies frequency is less than the lithofacies frequency limit value; the second sample data set includes sample data whose lithofacies density is less than the lithofacies density limit value and whose lithofacies frequency is greater than the lithofacies frequency limit value.

[0080] It can be understood that for samples with lithofacies density greater than the lithofacies density limit value and frequency less than the lithofacies frequency limit value, the logging response is relatively less affected by the surrounding rock, and the logging data distribution is relatively concentrated, so it is used as the training data of RNFNN; for samples with lithofacies density less than the lithofacies density limit value and frequency greater than the lithofacies frequency limit value, the logging response is relatively more affected by the surrounding rock, and the logging data distribution is relatively discrete, so it is used as the training data of MLPNN. Afterwards, according to the lithofacies density and frequency of the lithofacies to be identified, it can be determined which neural network should be selected for identification, which is convenient for improving the identification accuracy.

[0081] In specific implementation, in the above-mentioned intelligent identification method of complex lithofacies in river-facies tight sandstone reservoirs provided in an embodiment of the present invention, the hidden layer activation function of RBFNN is a Softmax type function, the output layer activation function is the identity, and the error function is the sum of squares; the hidden layer activation function of MLPNN is an S-type function, the output layer activation function is a Softmax type function, and the error function is cross entropy.

[0082] In specific implementation, in the above-mentioned complex lithofacies intelligent identification method of fluvial facies tight sandstone reservoir provided in the embodiment of the present invention, in the process of executing step S104 of training the intelligent identification model, the following steps may be specifically included:

[0083] The first step is to input the boundary value module according to the preliminary lithofacies density and frequency boundary values, and then input all the data in the lithofacies identification label library into the constructed intelligent recognition model. At this time, the intelligent recognition model will divide the database into two parts according to the input lithofacies density and frequency boundary values. In each part, part of the data (such as 70% of the data) is used for training, and part of the data (such as 30% of the data) is used for testing the results.

[0084] In the second step, if the training and testing results show that the two models have low accuracy in identifying various types of lithofacies, the number of hidden layer neurons, the number of hidden layer layers, the number of cycles and other parameters of the RBFNN and MLPNN in the intelligent recognition model are adjusted until the recognition accuracy of the RBFNN and MLPNN in the intelligent recognition model for the test data reaches the set threshold (such as 90%).

[0085] Step 3: If after repeated adjustments, the accuracy of a certain type of neural network in the intelligent recognition model still cannot reach the set threshold (such as 90%) for certain lithofacies, adjust the value in the boundary value module (that is, adjust the lithofacies density and frequency boundary values), and repeat the first to third steps. Until the lithofacies recognition accuracy of each type of lithofacies on the test set is greater than the set threshold (such as 90%), the model training is considered complete.

[0086] Step 4: Input the final input lithofacies density and frequency limit values ​​into the lithofacies identification threshold module, that is, use the final lithofacies density and frequency limit values ​​as the thresholds in the lithofacies identification threshold module in the intelligent identification model. In other words, the intelligent identification model can also include a lithofacies identification threshold module, which is used to save the final lithofacies density and frequency limit values. At this point, the intelligent identification model is built and trained.

[0087] In the specific implementation, in the above-mentioned intelligent identification method of complex lithofacies in river-facies tight sandstone reservoirs provided in the embodiment of the present invention, before executing step S105 to input the logging data and lithofacies characteristics of the lithofacies to be identified into the trained intelligent identification model, it can also include: creating a module for establishing the lithofacies layer to be identified, which is mainly based on the find_peaks function in the scipy library in python. Specifically, first use the find peaks function to find the extreme points of the GR logging curve, then average the depth of the two adjacent extreme points as the half-width point depth, and finally set the layer between the two half-width points as the lithofacies to be identified and extract the corresponding logging data. In addition, a formula for automatically calculating the lithofacies density and frequency of the lithofacies layer to be identified is added to the module for establishing the lithofacies layer to be identified. By inputting the logging curve data of the lithofacies layer to be identified into the module for establishing the lithofacies layer to be identified, the logging data and lithofacies characteristics (including lithofacies density and lithofacies frequency) of the lithofacies layer to be identified can be obtained. Specifically, the thickness of the lithofacies layer to be identified is automatically divided by the thickness of the already determined small layer (the same scale as the lithofacies combination mentioned above, which has been artificially divided), and the result is the lithofacies density of the lithofacies layer to be identified; the layers with similar logging responses corresponding to the lithofacies layer to be identified are automatically regarded as the same type of layers (such as category 1, category 2, category 3, etc.), and the number of the same type of layers within the already determined small layer (the same scale as the lithofacies combination mentioned above, which has been artificially divided) is automatically counted, which is the frequency of the lithofacies layer to be identified.

