Lithology identification method and device, electronic equipment and medium
By establishing a well logging interpretation volume model and a two-way long short-term memory network, the problem of unfocused identification of buried hill lithology was solved, the reliability and accuracy of lithology identification were improved, and the operation process was simplified.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA PETROLEUM & CHEMICAL CORP
- Filing Date
- 2024-10-31
- Publication Date
- 2026-05-01
AI Technical Summary
Existing technologies do not cover the identification of buried hill lithology, resulting in unfocused lithology identification results, imprecise theories, and complex operations.
By establishing a well logging interpretation volume model based on lithology logging geological classification, the reservoir mineral composition is optimized and inverted. A bidirectional long short-term memory network based on mineral composition-lithology sequence data is constructed for lithology identification. Combined with an attention-enhanced bidirectional long short-term memory network model, a lithology identification model considering the variation trend of mineral composition and the correlation between upper and lower data is constructed.
It improves the reliability and accuracy of lithology identification, making the lithology identification results more focused, with rigorous theory and simple and practical operation procedures.
Smart Images
Figure CN121960091A_ABST
Abstract
Description
Lithology identification methods, devices, electronic equipment and media Technical Field
[0001] This invention relates to the field of lithological identification and interpretation, and more specifically, to a lithological identification method, apparatus, electronic device, and medium. Background Technology
[0002] With rapid economic development, the demand for oil and gas resources is constantly rising. In recent years, unconventional oil and gas resources, deep-sea oil and gas resources, deep-water offshore oil and gas resources, and remaining oil and gas reserves in old oil and gas areas have shown good growth momentum, representing the future development direction of oil and gas resources and the main source of increased reserves and production. Deep-water areas have gradually become the main battlefield for the next stage of exploration and the replacement area for new reserves, and buried hill oil and gas reservoirs have become important potential exploration targets. However, existing technologies do not cover the identification technology of buried hill lithology.
[0003] Therefore, it is necessary to develop a lithology identification method, device, electronic equipment, and medium.
[0004] The information disclosed in the background section of this invention is intended only to enhance the understanding of the general background of this invention, and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art. Summary of the Invention
[0005] This invention proposes a lithology identification method, device, electronic equipment, and medium, which can improve lithology identification results, make anomalies more focused, and has a rigorous and reliable theory and a simple and practical operation process.
[0006] In a first aspect, embodiments of this disclosure provide a lithology identification method, including:
[0007] Establish a well logging interpretation volume model based on lithological well logging geological classification;
[0008] Optimization inversion of reservoir mineral composition based on well logging interpretation volume model;
[0009] Construct a training sample dataset based on "mineral composition-lithology" sequence data;
[0010] Based on the inversion results, a bidirectional long short-term memory network based on "mineral composition-lithology" sequence data is constructed for lithology identification.
[0011] As a specific implementation of this disclosure, establishing a well logging interpretation volume model based on lithological well logging geological classification includes:
[0012] Determine the lithology and corresponding minerals of a known area;
[0013] The skeletal lines of each mineral were drawn, and density-sound wave and density-neutron plates were then created;
[0014] A well logging interpretation volume model is established based on density-sonic wave and density-neutron charts.
[0015] As a specific implementation of this disclosure, the optimization inversion of reservoir mineral composition based on the well logging interpretation volume model includes:
[0016] Based on the well logging interpretation volume model, establish the well logging response equation;
[0017] Establish a constrained system of overdetermined linear equations.
[0018] As a specific implementation of this disclosure, the well logging response equation is as follows:
[0019]
[0020] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component.
[0021] As a specific implementation of this disclosure, the overdetermined linear equation system is as follows:
[0022]
[0023] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component;
[0024] The constraints of the overdetermined linear system of equations are:
[0025]
[0026] As a specific implementation of this disclosure, a lithology identification model that considers the variation trend of mineral composition and the correlation between upper and lower data is constructed by leveraging the memory capacity of an attention-enhanced bidirectional long short-term memory network model.
