A seismic-scale identification and planar distribution prediction method for shale lithofacies assemblages

Abstract

This invention discloses a method for seismic-scale identification and planar distribution prediction of shale lithofacies assemblages. The method includes: extracting geological characteristics of the target layer from drilling and conducting seismic forward modeling; determining the minimum thickness of identifiable lithofacies assemblages based on the simulation results; classifying lithofacies assemblages using geological data from drilling, logging, and analysis tests to obtain identifiable lithofacies assemblage types; extracting well-perimeter seismic attributes to obtain the seismic attributes corresponding to seismically identifiable lithofacies assemblage units; performing sensitivity analysis on the seismic attributes to determine seismic attributes that meet preset conditions; constructing a machine learning dataset using seismic attributes and lithofacies assemblage types; training the machine learning model to obtain a trained machine learning model; extracting seismic attributes of the target layer to create a prediction set; inputting this set into the trained machine learning model to obtain a planar distribution map of the lithofacies assemblages. This invention achieves low-cost, efficient, and accurate prediction of the planar distribution of shale lithofacies assemblages.

Description

Technical Field

[0001] This invention belongs to the field of shale oil and gas resource exploration technology, and is a method for seismic-scale identification and planar distribution prediction of shale lithofacies assemblages. Background Technology

[0002] Shale oil and gas, as an important unconventional oil and gas resource, shows broad prospects for exploration and development. The spatial development characteristics of shale lithofacies assemblages are a key geological factor influencing the distribution of shale oil and gas "sweet spots." Current research methods for shale lithofacies assemblages include identification based on core or facies logging data and machine learning prediction based on logging data. Existing methods mainly focus on the identification and prediction of shale lithofacies assemblages at single well points. Planar identification relies primarily on inferences between single well points, and its accuracy is affected by the density and spacing of well points. Seismic data provides the possibility for the planar distribution of shale lithofacies assemblages, but the scale of lithofacies assemblages identifiable by seismic data is currently unclear, limiting the seismic scale identification of shale lithofacies assemblages. Therefore, it is urgent to address the seismic identification scale of shale lithofacies assemblages and combine machine learning techniques to conduct efficient and accurate predictions of the planar distribution of shale lithofacies assemblages, providing strong technical support for research on the distribution of favorable shale oil and gas areas. Summary of the Invention

[0003] This invention addresses existing technical challenges by providing a method for seismic-scale identification and planar distribution prediction of shale lithofacies assemblages. The method specifically includes:

[0004] Step S1: Extract the geological characteristics of the target layer for drilling, conduct seismic forward modeling based on the geological characteristics, and determine the minimum thickness of the seismically identifiable lithofacies assemblages based on the simulation results;

[0005] Step S2: Combine geological data from drilling, logging, and analysis tests to classify lithofacies assemblages, and obtain seismically identifiable lithofacies assemblages based on the classification results;

[0006] Step S3: Extract well perimeter seismic attributes to obtain the seismic attributes corresponding to seismic-scale identifiable lithofacies assemblage units;

[0007] Step S4 uses the mRMR algorithm to perform sensitivity analysis on the seismic attributes and determine the seismic attributes that meet the preset conditions;

[0008] Step S5: Construct a machine learning dataset based on the seismic attributes that meet the preset conditions and the lithofacies combination type, and train the machine learning model based on the dataset to obtain a trained machine learning model.

[0009] Step S6: Extract the seismic attributes of the target layer that meet the preset conditions to create a prediction set, input it into the trained machine learning model, and obtain the planar distribution map of the lithofacies assemblage.

[0010] Optionally, in step S1, determining the minimum thickness of the identifiable lithofacies assemblages based on the simulation results specifically includes:

[0011] Based on the thickness and P-wave velocity of the target layer revealed by drilling, an initial seismic forward model was designed.

[0012] In the initial seismic forward model, the thickness of a single layer of the experimental layer's anomalous velocity geological body is set from large to small, and forward modeling experiments are carried out at a preset excitation frequency.