[0088] Furthermore, in the specific implementation, in the process of executing step S105, it can specifically include: inputting the lithofacies logging data, lithofacies density and lithofacies frequency to be identified into the intelligent identification model, and the lithofacies identification threshold module in the intelligent identification model will first automatically determine whether the lithofacies density and lithofacies frequency of the lithofacies to be identified are greater than (less than) the given lithofacies density and lithofacies frequency thresholds (i.e., saving the final lithofacies density and frequency limit values). If the lithofacies density is greater than the threshold and the lithofacies frequency is less than the threshold, the RBFNN model is automatically selected for identification; if the lithofacies density is less than the threshold and the lithofacies frequency is greater than the threshold, the MLPNN model is automatically selected for identification; thereby realizing the complex lithofacies identification of river facies tight sandstone reservoirs.

[0089] The following is an example of a typical river-facies deposited tight sandstone reservoir in a gas field to illustrate the complex lithofacies intelligent identification method of the river-facies tight sandstone reservoir provided by the embodiment of the present invention:

[0090] Step 1: Observe and describe the core of the target layer of the coring well, describe the lithology and bedding characteristics of the core, and determine a total of 8 lithofacies types. With reference to Mali's lithofacies division and naming scheme, each lithofacies type is given a specific symbol, namely, massive bedding conglomerate facies (GM), trough cross-bedding coarse sandstone facies (CSt), plate-like cross-bedding coarse sandstone facies (CSp), plate-like cross-bedding medium sandstone facies (MSp), parallel bedding medium sandstone facies (MSm), cross-bedding fine sandstone facies (Fc), wavy bedding siltstone facies (Sw), and massive bedding mudstone facies (Mm).

[0091] Step 2: Extract the maximum and minimum values ​​of each logging curve in the target layer, and then normalize each logging curve according to the min-max standardization method, so that the response values ​​contained in each logging curve are mapped to the range of [0-1].

[0092] Step 3: For each lithofacies type identified on the core, read the normalized response values ​​of all logging curves, and analyze the sensitivity of different logging curves to lithofacies based on the spider diagram. Figure 4 As shown, six types of logging data are selected, including normalized natural gamma curve (GRGYH), deep resistivity curve (LLDGYH), acoustic time difference curve (ACGYH), neutron curve (PHINGYH), density curve (DENGYH), and photoelectric absorption cross-section index (PEGYH).

[0093] Step 4: Establish a lithofacies identification label library.

[0094] Step 5: Analyze the lithofacies distribution characteristics and the logging response characteristics under lithofacies constraints.

[0095] like Figures 5 to 8As shown in the figure, it can be seen that there are certain differences in the lithofacies density and lithofacies frequency of the two types of river facies reservoirs. The overall lithofacies density of the braided river facies reservoir in the target layer is relatively large. Except for the conglomerate facies, wavy bedding siltstone facies, and mudstone facies, the densities of the remaining lithofacies are all greater than 25%; the lithofacies frequency is relatively low. Except for the mudstone facies, the frequencies of the remaining lithofacies are all less than 1.3. The overall lithofacies density of the meandering river facies reservoir in the target layer is relatively small. The densities of all lithofacies are less than 25%; the lithofacies frequency is relatively high. Except for the conglomerate facies, plate-like cross-bedding coarse sandstone facies, and trough-like cross-bedding coarse sandstone facies, the frequencies of the remaining lithofacies are all greater than 1.3.

[0096] For different lithofacies units, the degree to which their logging responses are affected by the surrounding rocks varies due to differences in their thickness and frequency of change. Figure 9a to Figure 9h As shown in the figure, the thickness of a single lithofacies in the meandering river reservoir is small, the lithofacies change frequently, each lithofacies is greatly affected by the adjacent lithofacies, and the distribution of logging data of each lithofacies is relatively scattered. Figure 10a to Figure 10h As shown, the single lithofacies of the braided river reservoir is thick, the lithofacies change is not frequent, each lithofacies is less affected by the adjacent lithofacies, and the distribution of logging data of each lithofacies is relatively concentrated.

[0097] Based on the comprehensive lithofacies distribution characteristics and logging response characteristics under their constraints, the initial limit values ​​of lithofacies characteristics are determined to be 25% lithofacies density and 1.3 lithofacies frequency.