[0027] The training sample dataset is fed into the lithology identification model for training, and lithology identification is performed based on the trained model.
[0028] Secondly, embodiments of this disclosure also provide a lithology identification device, comprising:
[0029] The well logging interpretation module establishes a well logging interpretation volume model based on lithological well logging geological classification;
[0030] The inversion module performs optimal inversion of reservoir mineral composition based on the well logging interpretation volume model.
[0031] The training sample construction module builds a training sample dataset based on the "mineral composition-lithology" sequence data;
[0032] The identification module constructs a bidirectional long short-term memory network based on the inversion results to identify lithology.
[0033] As a specific implementation of this disclosure, establishing a well logging interpretation volume model based on lithological well logging geological classification includes:
[0034] Determine the lithology and corresponding minerals of a known area;
[0035] The skeletal lines of each mineral were drawn, and density-sound wave and density-neutron plates were then created;
[0036] A well logging interpretation volume model is established based on density-sonic wave and density-neutron charts.
[0037] As a specific implementation of this disclosure, the optimization inversion of reservoir mineral composition based on the well logging interpretation volume model includes:
[0038] Based on the well logging interpretation volume model, establish the well logging response equation;
[0039] Establish a constrained system of overdetermined linear equations.
[0040] As a specific implementation of this disclosure, the well logging response equation is as follows:
[0041]
[0042] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component.
[0043] As a specific implementation of this disclosure, the overdetermined linear equation system is as follows:
[0044]
[0045] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component;
[0046] The constraints of the overdetermined linear system of equations are:
[0047]
[0048] As a specific implementation of this disclosure, a lithology identification model that considers the variation trend of mineral composition and the correlation between upper and lower data is constructed by leveraging the memory capacity of an attention-enhanced bidirectional long short-term memory network model.
[0049] The training sample dataset is fed into the lithology identification model for training, and lithology identification is performed based on the trained model.
[0050] Thirdly, embodiments of this disclosure also provide an electronic device, the electronic device comprising:
[0051] Memory, which stores executable instructions;
[0052] A processor that executes the executable instructions in the memory to implement the lithology identification method.
[0053] Fourthly, embodiments of this disclosure also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the described lithology identification method.
[0054] Its beneficial effects are as follows:
[0055] This invention introduces abstract lithological patterns into the network model training process in a concrete form, improving the reliability of intelligent prediction of complex lithology, and is of great significance for lithology identification and subsequent processing.
[0056] The methods and apparatus of the present invention have other features and advantages that will be apparent from or will be set forth in detail in the accompanying drawings and following detailed description, which together serve to explain the particular principles of the invention. Attached Figure Description
[0057] The above and other objects, features and advantages of the present invention will become more apparent from the more detailed description of exemplary embodiments of the invention in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same parts.
[0058] Figure 1 shows a flowchart of the steps of a lithology identification method according to an embodiment of the present invention.
[0059] Figure 2 shows a schematic diagram illustrating the relationship between the contents of quartz, potassium feldspar, and plagioclase in granite buried hills according to an embodiment of the present invention.
[0060] Figure 3 shows a schematic diagram of the relationship between the contents of quartz, feldspar, and clay in granite buried hills according to an embodiment of the present invention.
[0061] Figure 4 shows a schematic diagram of an LSTM cell structure according to an embodiment of the present invention.
[0062] Figure 5 shows a schematic diagram of the lithology identification results of three models according to an embodiment of the present invention.
[0063] Figure 6 shows a block diagram of a lithology identification device according to an embodiment of the present invention.
[0064] Explanation of reference numerals in the attached figures:
[0065] 201. Well Logging Interpretation Module; 202. Inversion Module; 203. Training Sample Construction Module; 204. Identification Module. Detailed Implementation
[0066] Preferred embodiments of the invention will now be described in more detail. While preferred embodiments of the invention are described below, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.