[0013] During the forward modeling experiment, the seismic reflection characteristics of geological bodies with different single-layer thicknesses and abnormal velocities in the experimental layer were compared to determine the smallest thickness unit that can be identified at the seismic scale.

[0014] Optionally, the initial seismic forward model is set to a background layer-experimental layer-background layer pattern;

[0015] The background layer velocity value was determined based on the measured formation velocity in the study area.

[0016] The experimental layer serves as a velocity model variation unit to simulate velocity anomalies caused by different lithofacies combinations in actual strata.

[0017] Optionally, in step S2, the seismically identifiable lithofacies assemblage types include:

[0018] Thick carbonate rock assemblages; thick carbonate rock assemblages interbedded with thin migmatite assemblages; thin migmatite assemblages interbedded with thick carbonate rock assemblages; alternating layers of carbonate rock assemblages and migmatite assemblages; thick migmatite assemblages interbedded with thin carbonate rock assemblages; thick carbonate rock assemblages interbedded with migmatite assemblages and siltstone assemblages; alternating layers of carbonate rock assemblages and migmatite assemblages interbedded with thin siltstone assemblages and thick migmatite assemblages interbedded with carbonate rock assemblages and siltstone assemblages.

[0019] Optionally, in step S3, the seismic properties corresponding to the lithofacies assemblage unit include:

[0020] Root mean square amplitude, amplitude kurtosis, amplitude squared difference, absolute amplitude sum, 90-degree phase, instantaneous phase, instantaneous bandwidth, instantaneous frequency, total energy, sweet spot.

[0021] Optionally, in step S4, determining the seismic attributes that meet the preset conditions specifically includes:

[0022] The mRMR algorithm was used to calculate the sensitivity of seismic attributes to the response of different lithofacies assemblage types. Based on the sensitivity analysis results, the seismic attribute types were sorted.

[0023] Based on the sorting results, earthquake attributes are selected from the earthquake attribute types as those that meet preset conditions.

[0024] Optionally, the seismic attributes that satisfy the preset conditions include:

[0025] Instantaneous bandwidth, instantaneous frequency, amplitude peak state, instantaneous phase, 90-degree phase, sweet spot, and amplitude squared difference properties.

[0026] Optionally, in step S6, the process of obtaining the planar distribution map of the lithofacies assemblage specifically involves:

[0027] The seismic data of the target layer in the research area are divided into layers of equal thickness according to the minimum thickness of the lithofacies assemblage that can be identified at the seismic scale. Seismic attribute data that meet the preset conditions are extracted for each layer according to the coordinate points. Based on the seismic attribute data that meet the preset conditions, a coordinate-based seismic attribute plane prediction set is made.

[0028] Seismic attribute data that meets the preset conditions are input into a trained machine learning model to obtain the lithofacies combination type corresponding to each coordinate point. A planar distribution map of the lithofacies combination type is then drawn using a programming language according to the coordinates and lithofacies combination type.

[0029] The beneficial effects of this invention are as follows:

[0030] This invention identifies the smallest seismically identifiable shale lithofacies assemblages through seismic forward modeling experiments. It then divides the target wells in the study area into lithofacies assemblages and extracts well-perimeter attributes. Seismic attribute sensitivity analysis is performed using the mRMR algorithm to select and optimize seismic attributes, constructing a dataset of lithofacies assemblages and optimized seismic attributes. A random forest algorithm is used to train and test a machine learning model, which is then applied to predict the planar distribution of shale lithofacies assemblages. The method provided by this invention effectively solves existing technical problems, identifies the smallest seismically identifiable shale lithofacies assemblages, and achieves low-cost, efficient, and accurate prediction of the planar distribution of shale lithofacies assemblages, providing strong technical support for the exploration and development of shale oil and gas. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart illustrating a method for seismic-scale identification and planar distribution prediction of shale lithofacies assemblages according to an embodiment of the present invention.