[0098] Step 6. Calculate the lithofacies density and lithofacies frequency of all labels in the lithofacies identification label library, and add the lithofacies density and lithofacies frequency of the labels to the lithofacies identification label library. Then, input the lithofacies density of 25% and the lithofacies frequency of 1.3 into the boundary value module, and input all the label data into the intelligent recognition model. At this time, the intelligent recognition model starts to automatically divide the label data and starts training and testing the model.

[0099] Table 1 Model training results

[0100]

[0101]

[0102] According to the training and test results of the intelligent recognition model in Table 1, when the lithofacies density limit value is set to 25% and the lithofacies frequency limit value is set to 1.3, the accuracy of each lithofacies test reaches 90%, and the model training is completed. At this time, the lithofacies density of 25% and the lithofacies frequency of 1.3 are used as thresholds to input the limit value module.

[0103] Step 7: Input the well logging curve data and the top and bottom depth data of the single layer into the lithofacies layer establishment module to be identified, and you can get Fig.11 The single well division diagram of the lithofacies layer to be identified is shown in FIG. 1 , and the lithofacies density, lithofacies frequency and logging data of the lithofacies layer to be identified are shown in Table 2 .

[0104] Table 2 Lithofacies density, lithofacies frequency and well logging data of the lithofacies layer to be identified in a well

[0105] Layer Category Lithofacies density Lithofacies frequency GRGHY LLDGHY ACG …… Lithofacies layer to be identified Category 1 14.30％ 1.86 0.83 0.23 0.53 …… Lithofacies layer to be identified Category 2 20.63％ 1.53 0.72 0.41 0.32 …… Lithofacies layer to be identified Category 6 30.33％ 1.19 0.42 0.83 0.73 …… Lithofacies layer to be identified Category 4 28.11％ 1.24 0.53 0.59 0.52 …… Lithofacies layer to be identified Category 5 34.30％ 1.12 0.31 0.73 0.63 …… Lithofacies layer to be identified Category 3 21.62％ 1.5 0.63 0.49 0.52 …… Lithofacies layer to be identified Category 4 27.93％ 1.26 0.55 0.62 0.59 …… Lithofacies layer to be identified Category 7 18.01％ 0.91 0.23 0.36 0.87 …… Lithofacies layer to be identified Category 3 22.31％ 1.6 0.67 0.45 0.49 …… Lithofacies layer to be identified Category 1 12.40％ 1.79 0.78 0.36 0.63 …… Lithofacies layer to be identified Category 4 28.66％ 1.21 0.57 0.63 0.60 …… Lithofacies layer to be identified Category 6 32.65％ 1.18 0.46 0.74 0.78 …… Lithofacies layer to be identified Category 3 20.93％ 1.63 0.65 0.42 0.57 …… Lithofacies layer to be identified Category 8 15.93％ 0.82 0.12 0.12 0.93 …… …… …… …… …… …… …… …… ……

[0106] Step 8: Input the well logging data, lithofacies density and lithofacies frequency to be identified into the trained intelligent recognition model to obtain Fig.12 The figure shows the results of complex lithofacies identification.

[0107] Based on the same inventive concept, an embodiment of the present invention also provides an intelligent identification system for complex lithofacies in river-facies tight sandstone reservoirs. Since the principle of solving the problem by this system is similar to the aforementioned intelligent identification method for complex lithofacies in river-facies tight sandstone reservoirs, the implementation of this system can refer to the implementation of the intelligent identification method for complex lithofacies in river-facies tight sandstone reservoirs, and the repeated parts will not be repeated.

[0108] In specific implementation, the complex lithofacies intelligent identification system of river-facies tight sandstone reservoirs provided by the embodiment of the present invention is as follows: Fig.13 As shown, specifically including:

[0109] A database establishment module 11 is used to establish a lithofacies identification label library;

[0110] The limit value determination module 12 is used to analyze the lithofacies characteristics and their logging response characteristics based on the lithofacies identification label library to determine the initial limit value of the lithofacies characteristics;

[0111] The model building module 13 is used to build an intelligent recognition model; the intelligent recognition model includes a boundary value module, a radial basis neural network and a multi-layer perceptron neural network; the boundary value module is used to receive the initial boundary value of the lithofacies feature, and divide the lithofacies recognition label library into a first sample data set and a second sample data set according to the initial boundary value of the lithofacies feature;

[0112] A model training module 14 is used to train the intelligent recognition model, input the first sample data set into a radial basis neural network for training, and input the second sample data set into a multi-layer perceptron neural network for training;

[0113] The model reasoning module 15 is used to input the lithofacies logging data and lithofacies characteristics to be identified into the trained intelligent identification model to obtain a complex lithofacies identification result map.