[0067] To facilitate understanding of the solutions and effects of the embodiments of the present invention, six specific application examples are given below. Those skilled in the art should understand that these examples are merely for the purpose of understanding the present invention, and any specific details therein are not intended to limit the present invention in any way.
[0068] Example 1
[0069] Figure 1 shows a flowchart of the steps of a lithology identification method according to an embodiment of the present invention.
[0070] As shown in Figure 1, this lithology identification method includes:
[0071] Step 101: Establish a well logging interpretation volume model based on lithological well logging geological classification;
[0072] Step 102: Optimize the inversion of reservoir mineral composition based on the well logging interpretation volume model;
[0073] Step 103: Construct a training sample dataset based on the "mineral composition-lithology" sequence data;
[0074] Step 104: Based on the inversion results, construct a bidirectional long short-term memory network based on the "mineral composition-lithology" sequence data for lithology identification.
[0075] In one example, establishing a well logging interpretation volumetric model based on lithology logging geological classification includes:
[0076] Determine the lithology and corresponding minerals of a known area;
[0077] The skeletal lines of each mineral were drawn, and density-sound wave and density-neutron plates were then created;
[0078] A well logging interpretation volume model is established based on density-sonic wave and density-neutron charts.
[0079] In one example, the optimal inversion of reservoir mineral composition based on the well logging interpretation volume model includes:
[0080] Based on the well logging interpretation volume model, establish the well logging response equation;
[0081] Establish a constrained system of overdetermined linear equations.
[0082] In one example, the well logging response equation is:
[0083]
[0084] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component.
[0085] In one example, the overdetermined linear equation system is:
[0086]
[0087] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component;
[0088] The constraints of the overdetermined linear system of equations are:
[0089]
[0090] In one example, the memory capacity of an attention-enhanced bidirectional long short-term memory network model is used to construct a lithology identification model that considers the variation trend of mineral composition and the correlation between upstream and downstream data.
[0091] The training sample dataset is fed into the lithology identification model for training, and lithology identification is performed based on the trained model.
[0092] Specifically, a well logging interpretation volumetric model based on lithological well logging geological classification is established. A combined triaxial compression and X-ray diffraction (TCD) experiment is designed to analyze the differences in brittleness indices among different lithologies in deep-water buried hill oil and gas reservoirs. This guides the lithological "composition-structure" secondary well logging geological classification. Based on the classification principles of "systematicity, clarity, and practicality," a well logging interpretation volumetric model for buried hills is established that can systematically explain the petrological mechanisms of buried hills, clearly reflect the differences in well logging responses among different lithologies, and also take into account practicality.
[0093] Analysis of logging and rock and mineral experiments reveals that the main lithologies in the buried hill area of HZ region include granite, diorite, diabase, gabbro, dacite, and andesite. Their main mineral components include quartz, potassium feldspar, plagioclase (including anorthite and sodium feldspar), and dark minerals. Based on this, framework lines for each mineral (quartz, potassium feldspar, sodium feldspar, anorthite, and dark minerals (using amphibole as an example) are plotted according to the relationship between mineral framework values and porosity (water saturation in pores). All framework lines are plotted on the same cross-plot, creating density-acoustic and density-neutron charts. Based on these charts, the influence of mineral composition and content on rock brittleness and their control over dominant reservoirs are analyzed. Several single minerals or mineral combinations are selected to establish a logging interpretation volumetric model.
[0094] Optimal inversion of mineral composition in buried hill reservoirs under the constraints of whole-rock mineral quantitative analysis data. Based on a multi-mineral, multi-component optimal interpretation model, this study conducts quantitative calculations of the mineral skeleton content in buried hill reservoirs. Discrete information such as two-dimensional minerals, thin sections, X-ray diffraction, and cuttings logging is introduced as constraints into the optimal quantitative calculation of mineral skeleton content, achieving organic integration with conventional logging and ECS elemental logging data, thereby improving the accuracy of reservoir mineral component content calculations.
