Total organic carbon logging interpretation method and system based on shale heterogeneity characteristics

Abstract

The embodiment of the present invention provides a total organic carbon logging interpretation method and system based on the heterogeneity characteristics of shale, including: obtaining relevant data of shale layer sections; obtaining the distribution of homogeneous unit types with depth in each single well coring section according to the core description results and core sample test data in the modeling data; for each homogeneous unit type, using a machine learning algorithm to establish a homogeneous unit logging identification model; using a machine learning algorithm to establish a respective TOC logging interpretation model for each homogeneous unit type; using the drilling data set to be predicted, using the homogeneous unit logging identification model to identify the homogeneous unit types of different depth sections of a single well, and conducting TOC logging interpretation of shale layer sections according to the TOC logging interpretation model of each homogeneous unit. By constructing a rock phase type in which other rock components are relatively homogeneous except TOC, TOC logging interpretation models are established for different rock types, so that TOC interpretation is more accurate.

Description

Technical Field

[0001] The invention relates to well logging evaluation of key parameters of shale reservoirs, and in particular to a total organic carbon well logging interpretation method and system based on shale heterogeneity characteristics. Background Art

[0002] Organic matter content is an indispensable basic data for source rock evaluation, shale oil and gas resource evaluation, and sweet spot prediction. At present, laboratory core test analysis technology is the most direct and accurate means to obtain the organic matter content of mud shale, among which total organic carbon (TOC) is the most commonly used indicator to characterize organic matter content. Due to the limitation that most drillings have no core data or incomplete coring, TOC interpretation through high-resolution and high-coverage logging data is an important means to evaluate the overall organic matter content of mud shale sections.

[0003] With the deepening of shale oil and gas reservoir exploration and development, a large number of TOC logging interpretation methods, technologies or models have been proposed. These methods can be divided into two categories: model-driven and data-driven. In the early development of TOC logging interpretation technology, model-driven interpretation methods were developed intensively, including classic interpretation methods such as formation density curve method, GR curve method, ΔlogR, etc. In the past two decades, with the development of artificial intelligence technology, data-driven methods have gradually become a research hotspot. TOC logging interpretation technologies based on BP neural network, SVM, naive Bayes and other algorithms have been proposed and widely used.

[0004] Whether it is model-driven or data-driven TOC logging interpretation technology, the purpose is to find a statistical relationship model between logging response value and TOC. Generally, the model-driven interpretation method takes the logging response characteristics of the organic-poor layer as the baseline, and builds a model based on the relationship between the amplitude of the logging response value deviation from the baseline and the TOC content (or supplemented by other parameters). For example, the ΔlogR method uses the water-bearing sandstone section as the baseline, calculates the ΔLogR value by the amplitude of the resistivity and acoustic wave time difference deviation from the baseline, and then establishes a correlation model between this value and TOC based on the maturity of the source rock; the data-driven interpretation method does not rely on the baseline of the organic-poor layer, but after selecting the logging series closely related to TOC, it uses artificial intelligence algorithms to explore the statistical relationship between logging response and TOC, and then uses the above statistical relationship to perform logging interpretation of TOC.

[0005] However, in actual situations, the logging response of the shale section is not only affected by the organic matter content, but also by other reservoir rock structures, including grain size, mineral content (especially the content of heavy minerals and radioactive elements), pore development, pore fluid properties, organic matter occurrence form, etc. These rock fabric characteristics will affect the logging response value together with the organic matter content, making the statistical relationship between TOC and logging response value extremely complex. However, in traditional model-driven and data-driven TOC interpretation methods, the above rock fabric characteristic data are not used to constrain the logging response-TOC relationship modeling. There are many reasons for the above situation, such as the low cost of obtaining TOC data and the high cost of obtaining other rock fabric data, which cannot form a large-scale data set required for modeling; there are some qualitative or semi-quantitative data in the rock fabric information, such as the occurrence form of organic matter, which cannot be used for quantitative modeling.

[0006] In addition, TOC test samples are often sampled at equal intervals or randomly, which inevitably results in fewer samples in certain numerical intervals, making the data set have the characteristics of unbalanced data distribution. Generally, data-driven methods focus on the data of high-distribution frequency samples, and the prediction accuracy of data with a small number of samples is low, resulting in low generalization ability of data-driven methods. Unfortunately, the TOC interval with low data density may be the key interval of shale reservoir research, such as the high-value organic carbon content interval. Summary of the invention

[0007] The embodiment of the present invention provides a total organic carbon logging interpretation method and system based on the heterogeneity characteristics of shale. The invention starts from the heterogeneity of the shale formation itself, constructs a homogeneous unit type in which other rock components except TOC are relatively homogeneous, and establishes TOC logging interpretation models for different homogeneous unit types, thereby avoiding underfitting or overfitting of the TOC interpretation model, enhancing the generalization ability of the interpretation model, making the TOC interpretation more accurate, and facilitating the detailed evaluation of the shale reservoir.

[0008] To achieve the above objectives, on the one hand, an embodiment of the present invention provides a total organic carbon logging interpretation method based on the heterogeneity characteristics of shale, comprising:

[0009] Obtaining shale layer-related data, the shale layer-related data includes two categories: a modeling data set and a drilling data set to be predicted; wherein the modeling data set includes core description results, core sample test data and logging data of the coring section, and the drilling data set to be predicted includes logging data; in the modeling data set, the core sample test data includes characteristic parameters after testing and analyzing the total organic carbon content TOC of mineral petrology and organic petrography, and the characteristic parameters are quantitative data and typified qualitative data; the total organic carbon content is a characterization index of the organic matter content in shale;

[0010] According to the core description results and core sample test data in the modeling data, a division benchmark of homogeneous unit types is established, and the homogeneous unit types of the coring section are divided using the division benchmark of the homogeneous unit types to obtain the distribution of the homogeneous unit types of the coring section of each single well with depth;

[0011] For each homogeneous unit type, a homogeneous unit logging identification model is established using a machine learning algorithm to predict the homogeneous unit type using the logging response of the shale layer.

[0012] A machine learning algorithm was used to establish a TOC logging interpretation model for each homogeneous unit type;

[0013] The drilling data set to be predicted is used to identify the homogeneous unit types at different depths of a single well using the homogeneous unit logging identification model, and the TOC logging interpretation of the shale layer is carried out based on the TOC logging interpretation model of each homogeneous unit.