[0114] In the above-mentioned intelligent identification system for complex lithofacies of fluvial tight sandstone reservoirs provided in the embodiment of the present invention, the identification of complex lithofacies of fluvial tight sandstone reservoirs can be effectively realized through the interaction of the above-mentioned five modules, with high identification accuracy and strong generalizability, providing an effective way for the detailed anatomical study of reservoir structure and heterogeneity characteristics, which is of great significance to the efficient development of fluvial tight sandstone reservoirs.

[0115] Furthermore, in a specific implementation, the complex lithofacies intelligent identification system of the fluvial facies tight sandstone reservoir provided in the embodiment of the present invention may also include:

[0116] The module for establishing the lithofacies layer to be identified is used to receive the logging curve data of the lithofacies layer to be identified, and obtain the logging data and lithofacies characteristics of the lithofacies layer to be identified through calculation; the lithofacies characteristics include lithofacies density and lithofacies frequency.

[0117] For more specific working processes of the above modules, please refer to the corresponding contents disclosed in the above embodiments, which will not be repeated here.

[0118] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the systems, devices, and storage media disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method part.

[0119] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0120] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0121] In summary, an intelligent identification method for complex lithofacies in river-facies tight sandstone reservoirs provided by an embodiment of the present invention includes: establishing a lithofacies identification label library; analyzing lithofacies characteristics and their logging response characteristics based on the lithofacies identification label library to determine the initial limit value of the lithofacies characteristics; constructing an intelligent identification model; the intelligent identification model includes a limit value module, a radial basis neural network and a multilayer perceptron neural network; the limit value module is used to receive the initial limit value of the lithofacies characteristics, and divide the lithofacies identification label library into a first sample data set and a second sample data set according to the initial limit value of the lithofacies characteristics; training the intelligent identification model, inputting the first sample data set into the radial basis neural network for training, and inputting the second sample data set into the multilayer perceptron neural network for training; inputting the lithofacies logging data and lithofacies characteristics to be identified into the trained intelligent identification model to obtain a complex lithofacies identification result map. The above method can effectively realize the identification of complex lithofacies in river-facies tight sandstone reservoirs, with high identification accuracy and strong generalizability, providing an effective way for fine anatomical research on reservoir structure and heterogeneity characteristics, and has great significance for the efficient development of river-facies tight sandstone reservoirs. In addition, the present invention also provides a corresponding system for the intelligent identification method of complex lithofacies in river-facies tight sandstone reservoirs, further making the above method more practical, and the system has corresponding advantages.

[0122] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0123] The above is a detailed introduction to the intelligent identification method and system for complex lithofacies in river-facies tight sandstone reservoirs provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technicians in this field, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. An intelligent identification method for complex lithofacies in river-facies tight sandstone reservoirs, characterized in that: include: Establish a petrographic identification label library; Analyze the lithofacies characteristics and their logging response characteristics based on the lithofacies identification label library to determine the initial limit value of the lithofacies characteristics; Constructing an intelligent recognition model; the intelligent recognition model includes a boundary value module, a radial basis neural network and a multilayer perceptron neural network; the boundary value module is used to receive the initial boundary value of the lithofacies feature, and divide the lithofacies identification label library into a first sample data set and a second sample data set according to the initial boundary value of the lithofacies feature; the first sample data set includes sample data whose lithofacies density is greater than the lithofacies density boundary value and whose lithofacies frequency is less than the lithofacies frequency boundary value; the second sample data set includes sample data whose lithofacies density is less than the lithofacies density boundary value and whose lithofacies frequency is greater than the lithofacies frequency boundary value; Training the intelligent recognition model, inputting the first sample data set into the radial basis neural network for training, and inputting the second sample data set into the multilayer perceptron neural network for training; The lithofacies logging data and lithofacies characteristics to be identified are input into the trained intelligent identification model to obtain a complex lithofacies identification result map.

2. The complex lithofacies intelligent identification method of fluvial facies tight sandstone reservoir according to claim 1 is characterized in that: Establish a petrographic identification label library, including: Observe and describe the core of the target layer of the coring well, identify the lithofacies type, and assign specific symbols to each lithofacies type; Extract the maximum and minimum values ​​of each logging curve in the target layer, and normalize each logging curve in a min-max standardization manner to map the response values ​​contained in each logging curve to between 0 and 1; For each identified lithofacies type, read the normalized logging response values ​​of all logging curves and draw a spider diagram; According to the spider web diagram, the sensitivity of different logging curves to different lithofacies types is analyzed, and only a lithofacies identification label library is established.