[0095] (1) Establish the logging response equation
[0096] Based on the lithological information from drilling cores, cuttings logging, and conventional logging data after depth repositioning, highly collinear curves were removed using correlation coefficient analysis methods such as Pearson's method, and conventional logging curves sensitive to lithology were selected. Based on the logging interpretation volume model, the logging response equation corresponding to the selected logging curve was established. The general formula of the response equation is as follows:
[0097]
[0098] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component.
[0099] (2) Establish a constrained system of overdetermined linear equations
[0100] ①Establish constraints for experimental data
[0101] Figure 2 shows a schematic diagram illustrating the relationship between the contents of quartz, potassium feldspar, and plagioclase in granite buried hills according to an embodiment of the present invention.
[0102] As shown in Figure 2, the analysis of the whole-rock X-ray diffraction results of granite buried hills in Huizhou, Weizhou, and Yongle areas reveals the following relationship among the three:
[0103]
[0104] In the formula: V sh V sand V P-fel and V K-fel The values are the clay content, quartz content, plagioclase content, and potassium feldspar content measured by X-ray diffraction, in % (%). These four minerals are the main minerals in this region, and their combined content is close to 1%.
[0105] The commonly used objective function form for optimal interpretation is as follows:
[0106]
[0107] In the formula: x j The content of the j-th stratigraphic component includes clay minerals, quartz, plagioclase, potassium feldspar, and pore fluids, etc.; A ij B is the skeleton response value of the i-th logging curve for the j-th mineral; i Let be the logging value of the i-th logging curve.
[0108] The constraint in this objective function is a rock physics model constraint, namely, the sum of the contents of all formation components is 1. Converting the contents of quartz, potassium feldspar, and plagioclase obtained from X-ray diffraction into the content relationships of formation mineral components yields the new optimization interpretation constraint:
[0109]
[0110] A new optimized interpretation model can calculate the mineral content of granite reservoirs in the northern South China Sea by integrating well logging data and X-ray diffraction experimental data. The introduction of direct geological data (rock cuttings data) will help improve the accuracy of mineral content calculations.
[0111] ②Constraints for establishing logging data
[0112] Figure 3 shows a schematic diagram of the relationship between the contents of quartz, feldspar, and clay in granite buried hills according to an embodiment of the present invention.
[0113] As shown in Figure 3, the analysis of the quartz, feldspar, and clay mineral contents in the X-ray diffraction results of granite buried hill logging in Huizhou and Yongle areas reveals the following relationship:
[0114]
[0115] Similar to the process of establishing constraints using experimental data, the above relationships are transformed into optimization constraints and added to the optimization calculation equations to obtain a new optimization interpretation objective function. By comprehensively integrating well logging information, the mineral composition of the formation can be jointly inverted, thereby improving the calculation accuracy.
[0116] Solving constrained optimization problems in mineral composition optimization inversion. Addressing issues such as the optimization results being heavily influenced by the initial point, susceptibility to local optima, and slow search speed in optimization interpretation, this paper proposes improved or combined optimization algorithms that offer faster global search speeds, higher computational efficiency and accuracy, and more effective solutions to constrained optimization problems through comparative analysis of different optimization algorithms.
[0117] Figure 4 shows a schematic diagram of an LSTM cell structure according to an embodiment of the present invention.
[0118] Lithology identification based on Bidirectional Long Short-Term Memory (Bi-LSTM) network of "mineral composition-lithology" sequence data. As shown in Figure 4, the Bi-LSTM lithology identification model was used to identify the lithology of buried hill reservoirs in the northern South China Sea, and the results were compared with those predicted by classical machine learning algorithms to test and evaluate the identification effect of the Bi-LSTM-based lithology identification model.
[0119] Example 2
[0120] The present invention also provides a lithology identification device, comprising:
[0121] The well logging interpretation module establishes a well logging interpretation volume model based on lithological well logging geological classification;
[0122] The inversion module performs optimal inversion of reservoir mineral composition based on the well logging interpretation volume model.