[0014] On the other hand, an embodiment of the present invention provides a total organic carbon logging interpretation system based on shale heterogeneity characteristics, comprising:

[0015] A data acquisition unit is used to acquire shale layer section related data, wherein the shale layer section related data includes two categories: a modeling data set and a drilling data set to be predicted; wherein the modeling data set includes core description results, core sample test data and logging data of the coring section, and the drilling data set to be predicted includes logging data; in the modeling data set, the core sample test data includes characteristic parameters after testing and analyzing the total organic carbon content TOC of mineral petrology and organic petrography, and the characteristic parameters are quantitative data and typified qualitative data; the total organic carbon content is a characterization index of the organic matter content in shale;

[0016] The homogeneous unit type division unit is used to establish a division basis of the homogeneous unit type according to the core description results and core sample test data in the modeling data, and to divide the homogeneous unit type of the coring section using the division basis of the homogeneous unit type to obtain the distribution of the homogeneous unit type of each single well coring section with depth;

[0017] A homogeneous unit type prediction unit is used to establish a homogeneous unit logging identification model for each homogeneous unit type using a machine learning algorithm, so as to predict the homogeneous unit type using the logging response of the shale layer segment;

[0018] The TOC logging interpretation model establishment unit uses a machine learning algorithm to establish a separate TOC logging interpretation model for each homogeneous unit type;

[0019] The TOC logging interpretation unit is used to use the drilling data set to be predicted and the homogeneous unit logging identification model to identify the homogeneous unit types in different depth sections of a single well, and to carry out TOC logging interpretation of shale layers based on the TOC logging interpretation model of each homogeneous unit.

[0020] The above technical solution has the following beneficial effects: the invention starts from the perspective of the heterogeneity of the shale formation itself, and by constructing a rock phase type in which other rock components except TOC are relatively homogeneous, a TOC logging interpretation model is established for different rock types, thereby avoiding underfitting or overfitting of the TOC interpretation model, enhancing the generalization ability of the interpretation model, making the TOC interpretation more accurate, and facilitating the detailed evaluation of the shale reservoir. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 is a flow chart of a total organic carbon logging interpretation method based on shale heterogeneity characteristics according to an embodiment of the present invention;

[0023] Figure 2 is a structural diagram of a total organic carbon logging interpretation system based on shale heterogeneity characteristics according to an embodiment of the invention;

[0024] Figure 3 are characteristics of different homogeneous unit types within the domain of the invention embodiment;

[0025] Figure 4 The homogeneous unit type, vertical distribution of TOC and its logging response of the YY22 well in the embodiment of the invention;

[0026] Figure 5 It is the construction of the homogeneous unit logging identification model of the YY22 well and its logging response in the embodiment of the invention;

[0027] Figure 6 It is the SVM algorithm hyperparameter optimization process of the embodiment of the invention;

[0028] Figure 7 is a cross plot of measured TOC and predicted TOC of an embodiment of the invention;

[0029] Figure 8 It is the homogeneous unit type and TOC prediction result of the FY3 well in the embodiment of the invention. DETAILED DESCRIPTION

[0030] 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.

[0031] like Figure 1 As shown, in combination with an embodiment of the present invention, a total organic carbon logging interpretation method based on the heterogeneity characteristics of shale is provided, comprising:

[0032] S101: Obtaining shale layer-related data, the shale layer-related data includes two categories: a modeling data set and a drilling data set to be predicted; wherein the modeling data set includes core description results, core sample test data and logging data of the coring section, and the drilling data set to be predicted includes logging data; in the modeling data set, the core sample test data includes characteristic parameters after testing and analyzing the total organic carbon content TOC of mineral petrology and organic petrography, and the characteristic parameters are quantitative data and typified qualitative data; the total organic carbon content is a characterization index of the organic matter content in shale;

[0033] S102: establishing a division criterion for homogeneous unit types according to the core description results in the modeling data and the core sample test data, and dividing the homogeneous unit types of the coring section using the division criterion for homogeneous unit types to obtain the distribution of the homogeneous unit types of the coring section of each single well with depth;

[0034] S103: For each homogeneous unit type, a homogeneous unit logging identification model is established using a machine learning algorithm to predict the homogeneous unit type using the logging response of the shale layer;

[0035] S104: using machine learning algorithms to establish a TOC logging interpretation model for each homogeneous unit type;

[0036] S105: Using the drilling data set to be predicted, the homogeneous unit logging identification model is used to identify the homogeneous unit types of different depth sections of a single well, and the TOC logging interpretation of the shale layer is carried out according to the TOC logging interpretation model of each homogeneous unit.

[0037] Preferably, the core description results of the shale include: lithology, sedimentary structure and sedimentary structure;

[0038] Step 102 specifically includes:

[0039] The lithology, sedimentary structure and sedimentary structure of shale are used as the basic basis for the division of homogeneous units, and the core sample test data is used as a judgment, so that any homogeneous unit to be divided has the following characteristics at the same time:

[0040] Except for TOC, the formation characteristic parameters are relatively homogeneous, that is, other formation characteristic parameters have a relatively small distribution space;

[0041] The thickness of the homogeneous unit should be at least a first thickness, which is greater than the minimum vertical resolution of conventional logging;

[0042] Using any of the homogeneous units as a basis for dividing homogeneous unit types;

[0043] According to the characteristics of any homogeneous unit, the well logging data corresponding to the modeling data set is divided into multiple different homogeneous units.

[0044] Preferably, step 103 specifically includes:

[0045] In the coring section with homogeneous unit type calibration, the homogeneous unit type of each logging response point is obtained from the modeling data set;

[0046] For each homogeneous unit type, a respective well logging response-homogeneous unit type data set A is established;

[0047] All the logging responses and homogeneous unit type data in the logging response-homogeneous unit data set A are taken as the supervision data set, and the well logging identification method F of homogeneous unit type is constructed based on the supervision data set using the data mining algorithm. litho After ensuring that the recognition accuracy of all logging responses and homogeneous unit type data reaches the preset accuracy, the homogeneous unit logging recognition model F is obtained. litho .

[0048] That is, it is equivalent to taking a sample from the coring section, testing and obtaining its total organic carbon content, and then obtaining the mean unit type of this sample based on the previous mean unit distribution, and obtaining the logging response corresponding to this depth based on the depth, thus forming a TOC-mean unit-well logging corresponding data pair, and combining these data pairs to form a TOC-mean unit type-well logging response data set. Among them, the TOC test sample is a sample with depth information collected in the coring section, and then the sample is tested using the TOC test method to obtain the corresponding TOC value.