3. The complex lithofacies intelligent identification method of fluvial tight sandstone reservoir according to claim 2 is characterized in that: Draw a spider web diagram, including: Draw multiple straight lines outward from a point, each straight line represents a well logging curve; For any lithofacies type, mark the logging response value point after normalization of the logging curve corresponding to the lithofacies type on each straight line; Connect the points of the lithofacies type on each straight line with straight lines to draw a spider web diagram.

4. The complex lithofacies intelligent identification method of fluvial facies tight sandstone reservoir according to claim 3 is characterized in that: According to the spider diagram, the sensitivity of different logging curves to different lithofacies types is analyzed, including: On any straight line of the spider web diagram, if the numerical points corresponding to different lithofacies types on the straight line are closer, the logging curve represented by the straight line is less sensitive to distinguishing different lithofacies types; if the numerical points corresponding to different lithofacies types on the straight line are farther away, the logging curve represented by the straight line is more sensitive to distinguishing different lithofacies types.

5. The complex lithofacies intelligent identification method of fluvial tight sandstone reservoir according to claim 4 is characterized in that: The lithofacies characteristics and their logging response characteristics are analyzed based on the lithofacies identification tag library to determine the initial limit value of the lithofacies characteristics, including: Calculating the lithofacies density and lithofacies frequency of the lithofacies units in the lithofacies identification tag library; Calculate the mean density and frequency of similar lithofacies; Draw the average lithofacies density distribution map and average lithofacies frequency distribution map of various lithofacies in different river phases; Analyze the different logging response characteristics of various lithofacies in different fluvial facies and draw frequency distribution histograms of different logging data; According to the average lithofacies density distribution diagram, average lithofacies frequency distribution diagram and different logging data frequency distribution histograms, the distribution of lithofacies characteristics, the influence of surrounding rocks on logging response and the distribution of logging data are analyzed. The lithofacies density limit value and lithofacies frequency limit value are obtained as the initial limit value of lithofacies characteristics based on the comprehensive analysis results.

6. The complex lithofacies intelligent identification method of fluvial tight sandstone reservoir according to claim 1 is characterized in that: The hidden layer activation function of the radial basis neural network is a Softmax type function, the output layer activation function is an identity, and the error function is a sum of squares; The hidden layer activation function of the multilayer perceptron neural network is a S-type function, the output layer activation function is a Softmax function, and the error function is cross entropy.

7. The complex lithofacies intelligent identification method of fluvial tight sandstone reservoirs according to claim 1 is characterized in that: The process of training the intelligent recognition model includes: Adjusting the number of hidden layer neurons, the number of hidden layer layers, and the number of cycles of the radial basis neural network and the multilayer perceptron neural network; Adjust the initial boundary value of the facies characteristic in the boundary value module.

8. The complex lithofacies intelligent identification method of fluvial tight sandstone reservoirs according to claim 1 is characterized in that: Before inputting the lithofacies logging data and lithofacies characteristics to be identified into the trained intelligent identification model, the method further includes: Create a module for establishing the lithofacies layer to be identified; The well logging curve data of the lithofacies layer to be identified is input into the lithofacies layer establishment module to obtain the well logging data and lithofacies characteristics of the lithofacies layer to be identified; the lithofacies characteristics include lithofacies density and lithofacies frequency.

9. An intelligent identification system for complex lithofacies of river-facies tight sandstone reservoirs, characterized in that: include: A database building module is used to build a lithofacies identification label library; A limit value determination module is used to analyze the lithofacies characteristics and their logging response characteristics based on the lithofacies identification label library to determine the initial limit value of the lithofacies characteristics; A model building module is used to build an intelligent recognition model; the intelligent recognition model includes a boundary value module, a radial basis neural network and a multilayer perceptron neural network; the boundary value module is used to receive the initial boundary value of the lithofacies feature, and divide the lithofacies identification label library into a first sample data set and a second sample data set according to the initial boundary value of the lithofacies feature; the first sample data set includes sample data whose lithofacies density is greater than the lithofacies density boundary value and whose lithofacies frequency is less than the lithofacies frequency boundary value; the second sample data set includes sample data whose lithofacies density is less than the lithofacies density boundary value and whose lithofacies frequency is greater than the lithofacies frequency boundary value; A model training module, used for training the intelligent recognition model, inputting the first sample data set into the radial basis neural network for training, and inputting the second sample data set into the multilayer perceptron neural network for training; The model reasoning module is used to input the lithofacies logging data and lithofacies characteristics to be identified into the trained intelligent identification model to obtain a complex lithofacies identification result map.

Images