[0123] The training sample construction module builds a training sample dataset based on the "mineral composition-lithology" sequence data;
[0124] The identification module constructs a bidirectional long short-term memory network based on the inversion results to identify lithology.
[0125] In one example, establishing a well logging interpretation volumetric model based on lithology logging geological classification includes:
[0126] Determine the lithology and corresponding minerals of a known area;
[0127] The skeletal lines of each mineral were drawn, and density-sound wave and density-neutron plates were then created;
[0128] A well logging interpretation volume model is established based on density-sonic wave and density-neutron charts.
[0129] In one example, the optimal inversion of reservoir mineral composition based on the well logging interpretation volume model includes:
[0130] Based on the well logging interpretation volume model, establish the well logging response equation;
[0131] Establish a constrained system of overdetermined linear equations.
[0132] In one example, the well logging response equation is:
[0133]
[0134] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component.
[0135] In one example, the overdetermined linear equation system is:
[0136]
[0137] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component;
[0138] The constraints of the overdetermined linear system of equations are:
[0139]
[0140] In one example, the memory capacity of an attention-enhanced bidirectional long short-term memory network model is used to construct a lithology identification model that considers the variation trend of mineral composition and the correlation between upstream and downstream data.
[0141] The training sample dataset is fed into the lithology identification model for training, and lithology identification is performed based on the trained model.
[0142] Specifically, a well logging interpretation volumetric model based on lithological well logging geological classification is established. A combined triaxial compression and X-ray diffraction (TCD) experiment is designed to analyze the differences in brittleness indices among different lithologies in deep-water buried hill oil and gas reservoirs. This guides the lithological "composition-structure" secondary well logging geological classification. Based on the classification principles of "systematicity, clarity, and practicality," a well logging interpretation volumetric model for buried hills is established that can systematically explain the petrological mechanisms of buried hills, clearly reflect the differences in well logging responses among different lithologies, and also take into account practicality.
[0143] Analysis of logging and rock and mineral experiments reveals that the main lithologies in the buried hill area of HZ region include granite, diorite, diabase, gabbro, dacite, and andesite. Their main mineral components include quartz, potassium feldspar, plagioclase (including anorthite and sodium feldspar), and dark minerals. Based on this, framework lines for each mineral (quartz, potassium feldspar, sodium feldspar, anorthite, and dark minerals (using amphibole as an example) are plotted according to the relationship between mineral framework values and porosity (water saturation in pores). All framework lines are plotted on the same cross-plot, creating density-acoustic and density-neutron charts. Based on these charts, the influence of mineral composition and content on rock brittleness and their control over dominant reservoirs are analyzed. Several single minerals or mineral combinations are selected to establish a logging interpretation volumetric model.
[0144] Optimal inversion of mineral composition in buried hill reservoirs under the constraints of whole-rock mineral quantitative analysis data. Based on a multi-mineral, multi-component optimal interpretation model, this study conducts quantitative calculations of the mineral skeleton content in buried hill reservoirs. Discrete information such as two-dimensional minerals, thin sections, X-ray diffraction, and cuttings logging is introduced as constraints into the optimal quantitative calculation of mineral skeleton content, achieving organic integration with conventional logging and ECS elemental logging data, thereby improving the accuracy of reservoir mineral component content calculations.
[0145] (1) Establish the logging response equation
[0146] Based on the lithological information from drilling cores, cuttings logging, and conventional logging data after depth repositioning, highly collinear curves were removed using correlation coefficient analysis methods such as Pearson's method, and conventional logging curves sensitive to lithology were selected. Based on the logging interpretation volume model, the logging response equation corresponding to the selected logging curve was established. The general formula of the response equation is as follows:
[0147]
[0148] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component.
[0149] (2) Establish a constrained system of overdetermined linear equations
[0150] ①Establish constraints for experimental data
[0151] Figure 2 shows a schematic diagram illustrating the relationship between the contents of quartz, potassium feldspar, and plagioclase in granite buried hills according to an embodiment of the present invention.