[0033] Figure 2 This is a schematic diagram of the initial model for earthquake forward modeling according to an embodiment of the present invention;

[0034] Figure 3 This is a schematic diagram of the seismic forward modeling process according to an embodiment of the present invention;

[0035] Figure 4 This is a diagram of seismic-scale shale lithofacies assemblage types according to an embodiment of the present invention, wherein, Figure 4 (a) is a schematic diagram of the type of thick carbonate rock assemblage. Figure 4 (b) is a schematic diagram of a thick carbonate rock assemblage interbedded with a thin migmatite assemblage. Figure 4 (c) is a schematic diagram of a thin-layered migmatite assemblage interbedded with a thick-layered carbonate rock assemblage. Figure 4 (d) is a schematic diagram of the interbedded combination of carbonate rock assemblages and migmatite assemblages. Figure 4 (e) is a schematic diagram of a thick-layered migmatite assemblage interbedded with a thin-layered carbonate assemblage. Figure 4 (f) is a schematic diagram of thick-layered carbonate rock assemblages interbedded with migmatite and siltstone assemblages. Figure 4 (g) is a schematic diagram of an interbedded carbonate rock assemblage and a migmatite assemblage with thin layers of siltstone. Figure 4 (h) is a schematic diagram of the types of thick-layered mixed rock assemblages interbedded with carbonate rock assemblages and siltstone assemblages;

[0036] Figure 5 This is a ranking diagram of seismic attribute sensitivity based on the mRMR algorithm according to an embodiment of the present invention;

[0037] Figure 6 This is a planar distribution diagram of the lithofacies assemblage according to an embodiment of the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] This embodiment provides a method for seismic-scale identification and planar distribution prediction of shale lithofacies assemblages, such as... Figure 1 As shown, it specifically includes:

[0041] Step S1: Extract the geological characteristics of the target layer for drilling, conduct seismic forward modeling based on the geological characteristics, and determine the minimum thickness of the seismically identifiable lithofacies assemblages based on the simulation results.

[0042] Based on the characteristics of the target layer revealed by drilling, such as thickness and P-wave velocity, an initial seismic forward model was designed, and Tesseral software was selected for forward modeling. Figure 3 As shown, the initial model has a length of 2000m and a width of 1000m. The initial model consists of three parts: a background layer, an experimental layer, and another background layer. The initial model is as follows: Figure 2 As shown, the thicknesses are 400m, 200m and 400m respectively. The velocity values ​​of the background layer are determined based on the measured strata velocities in the study area. The experimental layer is used as a velocity model variation unit to simulate velocity anomalies caused by different lithofacies combinations in the actual strata.

[0043] In this embodiment, the minimum thickness is "the minimum thickness of identifiable lithofacies assemblages under the resolution conditions of existing seismic data quality control".

[0044] Regarding the drilling mentioned in this embodiment: Drilling is a systematic project that includes well site deployment, foundation construction, tripping, electrical logging (well logging), cementing operations, and other processes. The logging data obtained during the drilling process, combined with the geological background, can determine information such as the depth and thickness of the target layer.

[0045] like Figure 3 As shown, the single-layer thickness of the anomalous velocity geological bodies in the experimental layer was set from large to small, and forward modeling experiments were carried out at an excitation frequency of 45 Hz. The seismic reflection characteristics of anomalous velocity geological bodies with different single-layer thicknesses in the experimental layer were compared to determine the smallest identifiable thickness unit at the seismic scale. The smallest identifiable thickness unit is 10-15 m.

[0046] Step S2: Combine geological data from drilling, logging, and analysis tests to classify lithofacies assemblages, and obtain seismically identifiable lithofacies assemblages based on the classification results.