[0049] Preferably, all the logging responses and homogeneous unit type data in the logging response-homogeneous unit data set A are used as a supervisory data set, and a well logging identification method F of homogeneous unit type is constructed based on the supervisory data set using a data mining algorithm. litho After ensuring that the recognition accuracy of all logging responses and homogeneous unit type data reaches the preset accuracy, the homogeneous unit logging recognition model F is obtained. litho , including:

[0050] The extreme gradient boosting tree XGboost algorithm is used to establish a well logging identification model for homogeneous unit types;

[0051] The logging response-rock phase data of the logging response-mean unit data set A is used as a supervised data set, and the supervised data set is used as the input data of the well logging recognition model of the homogeneous unit type to train the XGboost classification model, and the K-fold crossover operation is used during the training;

[0052] The fitness function of the genetic algorithm is set to the cross-validation accuracy of XGboost; the genetic algorithm is used to optimize the hyperparameter series so that the cross-validation accuracy of XGboost reaches the minimum, ensuring that the recognition accuracy reaches the preset accuracy; specifically including:

[0053] After a population with an initial hyperparameter series is given, the XGboost cross-validation accuracy corresponding to each individual in the population is obtained, and individuals in the initial population are selected according to the elite retention mode. A new generation of population is generated through operations such as crossover and mutation. The above process is iterated repeatedly until the set number of iterations is reached or the fitness function reaches the set threshold. The hyperparameter system that makes the fitness function reach the lowest is selected to ensure that the recognition accuracy reaches the preset accuracy.

[0054] Preferably, step 104 specifically includes:

[0055] The TOC data of the coring section is obtained, and the corresponding mean unit type is obtained by using the depth information of each TOC test sample. The logging response value corresponding to each TOC logging sample is collected to form a TOC-homogeneous unit type-logging response data set B; wherein the TOC data of the coring section comes from multiple TOC test samples, the TOC data of the coring section is the core sample test data, and the logging response value refers to the logging data;

[0056] According to the number m of all mean unit types in all TOC logging samples, the TOC-homogeneous unit type-logging response dataset B is divided into m TOC-logging response datasets B i , where each TOC-logging response data set B iThe samples in the same homogeneous unit type, where i∈[1, m], are all from the same homogeneous unit type; the TOC-logging response dataset B of the target lithology of non-shale oil and gas i Set to an empty set;

[0057] Using data mining algorithms, each TOC-logging response data set B i The data in the well are used as the supervision data set to construct the TOC logging interpretation method F for each homogeneous unit type. TOC，i After ensuring that the recognition accuracy reaches the preset accuracy, the TOC logging interpretation model F corresponding to each homogeneous unit type is obtained. TOC，i .

[0058] Preferably, a data mining algorithm is used to convert each TOC-logging response data set B i The data in the well are used as the supervision data set to construct the TOC logging interpretation method F for each homogeneous unit type. TOC，i After ensuring that the recognition accuracy reaches the preset accuracy, the TOC logging interpretation model F corresponding to each homogeneous unit type is obtained. TOC，i , including:

[0059] The data in each TOC-logging response data set Bi is used as a supervised data set, each supervised data set is used as input data, the SVM regression algorithm is trained, and K-fold cross validation is used in the training of the SVM regression algorithm;

[0060] The fitness function of the genetic algorithm is set to the cross-square error of SVM under cross-validation; the genetic algorithm is used to optimize the hyperparameter series that minimizes the cross-square error, and the hyperparameter series that minimizes the cross-square error is used as the hyperparameter in the SVM regression algorithm to obtain the TOC logging data interpretation model corresponding to various homogeneous unit types;

[0061] The fitness function of the genetic algorithm is set to the cross-square error of the SVM under cross-validation; the genetic algorithm is used to optimize the hyperparameter series that minimizes the cross-square error, specifically including:

[0062] After a population with an initial hyperparameter series is given, the cross-square error corresponding to each individual in the population is obtained, and individuals in the initial population are selected according to the elite retention mode. A new generation of population is generated through operations such as crossover and mutation. The above process is iterated repeatedly until the set number of iterations is reached or the fitness function reaches the set threshold. The hyperparameter series that makes the fitness function reach the lowest is selected, and the recognition accuracy is guaranteed to reach the preset accuracy.

[0063] Preferably, step 105 specifically includes:

[0064] The homogeneous unit logging identification model is used to identify the homogeneous unit type of the predicted drilling data set, and the distribution of the homogeneous unit type of the predicted drilling with depth is obtained;

[0065] For each homogeneous unit type to be predicted for drilling, the respective TOC logging interpretation model is used for TOC interpretation to obtain the TOC interpretation corresponding to each homogeneous unit type, and all TOC interpretations corresponding to the homogeneous unit type are used as the TOC logging interpretation of the shale to be predicted for drilling.

[0066] like Figure 2 As shown, in combination with an embodiment of the present invention, a total organic carbon logging interpretation system based on the heterogeneity characteristics of shale is provided, comprising:

[0067] The data acquisition unit 21 is used to acquire the relevant data of the shale layer, and the relevant data of the shale layer includes two categories: a modeling data set and a drilling data set to be predicted; wherein the modeling data set includes the core description results of the coring section, the core sample test data and the logging data, and the drilling data set to be predicted includes the logging data; in the modeling data set, the core sample test data includes the characteristic parameters after the test and analysis of the total organic carbon content TOC of the mineral petrology and organic petrography, and the characteristic parameters are quantitative data and typified qualitative data; the total organic carbon content is a characterization index of the organic matter content in the shale;

[0068] The homogeneous unit type division unit 22 is used to establish a division basis of the homogeneous unit type according to the core description results and the core sample test data in the modeling data, and divide the homogeneous unit type of the coring section using the division basis of the homogeneous unit type to obtain the distribution of the homogeneous unit type of each single well coring section with depth;

[0069] A homogeneous unit type prediction unit 23 is used to establish a homogeneous unit logging identification model for each homogeneous unit type using a machine learning algorithm, so as to predict the homogeneous unit type using the logging response of the shale layer segment;

[0070] A TOC logging interpretation model building unit 24, which uses a machine learning algorithm to build a respective TOC logging interpretation model for each homogeneous unit type;

[0071] The TOC logging interpretation unit 25 is used to use the drilling data set to be predicted and the homogeneous unit logging identification model to identify the homogeneous unit types of different depth sections of a single well, and to carry out TOC logging interpretation of shale layers based on the TOC logging interpretation model of each homogeneous unit.