[0152] As shown in Figure 2, the analysis of the whole-rock X-ray diffraction results of granite buried hills in Huizhou, Weizhou, and Yongle areas reveals the following relationship among the three:
[0153]
[0154] In the formula: V sh V sand V P-fel and V K-fel The values are the clay content, quartz content, plagioclase content, and potassium feldspar content measured by X-ray diffraction, in % (%). These four minerals are the main minerals in this region, and their combined content is close to 1%.
[0155] The commonly used objective function form for optimal interpretation is as follows:
[0156]
[0157] In the formula: x j The content of the j-th stratigraphic component includes clay minerals, quartz, plagioclase, potassium feldspar, and pore fluids, etc.; A ij B is the skeleton response value of the i-th logging curve for the j-th mineral; i Let be the logging value of the i-th logging curve.
[0158] The constraint in this objective function is a rock physics model constraint, namely, the sum of the contents of all formation components is 1. Converting the contents of quartz, potassium feldspar, and plagioclase obtained from X-ray diffraction into the content relationships of formation mineral components yields the new optimization interpretation constraint:
[0159]
[0160] A new optimized interpretation model can calculate the mineral content of granite reservoirs in the northern South China Sea by integrating well logging data and X-ray diffraction experimental data. The introduction of direct geological data (rock cuttings data) will help improve the accuracy of mineral content calculations.
[0161] ②Constraints for establishing logging data
[0162] Figure 3 shows a schematic diagram of the relationship between the contents of quartz, feldspar, and clay in granite buried hills according to an embodiment of the present invention.
[0163] As shown in Figure 3, the analysis of the quartz, feldspar, and clay mineral contents in the X-ray diffraction results of granite buried hill logging in Huizhou and Yongle areas reveals the following relationship:
[0164]
[0165] Similar to the process of establishing constraints using experimental data, the above relationships are transformed into optimization constraints and added to the optimization calculation equations to obtain a new optimization interpretation objective function. By comprehensively integrating well logging information, the mineral composition of the formation can be jointly inverted, thereby improving the calculation accuracy.
[0166] Solving constrained optimization problems in mineral composition optimization inversion. Addressing issues such as the optimization results being heavily influenced by the initial point, susceptibility to local optima, and slow search speed in optimization interpretation, this paper proposes improved or combined optimization algorithms that offer faster global search speeds, higher computational efficiency and accuracy, and more effective solutions to constrained optimization problems through comparative analysis of different optimization algorithms.
[0167] Figure 4 shows a schematic diagram of an LSTM cell structure according to an embodiment of the present invention.
[0168] Lithology identification based on Bidirectional Long Short-Term Memory (Bi-LSTM) network of "mineral composition-lithology" sequence data. As shown in Figure 4, the Bi-LSTM lithology identification model was used to identify the lithology of buried hill reservoirs in the northern South China Sea, and the results were compared with those predicted by classical machine learning algorithms to test and evaluate the identification effect of the Bi-LSTM-based lithology identification model.
[0169] Example 3
[0170] This invention provides a lithology identification method, comprising the following four steps: First, establishing a volumetric model for interpreting mineral components in deep-water buried hill reservoirs through well logging; second, combining constrained optimization algorithms to achieve optimal solutions for overdetermined linear equations with multiple constraints; third, constructing a training sample dataset based on "mineral component-lithology" sequence data; and fourth, utilizing the memory capacity of an attention-enhanced bidirectional long short-term memory network (Attention-BiLSTM) model to achieve lithology identification.
[0171] This invention introduces abstract lithological patterns into the network model training process in a concrete form, improving the reliability of intelligent prediction of complex lithology, and is of great significance for lithology identification and subsequent processing.