[0047] like Figure 4 (a)- Figure 4 As shown in (h), seismically identifiable lithofacies assemblages include thick carbonate rock assemblages, thick carbonate rock assemblages interbedded with thin migmatite assemblages, thin migmatite assemblages interbedded with thick carbonate rock assemblages, alternating layers of carbonate rock assemblages and migmatite assemblages, thick migmatite assemblages interbedded with thin carbonate rock assemblages, thick carbonate rock assemblages interbedded with migmatite assemblages, siltstone assemblages, alternating layers of carbonate rock assemblages and migmatite assemblages interbedded with thin siltstone assemblages, and thick migmatite assemblages interbedded with carbonate rock assemblages and siltstone assemblages.

[0048] Step S3: Extract the well perimeter seismic attributes to obtain the seismic attributes corresponding to the seismic facies assemblage units that can be identified at the seismic scale.

[0049] The method for obtaining well perimeter seismic attributes is as follows: Jason software is used to collect regional well seismic data and establish a software research area. After time-depth conversion and well seismic calibration, the seismic attributes of the well perimeter can be directly extracted using the software's basic functions, and the data can be exported for later data processing and the creation of related training sets.

[0050] Earthquake properties include root mean square amplitude, amplitude kurtosis, amplitude squared difference, absolute amplitude sum, 90-degree phase, instantaneous phase, instantaneous bandwidth, instantaneous frequency, total energy, and sweet spot.

[0051] Step S4: Use the mRMR algorithm to perform sensitivity analysis on the seismic attributes and determine the seismic attributes that meet the preset conditions.

[0052] The mRMR algorithm was used to calculate the sensitivity of different types of seismic attributes to lithofacies assemblages, such as... Figure 5 As shown, earthquake attribute types are sorted according to importance scores; earthquake attributes with sensitivity parameters greater than 0.08 are selected as preferred earthquake attributes, including instantaneous bandwidth, instantaneous frequency, amplitude kurtosis, instantaneous phase, 90-degree phase, sweet spot, and amplitude squared difference attributes.

[0053] Step S5: Construct a machine learning dataset based on the earthquake attributes that meet the preset conditions and the lithofacies combination type, and train the machine learning model based on the dataset to obtain a trained machine learning model.

[0054] Specifically, a machine learning model was built using the random forest algorithm. The dataset for machine learning was constructed with the preferred seismic attributes as feature values ​​and the lithofacies as labels. The dataset contained a total of 86 sets of data.

[0055] The dataset was divided into a training set and a test set in a 7:3 ratio. The training set contained 60 sets of data for training the machine learning model, while the test set contained 26 sets of data for accuracy validation. The model's test accuracy was 84.6%.

[0056] Step S6: Extract the seismic attributes of the target layer that meet the preset conditions to create a prediction set, input it into the trained machine learning model, and obtain the planar distribution map of the lithofacies assemblage.

[0057] The seismic data volume of the target layer in the research area is divided into layers of equal thickness according to the smallest identifiable thickness unit at the seismic scale. The preferred seismic attribute data of each layer is extracted according to the coordinate points, and a coordinate-based seismic attribute plane prediction set is created.

[0058] The selected seismic attribute data is input into the machine learning model described in step S5 to obtain the lithofacies combination type corresponding to each coordinate point. This is done using the pandas, numpy, and matplotlib libraries of the Python programming language, such as... Figure 6 As shown, a planar distribution map of lithofacies assemblages is drawn according to coordinates and lithofacies assemblages.