[0072] Preferably, the core description results of the shale include: lithology, sedimentary structure and sedimentary structure;

[0073] The homogeneous unit type classification unit 22 is specifically used for:

[0074] The lithology, sedimentary structure and sedimentary structure of shale are used as the basic basis for the division of homogeneous units, and the core sample test data is used as a judgment, so that any homogeneous unit to be divided has the following characteristics at the same time:

[0075] Except for TOC, the formation characteristic parameters are relatively homogeneous, that is, other formation characteristic parameters have a relatively small distribution space;

[0076] The thickness of the homogeneous unit should be at least a first thickness, which is greater than the minimum vertical resolution of conventional logging;

[0077] Using any of the homogeneous units as a basis for dividing homogeneous unit types;

[0078] According to the characteristics of any homogeneous unit, the well logging data corresponding to the modeling data set is divided into multiple different homogeneous units.

[0079] Preferably, the homogeneous unit type prediction unit 23 is specifically used for:

[0080] In the coring section with homogeneous unit type calibration, the homogeneous unit type of each logging response point is obtained from the modeling data set;

[0081] For each homogeneous unit type, a respective well logging response-homogeneous unit type data set A is established;

[0082] All the logging responses and homogeneous unit type data in the logging response-homogeneous unit data set A are taken as the supervision data set, and the well logging identification method F of homogeneous unit type is constructed based on the supervision data set using the data mining algorithm. litho After ensuring that the recognition accuracy of all logging responses and homogeneous unit type data reaches the preset accuracy, the homogeneous unit logging recognition model F is obtained. litho .

[0083] Preferably, the TOC logging interpretation model building unit 24 is specifically used for:

[0084] The TOC data of the coring section is obtained, and the corresponding mean unit type is obtained by using the depth information of each TOC test sample. The logging response value corresponding to each TOC logging sample is collected to form a TOC-homogeneous unit type-logging response data set B; wherein the TOC data of the coring section comes from multiple TOC test samples, the TOC data of the coring section is the core sample test data, and the logging response value refers to the logging data;

[0085] According to the number m of all mean unit types in all TOC logging samples, the TOC-homogeneous unit type-logging response dataset B is divided into m TOC-logging response datasets Bi , where each TOC-logging response data set B i The samples in the same homogeneous unit type, where i∈[1, m], are all from the same homogeneous unit type; the TOC-logging response dataset B of the target lithology of non-shale oil and gas i Set to an empty set;

[0086] Using data mining algorithms, each TOC-logging response data set B i The data in the well are used as the supervision data set to construct the TOC logging interpretation method F for each homogeneous unit type. TOC，i After ensuring that the recognition accuracy reaches the preset accuracy, the TOC logging interpretation model F corresponding to each homogeneous unit type is obtained. TOC，i .

[0087] Preferably, the TOC logging interpretation unit 25 is specifically used for:

[0088] The homogeneous unit logging identification model is used to identify the homogeneous unit type of the predicted drilling data set, and the distribution of the homogeneous unit type of the predicted drilling with depth is obtained;

[0089] For each homogeneous unit type to be predicted for drilling, the respective TOC logging interpretation model is used for TOC interpretation to obtain the TOC interpretation corresponding to each homogeneous unit type, and all TOC interpretations corresponding to the homogeneous unit type are used as the TOC logging interpretation of the shale to be predicted for drilling.

[0090] Beneficial effects achieved by the present invention: Starting from the heterogeneity of the shale formation itself, the present invention constructs a rock phase type in which other rock components except TOC are relatively homogeneous, and establishes TOC logging interpretation models for different rock types respectively, thereby avoiding underfitting or overfitting of the TOC interpretation model, enhancing the generalization ability of the interpretation model, making the TOC interpretation more accurate, and facilitating the fine evaluation of the shale reservoir.

[0091] Therefore, oil companies and other related institutions can use this method to perform fine TOC interpretation and evaluation on highly heterogeneous shale reservoirs, thereby greatly reducing the risk of shale oil and shale gas reservoirs in exploration and development. This method has broad application prospects.

[0092] The above technical solution of the embodiment of the present invention is described in detail below in conjunction with specific application examples. For technical details not introduced during the implementation process, please refer to the relevant description in the previous text.

[0093] A total organic carbon logging interpretation method based on the heterogeneity characteristics of mudstone. In view of the problem that the strong heterogeneity characteristics of shale have a great influence on the organic carbon logging interpretation results, the present invention provides an organic carbon logging interpretation technology based on the heterogeneity characteristics of shale. The basic idea is: divide the shale layer into several layers, so that in the conventional logging vertical scale (about 30 to 80 cm), other structural information except TOC has relatively homogeneous characteristics and a small range of variation, so that this part of structural information becomes redundant information and has a relatively small impact on the logging response; divide several layers into homogeneous types according to the similarity of structural information, establish a TOC interpretation model of logging response for each type, and ensure that the interpretation model in each homogeneous unit type has good generalization ability; for the shale layer to be predicted for drilling, first carry out homogeneous unit type identification, divide different types of layers according to the homogeneous unit type, and then use the established TOC interpretation model to carry out TOC logging interpretation for each layer.

[0094] The principle of the present invention is as follows:

[0095] In essence, TOC interpretation is a typical regression problem. Ideally, assume that X is a set of well logging information and Y is a set of formation information (which contains all the formation information). Let x∈X, y∈Y, both from the joint distribution P X×Y The fundamental goal of TOC logging interpretation is to establish a mapping relationship f∈F:X->Y so that the expected error ε ex =E (x，y)～P(X×Y) L(f(x),y toc ) is minimized, where L(f(x), y toc ) is the loss function, which represents the difference between the predicted TOC and the measured TOC.

[0096] In practice, the joint distribution P X×Y is often unknown. Both x and y come from a finite discrete supervised dataset Among them, x i ∈X,y i ∈Y, n represents D t The number of discrete data in the data set. In this case, the goal of TOC logging interpretation is set to establish a mapping relationship f∈F:D t，x ->D t，y , so that ε em =E (x，y)～Dt L(f(x),y toc ). Therefore, in order to reduce the error of TOC interpretation results, the most direct way is to increase the size of the Dt data set (i.e., the value of n) so that D t Can cover the joint distribution P X×YThis method can ensure that ε=|ε ex -ε em | is small enough to make the explanation model have good generalization ability and ensure that the explanation result has a small error.

[0097] However, the supervised dataset D containing complete geological information t In most cases, the supervised data set often only contains the TOC information in the well logging information and formation information. The data set can be expressed as Compared to D t Dataset, D t 'The data set is an incomplete data set, which lacks other stratigraphic information except TOC -toc Therefore, the goal of TOC logging interpretation is set as: to establish a mapping relationship f'∈F': D t，x ->D t，ytoc , so that the expected error ε em '=E (x，y)～Dt ,L(f(x),y toc ) minimized.