[0172] The training sample dataset establishment includes reliability evaluation of the input data for the Attention-BiLSTM model and training sample generation. The input layer of the Attention-BiLSTM model converts the conventional point-to-point mapping into a sequence-to-point mapping. The continuous mineral component content data of a single well obtained through optimized inversion is used as the training sample data. Based on XRD whole-rock analysis of mineral content, evaluation metrics such as mean square error, root mean square error, and mean absolute percentage error are used to conduct reliability analysis on the optimized inverted mineral component content, ensuring the accuracy of the model input data. Normalized inverted mineral component content data are used to form a depth-corresponding sequence dataset. Samples are randomly selected according to a certain ratio of training samples to validation samples to establish training and validation sample sets respectively. The Attention-BiGRU lithology identification model is applied to identify the lithology of buried hill reservoirs in the northern South China Sea. The results are compared with those of classic machine learning algorithms such as support vector machines and random forests to test and evaluate the identification effect of the Attention-BiGRU-based lithology identification model.
[0173] Figure 5 shows a schematic diagram of the lithology identification results of three models according to an embodiment of the present invention.
[0174] A "mineral-lithology" dataset was constructed using seven wells from a certain area. Three models, Bi-LSTM, Bi-GRU, and Attention-BiGRU, were trained on each, as shown in Figure 5. The model accuracies were 0.85, 0.86, and 0.89, respectively. As shown in Table 1, the identification accuracy for the four lithologies, except for diabase, was relatively high. Furthermore, the Attention-BiLSTM model generally achieved higher lithology identification accuracy than the other models.
[0175] Table 1. Accuracy of lithology identification for each model
[0176] Model Accuracy: BI-LSTM 0.85 0.78 0.88 0.96 0.78 0.35 Bi-GRU 0.86 0.93 0.93 0.95 0.79 0.66 Attention-BiGRU 0.89 0.90 0.95 0.98 0.82 0.70 surface
[0177] In summary, this invention can improve lithological identification results, make anomalies more focused to a certain extent, and has a rigorous and reliable theory as well as a simple and practical operation process.
[0178] Example 4
[0179] Figure 6 shows a block diagram of a lithology identification device according to an embodiment of the present invention.
[0180] As shown in Figure 6, the lithology identification device includes:
[0181] Well logging interpretation module 201 establishes a well logging interpretation volume model based on lithological well logging geological classification;
[0182] Inversion module 202 performs optimal inversion of reservoir mineral composition based on the well logging interpretation volume model;
[0183] Training sample construction module 203 constructs a training sample dataset based on "mineral composition-lithology" sequence data;
[0184] The identification module 204 constructs a bidirectional long short-term memory network based on the inversion results to identify lithology.
[0185] In one example, establishing a well logging interpretation volumetric model based on lithology logging geological classification includes:
[0186] Determine the lithology and corresponding minerals of a known area;
[0187] The skeletal lines of each mineral were drawn, and density-sound wave and density-neutron plates were then created;
[0188] A well logging interpretation volume model is established based on density-sonic wave and density-neutron charts.
[0189] In one example, the optimal inversion of reservoir mineral composition based on the well logging interpretation volume model includes:
[0190] Based on the well logging interpretation volume model, establish the well logging response equation;
[0191] Establish a constrained system of overdetermined linear equations.
[0192] In one example, the well logging response equation is:
[0193]
[0194] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component.
[0195] In one example, the overdetermined linear equation system is:
[0196]
[0197] In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component;
[0198] The constraints of the overdetermined linear system of equations are:
[0199]
[0200] In one example, the memory capacity of an attention-enhanced bidirectional long short-term memory network model is used to construct a lithology identification model that considers the variation trend of mineral composition and the correlation between upstream and downstream data.
[0201] The training sample dataset is fed into the lithology identification model for training, and lithology identification is performed based on the trained model.
[0202] Example 5
[0203] This disclosure provides an electronic device, comprising: a memory storing executable instructions; and a processor executing the executable instructions in the memory to implement the aforementioned lithology identification method.
[0204] An electronic device according to an embodiment of the present disclosure includes a memory and a processor.