[0059] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for seismic scale identification and planar distribution prediction of shale facies assemblages, characterized in that, Specifically comprising the following steps: Step S1, extracting the geological characteristics of the target layer of the well, carrying out seismic forward modeling based on the geological characteristics, and determining the minimum thickness of the seismic scale recognizable lithofacies combination based on the simulation results; Step S2, combining the geological data of drilling, logging and analysis testing, dividing the lithofacies combination, and obtaining the seismic scale recognizable lithofacies combination type based on the division results; Step S3, extracting the seismic attributes around the well to obtain the seismic attributes corresponding to the seismic scale recognizable lithofacies combination unit; Step S4, using the mRMR algorithm to perform sensitivity analysis on the seismic attributes to determine the seismic attributes meeting the preset conditions; Step S5, constructing a data set for machine learning based on the seismic attributes meeting the preset conditions and the lithofacies combination type, training the machine learning model based on the data set, and obtaining the trained machine learning model; Step S6, extracting the seismic attributes of the target layer meeting the preset conditions to make a prediction set, inputting the trained machine learning model, and obtaining the planar distribution map of the lithofacies combination.

2. The method according to claim 1, wherein, In step S1, the content of determining the minimum thickness of the recognizable lithofacies combination based on the simulation results specifically comprises: Based on the thickness and P-wave velocity of the target layer revealed by drilling, an initial model of seismic forward modeling is designed; In the initial model of seismic forward modeling, the single-layer thickness of the abnormal velocity geological body in the experimental layer is set, and the setting method is from large to small, and the forward modeling experiment is carried out under the preset excitation frequency; During the forward modeling experiment, the seismic reflection characteristics of the abnormal velocity geological body with different single-layer thicknesses in the experimental layer are compared to determine the minimum thickness unit recognizable in the seismic scale.

3. The method according to claim 2, wherein the method further comprises: The initial model of seismic forward modeling is set as a mode of background layer-experimental layer-background layer; The background layer velocity value is determined according to the measured stratum velocity in the research area; The experimental layer is used as a velocity model change unit to simulate the velocity anomaly body caused by different lithofacies combination stacking types in the actual stratum.

4. The method according to claim 1, wherein, In step S2, the seismic scale recognizable lithofacies combination type includes: Thick carbonate rock combination, thick carbonate rock combination with thin mixed rock combination, thin mixed rock combination with thick carbonate rock combination, interbedded carbonate rock combination and mixed rock combination, thick mixed rock combination with thin carbonate combination, thick carbonate rock combination with mixed rock combination, siltstone combination, interbedded carbonate rock combination and mixed rock combination with thin siltstone combination, and thick mixed rock combination with carbonate rock combination and siltstone combination.

5. The method according to claim 1, wherein, In step S3, the seismic attributes corresponding to the lithofacies combination unit include: Root mean square amplitude, amplitude kurtosis, amplitude square difference, absolute amplitude sum, 90-degree phase, instantaneous phase, instantaneous bandwidth, instantaneous frequency, total energy, and sweet spot.

6. The method according to claim 1, wherein, In step S4, the content of determining the seismic attributes meeting the preset conditions specifically comprises: Using the mRMR algorithm to calculate the sensitivity of the seismic attributes of different lithofacies combination types to the lithofacies combination response, and sorting the seismic attribute types based on the sensitivity analysis results; Based on the sorting results, the seismic attributes are selected from the seismic attribute types as the seismic attributes meeting the preset conditions.

7. The method according to claim 6, wherein, The seismic attributes meeting the preset conditions include: Instantaneous bandwidth, instantaneous frequency, amplitude peak, instantaneous phase, 90 degree phase, sweet spot, and amplitude square difference attributes.

8. The method according to claim 1, wherein, In the step S6, the process of obtaining the planar distribution map of the lithofacies assemblage is specifically: According to the minimum thickness of the lithofacies assemblage identifiable in the seismic scale, the seismic data of the target area of the target layer is divided into layers with equal thickness, the seismic attribute data of each layer meeting the preset condition is extracted according to the coordinate points, and the seismic attribute planar prediction set based on the coordinates is made based on the seismic attribute data meeting the preset condition; The seismic attribute data meeting the preset condition is input into the trained machine learning model, the lithofacies assemblage type corresponding to each coordinate point is obtained, and the planar distribution map of the lithofacies assemblage type is drawn according to the coordinates and the lithofacies assemblage type by using a programming language.

Images