[0098] The prerequisite for the generalization ability of this interpretation model is that the other structural information in the shale layer except the organic matter content is relatively homogeneous and has a small range of variation. For TOC interpretation construction, this part of rock structural information is redundant information. t 'Can provide with D t Obviously, this is inconsistent with the extremely strong rock fabric heterogeneity of the shale layer. t 'The size of the data set cannot guarantee that ε=|ε ex -ε em '| is small enough, which makes the explanation model very easy to underfit or overfit, resulting in poor generalization ability, causing large errors in the TOC explanation results.

[0099] This method constructs a different t 'Supervised dataset D t，j ', which is D t 'Divided into m subsets D' t，j , j is a non-zero natural number less than m. These data sets satisfy 3 basic conditions*:

[0100] 1. All subsets DD't, j t，j The union of is equal to D t ', that is, ∪ j＝1 m D t，j '=D t '(Note: D' t，j Can be an empty set)

[0101] 2. Any two subsets are disjoint sets, i.e. D tj＝a1 '∩D t，j＝a2 '=φ, a1 and a2 are non-zero natural numbers less than m.

[0102] 3. Dataset consisting of TOC data in a single subset Each corresponds to a data set containing all stratigraphic information Among them, y -toc represents other stratigraphic information except TOC. Then, in a single subset, y -toc It has the characteristic of low dispersion, that is, σ(y -toc )<ε.

[0103] Therefore, when the above three basic conditions are met, the goal of establishing the TOC logging interpretation model is set as follows: j ∈F:D t，j，x ->D t，j，y , so that ε em，j =E (x，y)～Dt，j’ L(f(x),y toc ), so that every ε j = / ε em，j -ε ex，j / Small enough and has good generalization properties.

[0104] The specific steps of the present invention are as follows:

[0105] 1. Obtain relevant data of shale layers. The sample data set includes two categories: modeling data set (multiple wells) and prediction drilling data set. The modeling data set should include core description results of the coring section, core sample test data and logging data, while the prediction drilling data set only needs to include logging data. The core description results include lithology, sedimentary structure and mineral composition (sedimentary structure).

[0106] 2. The core sample test data in the modeling data set includes characteristic parameters after testing and analysis of mineral petrology, organic petrography, physical properties, TOC, etc. The characteristic parameters are quantitative data and typified qualitative data;

[0107] 3. Based on the core description results and core sample test data in the modeling data, a homogeneous unit division scheme is established in the core description results according to the differences in lithology, sedimentary structure and mineral composition. The scheme is used to divide the homogeneous unit types of core samples and establish the vertical distribution of homogeneous unit types of single wells (vertical distribution refers to the distribution of homogeneous unit types with depth); among them, the division standard of homogeneous unit types is determined based on the data of all drilling coring. After determination, the coring of each well is divided into single wells according to this standard.

[0108] 4. In the coring section with homogeneous unit type calibration, obtain the homogeneous unit type of each logging response point, and establish the logging response-mean unit data set A (also called: homogeneous unit-logging response data set A);

[0109] 5. Taking the well logging response-homogeneous unit type data (rock phase well logging response-rock phase data) in data set A as the supervised data set, a well logging identification method F of homogeneous unit type is constructed based on the supervised data set using the data mining algorithm. litho , and ensure that the recognition accuracy reaches the preset accuracy; obtain the homogeneous unit logging recognition model.

[0110] 6. Obtain the homogeneous unit type corresponding to each TOC test sample, collect the logging response value corresponding to each TOC logging sample, and form a TOC-homogeneous unit type-logging response data set (the data set is represented by B). Then, according to the number of mean unit types m, the data set B is divided into m TOC-logging response data sets, where the samples in each TOC-logging response data set are from the same homogeneous unit type. The above TOC-logging response data sets are represented by B. i Represents, where i∈[1,m];

[0111] 7. Using data mining algorithms, the data in each TOC-logging response data set Bi is used as a supervision data set to construct a TOC logging interpretation method F for each homogeneous unit type. TOC，i , and ensure that the recognition accuracy reaches the preset accuracy;

[0112] 8. Using the drilling data set to be predicted, first use F litho The homogeneous unit type is identified, and then, based on the above interpretation results, each rock phase type is analyzed using its own independent F TOC，i The interpretation model is used to interpret TOC, obtain the TOC interpretation results of a single well, and obtain the TOC logging interpretation model.

[0113] In summary, (1) classify homogeneous unit types; (2) use machine learning algorithms to establish a homogeneous unit logging identification model; (3) establish a TOC logging data interpretation model for each homogeneous unit; (4) use (2) and (3) to carry out TOC logging interpretation of the entire shale layer. Geology has a certain degree of complexity, and each region may have different mean unit types, so (1) and (2) are necessary.

[0114] In combination with the principle of the present invention, the steps of selecting 10 wells in the south-central region of the Ordos Basin for TOCC logging interpretation are as follows:

[0115] 1. Ten wells were selected in the central and southern part of the Ordos Basin, of which nine were used as modeling well data sets and one was used as a verification well data set to be predicted, as shown in Table 1. Table 1 shows that the modeling data set includes the core description results of the shale coring sections of the nine wells, 429 core test data and 7 basic logging curves, and the verification well contains 33 core data and 7 basic logging curves.

[0116] Table 1 Example data set display

[0117]

[0118] 2. Establish the standard for the division of homogeneous units. The basis of the division comes from core observation. In order to ensure that the homogeneous unit type is easy to obtain, the lithology, sedimentary structure and sedimentary structure of mud shale are used as the basic basis for the division of homogeneous units, and the test data are used as the basis for judging the rationality of the division. The division standards include two categories. One is that the homogeneous unit has the relative homogeneity of the stratigraphic characteristics other than TOC, that is, other stratigraphic characteristics have a relatively small distribution space; the second is that the homogeneous unit should correspond to the well logging scale, and the thickness of the homogeneous unit should be above the vertical resolution of the well logging. Therefore, in the core observation, the DEN with the minimum vertical resolution in the well logging series (the minimum vertical resolution also refers to a preset thickness) shall prevail, that is, the homogeneous unit should maintain homogeneity within 30cm of the core. Based on the core observation of 9 wells, five types of rock phases are divided in the target area, namely, silt / fine sandstone phase, clay shale phase, tuff phase, silty and clay interbedded shale, tuffaceous and clay interbedded shale. Among them, the first three are composed of a single rock type, and the last two are composed of two types of rocks. On the logging scale, the last two are relatively homogeneous. Figure 3 It is the statistical result of the above five homogeneous units in core, thin section, mineral composition, organic matter content and organic matter occurrence state, which shows that there are obvious differences between different homogeneous unit types, while the stratigraphic characteristics other than TOC within the homogeneous unit type are relatively similar.