[0205] This memory is used to store non-transitory computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.
[0206] The processor may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In one embodiment of this disclosure, the processor is used to execute computer-readable instructions stored in the memory.
[0207] Those skilled in the art will understand that, in order to solve the technical problem of how to achieve a good user experience, this embodiment may also include well-known structures such as communication buses and interfaces, and these well-known structures should also be included within the protection scope of this disclosure.
[0208] For a detailed description of this embodiment, please refer to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.
[0209] Example 6
[0210] This disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the lithology identification method.
[0211] A computer-readable storage medium according to embodiments of the present disclosure stores non-transitory computer-readable instructions. When these non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the methods described in the foregoing embodiments of the present disclosure are performed.
[0212] The aforementioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or portable hard drive), media with built-in rewritable non-volatile memory (e.g., memory card), and media with built-in ROM (e.g., ROM cartridge).
[0213] Those skilled in the art should understand that the above description of the embodiments of the present invention is only intended to illustrate the beneficial effects of the embodiments of the present invention, and is not intended to limit the embodiments of the present invention to any of the examples given.
[0214] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.
Claims
1. A lithology identification method, characterized in that, include: Establish a well logging interpretation volume model based on lithological well logging geological classification; Optimization inversion of reservoir mineral composition based on well logging interpretation volume model; A training sample dataset based on "mineral composition-lithology" sequence data is constructed; based on the inversion results, a bidirectional long short-term memory network based on "mineral composition-lithology" sequence data is constructed for lithology identification.
2. The lithology identification method according to claim 1, wherein, Establishing a well logging interpretation volume model based on lithological well logging geological classification includes: determining the lithology and corresponding minerals of a known area; drawing the framework lines of each mineral, and then creating density-sonic and density-neutron charts; and establishing a well logging interpretation volume model based on the density-sonic and density-neutron charts.
3. The lithology identification method according to claim 1, wherein, The optimization inversion of reservoir mineral composition based on the well logging interpretation volume model includes: establishing the well logging response equation based on the well logging interpretation volume model; and establishing a constrained overdetermined linear equation system.
4. The lithology identification method according to claim 3, wherein, The well logging response equation is: In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component.
5. The lithology identification method according to claim 3, wherein, The overdetermined linear equation system is as follows: In the formula, i = 1, 2, ..., m, j = 1, 2, ..., n, where n is the number of components constituting the volume model; x j is the relative content of the j-th component; m is the number of well logging curves selected; B i Let A be the response value of the i-th logging curve to the formation; ij This represents the theoretical value of the i-th logging curve for the j-th component; The constraints of the overdetermined linear system of equations are:
6. The lithology identification method according to claim 1, wherein, By leveraging the memory capacity of an attention-enhanced bidirectional long short-term memory network model, a lithology identification model is constructed that considers the changing trends of mineral composition and the correlation between upstream and downstream data. The training sample dataset is then fed into the lithology identification model for training, and lithology identification is performed based on the trained model.
7. A lithology identification device, characterized in that, include: The well logging interpretation module establishes a well logging interpretation volume model based on lithological well logging geological classification; The inversion module performs optimal inversion of reservoir mineral composition based on the well logging interpretation volume model. The training sample construction module builds a training sample dataset based on the "mineral composition-lithology" sequence data; The identification module constructs a bidirectional long short-term memory network based on the inversion results to identify lithology.
8. The lithology identification device according to claim 7, wherein, Establishing a well logging interpretation volume model based on lithological well logging geological classification includes: determining the lithology and corresponding minerals of a known area; drawing the framework lines of each mineral, and then creating density-sonic and density-neutron charts; and establishing a well logging interpretation volume model based on the density-sonic and density-neutron charts.
9. An electronic device, characterized in that, The electronic device includes: a memory storing executable instructions; and a processor that executes the executable instructions in the memory to implement the lithology identification method according to any one of claims 1-6.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the lithology identification method according to any one of claims 1-6.