[0119] 3. According to the homogeneous unit division standard, determine the vertical homogeneous unit distribution of 9 core observation wells and determine the mean unit attribute of the core data ( Figure 4 The vertical distribution of homogeneous units, TOC data points and logging information of Well YY22 is shown).

[0120] 4. Based on the division of single well homogeneous units, first extract the logging attributes of the homogeneous units (here, the logging attributes are logging response values, including AC, CAL, DEN, GR, K, Rt, PE, RD, RS, Th, U) to form a homogeneous unit type-logging response data set. In this embodiment, the length of the observed core is 230m, and 1836 sets of homogeneous unit type-logging response data sets are obtained. The mean unit type is numerically coded, where tuff is 1, tuffaceous and clay interlayer shale is coded as 2, clay shale is coded as 3, silt and clay interlayer shale is coded as 4, and siltstone / fine sandstone is coded as 5. Table 2 shows some data in the homogeneous unit-logging response data set.

[0121] 5. Then extract the logging attributes corresponding to the TOC data points in the core in different homogeneous units (the logging attributes here are logging response values, including AC, CAL, DEN, GR, PE and RT. As shown in Table 3, except for TOC, together with TOC, they constitute the modeling supervision data set of the TOC logging interpretation model) to form the TOC-logging response data set under different homogeneous unit types. Since tuff is not the target lithology of shale oil and gas, its data is relatively small, and its TOC values ​​are all below 1.0wt.%. Therefore, the TOC-logging response data set described for tuff is an empty set. In the construction of the data set, three comprehensive parameters, Δ1gR, U / Thh and Th / K, were added as additional logging parameters. The above three parameters are often used by predecessors to interpret logging TOC. After the above division process, the TOC-test response data sets of four homogeneous units, namely, tuffaceous and clayey interbedded shale, clayey shale, silty and clayey interbedded shale, and siltstone / fine sandstone, were obtained, with data volumes of 171, 214, 72, and 55, respectively. Table 3 shows the data set of clayey shale in the TOC-logging response data set.

[0122]

[0123]

[0124] 6. The homogeneous unit-logging response data set is divided into two parts. The corresponding data of WY1 and FY3 are extracted as the validation well data set (the number of data sets is 144 groups), and the other well data are allocated as the training data set (the number of data sets is 1353 groups).

[0125] 7. Use machine learning algorithms to predict homogeneous unit types based on logging information. In this embodiment, the XGboost algorithm (extreme gradient boosting tree) is used to establish a well logging identification model for homogeneous unit types. The XGboost classification model is trained using the modeling data set as input data, and the K-fold crossover operation is used to ensure that the trained model has the best robustness. In view of the problem of hyperparameter selection in XGboost, this embodiment uses a genetic algorithm for optimization, in which the fitness function of the genetic algorithm is set to the cross-validation accuracy of XGboost (mlogloss is selected in this paper), and the goal of the genetic algorithm is to find a hyperparameter series that minimizes the cross-validation accuracy. After the population of the initial hyperparameter series is given, the XGboost cross-validation accuracy corresponding to each individual in the population is obtained, and individuals in the first generation population are selected according to the elite retention mode, and a new generation population is generated through operations such as crossover and mutation. The above process is iterated repeatedly until the set number of iterations is reached or the fitness function reaches the set threshold, and the hyperparameter system that makes the fitness function reach the lowest is selected.

[0126] Figure 5 (a) shows the hyperparameter settings of the XGboost algorithm in the embodiment and the hyperparameter range to be optimized by the genetic algorithm, where K=7 in K-fold; Figure 5 (b) is the iterative process of the average cross-validation accuracy and the optimal cross-validation accuracy of XGboost when the genetic algorithm is used for optimization (the number of populations per generation is 150 and the number of genetic iterations is 200). The above process determines Figure 5 (a) The optimal hyperparameter series and XGboost model shown. Figure 5 (c) shows the comparison between the real homogeneous unit type and the predicted homogeneous unit type in the WY1 well, indicating that the homogeneous unit logging identification method has excellent recognition accuracy, and the recognition accuracy rate can reach 88.3%.

[0127] 8. The TOC-logging response data set is divided. First, 20% of the data are extracted from the TOC-logging response data set of each homogeneous unit type to form a verification data set of each homogeneous unit type (the number of data is 86), and the remaining data are used to form a modeling supervision data set of each homogeneous unit. A total of 8 sub-datasets are constructed.

[0128] 9. Use a machine learning algorithm to establish a TOC well logging data interpretation model (TOC well logging interpretation model) for each homogeneous unit type. In this embodiment, the SVM regression algorithm is used to construct a TOC well logging data prediction model (TOC well logging interpretation model). The training data set composed of each mean unit type is used as input data to train the SVM regression algorithm, thereby training and obtaining the TOC well logging data prediction model of the four unit types. K-fold cross validation is used in training to improve the robustness of the interpretation model. For the hyperparameters in the SVM, this embodiment uses a genetic algorithm for optimization, and the fitness function of the genetic algorithm is set to the cross-square error of the SVM under cross validation. The goal of the genetic algorithm is to find a hyperparameter series that minimizes the cross-square error. After the population of the initial hyperparameter series is given, the cross-square error corresponding to each individual in the population is obtained, and the individuals in the first generation population are selected according to the elite retention mode. A new generation of population is generated through operations such as crossover and mutation. The above process is iterated repeatedly until the set number of iterations is reached or the fitness function reaches the set threshold. The hyperparameter system that makes the fitness function reach the lowest is selected, such as Figure 6 shown.

[0129] Table 4 shows the main parameter settings in the SVM algorithm and genetic algorithm in the embodiment. Figure 5 This is the iterative process of the average recognition accuracy and the optimal recognition accuracy of SVM when the genetic algorithm is used for optimization (the number of populations per generation is 150, the number of genetic iterations is 200, and K = 7). The above process determines the optimal hyperparameter series and SVM explanation model shown in Table 5. Figure 7 As shown, the accuracy indicators MSE=1.18 and MAPE=20.57% in the validation set using this method are much higher than the accuracy indicators MSE=3.27 and MAPE=29.54% obtained by the traditional method.

[0130]

[0131] 10. Based on the establishment of the homogeneous unit logging identification method and the TOC logging interpretation method of each homogeneous unit, the TOC logging interpretation of the whole well section of the shale layer is carried out. First, the homogeneous unit is identified by using the homogeneous unit logging identification method and logging data, and then the TOC in each homogeneous unit is interpreted by using the TOC logging interpretation method of each homogeneous unit. Figure 8 shown.

[0132] The beneficial effects achieved by the present invention are:

[0133] The present invention starts from the heterogeneity of the shale formation itself, constructs rock phase types in which other rock components except TOC are relatively homogeneous, and establishes TOC logging interpretation models for different rock types respectively, so as to avoid underfitting or overfitting of the TOC interpretation model, enhance the generalization ability of the interpretation model, make TOC interpretation more accurate, and facilitate the fine evaluation of shale reservoirs.

[0134] Therefore, oil companies and other related institutions can use this method to perform fine TOC interpretation and evaluation on highly heterogeneous shale reservoirs, thereby greatly reducing the risk of shale oil and shale gas reservoirs in exploration and development. This method has broad application prospects.

[0135] It should be understood that the specific order or hierarchy of steps in the disclosed process is an example of an exemplary method. Based on design preferences, it should be understood that the specific order or hierarchy of steps in the process can be rearranged without departing from the scope of protection of the present disclosure. The attached method claims present the elements of the various steps in an exemplary order and are not intended to be limited to the specific order or hierarchy described.

[0136] In the above detailed description, various features are grouped together in a single embodiment to simplify the disclosure. This method of disclosure should not be interpreted as reflecting an intention that the embodiments of the claimed subject matter require more features than are clearly stated in each claim. On the contrary, as reflected in the appended claims, the invention is in a state of having less than all the features of the disclosed individual embodiments. Therefore, the appended claims are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate preferred embodiment of the invention.

[0137] The disclosed embodiments are described above to enable any person skilled in the art to implement or use the present invention. Various modifications of these embodiments are obvious to those skilled in the art, and the general principles defined herein may also be applied to other embodiments without departing from the spirit and scope of the present disclosure. Therefore, the present disclosure is not limited to the embodiments given herein, but is consistent with the broadest scope of the principles and novel features disclosed in this application.

[0138] The above description includes examples of one or more embodiments. Of course, it is impossible to describe all possible combinations of components or methods for the purpose of describing the above embodiments, but it should be recognized by those skilled in the art that the various embodiments may be further combined and arranged. Therefore, the embodiments described herein are intended to cover all such changes, modifications and variations that fall within the scope of protection of the appended claims. In addition, with respect to the term "comprising" used in the specification or claims, the word is covered in a manner similar to the term "including", just as "including," is explained as a transitional word in the claims. In addition, any term "or" used in the specification of the claims is intended to mean "non-exclusive or".

[0139] Those skilled in the art may also understand that the various illustrative logical blocks, units, and steps listed in the embodiments of the present invention may be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly demonstrate the interchangeability of hardware and software, the various illustrative components, units, and steps described above have generally described their functions. Whether such functions are implemented by hardware or software depends on the specific application and the design requirements of the entire system. Those skilled in the art may use various methods to implement the described functions for each specific application, but such implementation should not be understood as exceeding the scope of protection of the embodiments of the present invention.

[0140] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A total organic carbon logging interpretation method based on the heterogeneity characteristics of shale, characterized in that: include: Obtaining shale layer-related data, the shale layer-related data includes two categories: a modeling data set and a drilling data set to be predicted; wherein the modeling data set includes core description results, core sample test data and logging data of the coring section, and the drilling data set to be predicted includes logging data; in the modeling data set, the core sample test data includes characteristic parameters after testing and analyzing the total organic carbon content TOC of mineral petrology and organic petrography, and the characteristic parameters are quantitative data and typified qualitative data; the total organic carbon content is a characterization index of the organic matter content in shale; According to the core description results and core sample test data in the modeling data, a division benchmark of homogeneous unit types is established, and the homogeneous unit types of the coring section are divided using the division benchmark of the homogeneous unit types to obtain the distribution of the homogeneous unit types of the coring section of each single well with depth; For each homogeneous unit type, a homogeneous unit logging identification model is established using a machine learning algorithm to predict the homogeneous unit type using the logging response of the shale layer. A machine learning algorithm was used to establish a TOC logging interpretation model for each homogeneous unit type; Using the drilling data set to be predicted, the homogeneous unit logging identification model is used to identify the homogeneous unit types at different depths of a single well, and the TOC logging interpretation of the shale layer is carried out based on the TOC logging interpretation model of each homogeneous unit; For each homogeneous unit type, a homogeneous unit logging identification model is established by using a machine learning algorithm to predict the homogeneous unit type by using the logging response of the shale layer segment, specifically including: In the coring section with homogeneous unit type calibration, the homogeneous unit type of each logging response point is obtained from the modeling data set; For each homogeneous unit type, a respective well logging response-homogeneous unit type data set A is established; All the logging responses and homogeneous unit type data in the logging response-homogeneous unit data set A are taken as the supervision data set, and the well logging identification method F of homogeneous unit type is constructed based on the supervision data set using the data mining algorithm. litho After ensuring that the recognition accuracy of all logging responses and homogeneous unit type data reaches the preset accuracy, the homogeneous unit logging recognition model F is obtained. litho ; The machine learning algorithm is used to establish a TOC logging interpretation model for each homogeneous unit type, specifically including: The TOC data of the coring section is obtained, and the corresponding mean unit type is obtained by using the depth information of each TOC test sample. The logging response value corresponding to each TOC logging sample is collected to form a TOC-homogeneous unit type-logging response data set B; wherein the TOC data of the coring section comes from multiple TOC test samples, the TOC data of the coring section is the core sample test data, and the logging response value refers to the logging data; According to the number m of all mean unit types in all TOC logging samples, the TOC-homogeneous unit type-logging response dataset B is divided into m TOC-logging response datasets B i ; Each TOC-logging response data set B i The samples in the same homogeneous unit type, where i∈[1,m], are the TOC-logging response dataset B of the target lithology of non-shale oil and gas. i Set to an empty set; Using data mining algorithms, each TOC-logging response data set B i The data in the well are used as the supervision data set to construct the TOC logging interpretation method F for each homogeneous unit type. TOC，i After ensuring that the recognition accuracy reaches the preset accuracy, the TOC logging interpretation model F corresponding to each homogeneous unit type is obtained. TOC，i .

2. The total organic carbon logging interpretation method based on shale heterogeneity characteristics according to claim 1 is characterized in that: The core description results of the shale include: lithology, sedimentary structure and sedimentary structure; The division benchmark of the homogeneous unit type is established according to the core description results and core sample test data in the modeling data; The core sample homogeneous unit types are divided using the division criteria of the homogeneous unit types. The distribution of the homogeneous unit types of each single well with depth specifically includes: The lithology, sedimentary structure and sedimentary structure of shale are used as the basic basis for the division of homogeneous units, and the core sample test data is used as a judgment, so that any homogeneous unit to be divided has the following characteristics at the same time: Except for TOC, the formation characteristic parameters are relatively homogeneous, that is, other formation characteristic parameters have a relatively small distribution space; The thickness of the homogeneous unit should be at least a first thickness, and the first thickness is greater than the minimum vertical resolution of conventional well logging; any of the homogeneous units is used as a basis for dividing homogeneous unit types; According to the characteristics of any homogeneous unit, the well logging data corresponding to the modeling data set is divided into multiple different homogeneous units.

3. The total organic carbon logging interpretation method based on shale heterogeneity characteristics according to claim 1 is characterized in that: The method uses the drilling data set to be predicted and the homogeneous unit logging identification model to identify the homogeneous unit types of different depth sections of a single well, and carries out TOC logging interpretation of shale layers according to the TOC logging interpretation model of each homogeneous unit, specifically including: The homogeneous unit logging identification model is used to identify the homogeneous unit type of the predicted drilling data set, and the distribution of the homogeneous unit type of the predicted drilling with depth is obtained; For each homogeneous unit type to be predicted for drilling, the respective TOC logging interpretation model is used for TOC interpretation to obtain the TOC interpretation corresponding to each homogeneous unit type, and all TOC interpretations corresponding to the homogeneous unit type are used as the TOC logging interpretation of the shale to be predicted for drilling.

4. A total organic carbon logging interpretation system based on the heterogeneity characteristics of shale, characterized in that: include: A data acquisition unit is used to acquire shale interval related data, wherein the shale interval related data includes two categories: a modeling data set and a drilling data set to be predicted; wherein the modeling data set includes core description results, core sample test data and logging data of the coring section, and the drilling data set to be predicted includes logging data; in the modeling data set, the core sample test data includes characteristic parameters after testing and analyzing the total organic carbon content TOC of mineral petrology and organic petrography, and the characteristic parameters are quantitative data and typified qualitative data; the total organic carbon content is a characterization index of the organic matter content in the shale; The homogeneous unit type division unit is used to establish a division basis of the homogeneous unit type according to the core description results and core sample test data in the modeling data, and to divide the homogeneous unit type of the coring section using the division basis of the homogeneous unit type to obtain the distribution of the homogeneous unit type of each single well coring section with depth; The homogeneous unit type prediction unit is used to establish a homogeneous unit logging identification model for each homogeneous unit type using a machine learning algorithm, so as to predict the homogeneous unit type using the logging response of the shale layer segment; The TOC logging interpretation model establishment unit uses a machine learning algorithm to establish a separate TOC logging interpretation model for each homogeneous unit type; The TOC logging interpretation unit is used to use the drilling data set to be predicted and the homogeneous unit logging identification model to identify the homogeneous unit types at different depths of a single well, and to carry out TOC logging interpretation of shale layers based on the TOC logging interpretation model of each homogeneous unit; The homogeneous unit type prediction unit is specifically used for: In the coring section with homogeneous unit type calibration, the homogeneous unit type of each logging response point is obtained from the modeling data set; For each homogeneous unit type, a respective well logging response-homogeneous unit type data set A is established; All the logging responses and homogeneous unit type data in the logging response-homogeneous unit data set A are taken as the supervision data set, and the well logging identification method F of homogeneous unit type is constructed based on the supervision data set using the data mining algorithm. litho After ensuring that the recognition accuracy of all logging responses and homogeneous unit type data reaches the preset accuracy, the homogeneous unit logging recognition model F is obtained. litho ; The TOC logging interpretation model building unit is specifically used for: The TOC data of the coring section is obtained, and the corresponding mean unit type is obtained by using the depth information of each TOC test sample. The logging response value corresponding to each TOC logging sample is collected to form a TOC-homogeneous unit type-logging response data set B; wherein the TOC data of the coring section comes from multiple TOC test samples, the TOC data of the coring section is the core sample test data, and the logging response value refers to the logging data; According to the number m of all mean unit types in all TOC logging samples, the TOC-homogeneous unit type-logging response dataset B is divided into m TOC-logging response datasets B i ; Each TOC-logging response data set B i The samples in the same homogeneous unit type, where i∈[1,m], are the TOC-logging response dataset B of the target lithology of non-shale oil and gas. i Set to an empty set; Using data mining algorithms, each TOC-logging response data set B i The data in the well are used as the supervision data set to construct the TOC logging interpretation method F for each homogeneous unit type. TOC，i After ensuring that the recognition accuracy reaches the preset accuracy, the TOC logging interpretation model F corresponding to each homogeneous unit type is obtained. TOC，i .

5. The total organic carbon logging interpretation system based on shale heterogeneity characteristics according to claim 4 is characterized in that: The core description results of the shale include: lithology, sedimentary structure and sedimentary structure; The homogeneous unit type division unit is specifically used for: The lithology, sedimentary structure and sedimentary structure of shale are used as the basic basis for the division of homogeneous units, and the core sample test data is used as a judgment, so that any homogeneous unit to be divided has the following characteristics at the same time: Except for TOC, the formation characteristic parameters are relatively homogeneous, that is, other formation characteristic parameters have a relatively small distribution space; The thickness of the homogeneous unit should be at least a first thickness, and the first thickness is greater than the minimum vertical resolution of conventional well logging; any of the homogeneous units is used as a basis for dividing homogeneous unit types; According to the characteristics of any homogeneous unit, the well logging data corresponding to the modeling data set is divided into multiple different homogeneous units.

6. The total organic carbon logging interpretation system based on shale heterogeneity characteristics according to claim 4 is characterized in that: The TOC logging interpretation unit is specifically used for: The homogeneous unit logging identification model is used to identify the homogeneous unit type of the predicted drilling data set, and the distribution of the homogeneous unit type of the predicted drilling with depth is obtained; For each homogeneous unit type to be predicted for drilling, the respective TOC logging interpretation model is used for TOC interpretation to obtain the TOC interpretation corresponding to each homogeneous unit type, and all TOC interpretations corresponding to the homogeneous unit type are used as the TOC logging interpretation of the shale to be predicted for drilling.

Images