A lithology identification method based on geology constraint deep learning and related equipment

Abstract

The application discloses a lithology identification method based on geological constraint deep learning and related equipment, and belongs to the technical field of deep learning. The method comprises the following steps: acquiring logging curves of a plurality of wells in the same work area, and dividing the logging curves into a training set and a test set; performing a geological constraint operation on the training set to obtain training samples; performing a geological constraint operation on the test set to obtain test samples; training a long short-term memory network using the training samples to obtain a long short-term memory network based on geological constraint; testing the accuracy of the long short-term memory network based on geological constraint using the test samples, and if the accuracy meets the requirements, the training is completed; and identifying the lithology of the well using the long short-term memory network based on geological constraint which has completed the training. The application proposes to enhance the robustness of the long short-term memory network by using geological constraint according to the network structure, so that the lithology identification accuracy of the method is improved.

Description

Technical Field

[0001] This invention belongs to the field of deep learning technology and relates to a lithology identification method and related equipment based on geologically constrained deep learning. Background Technology

[0002] Accurate identification of reservoir lithology is crucial for sweet spot characterization and exploration risk assessment. Rock lithology is closely related to its physical parameters and elastic properties. Therefore, to clearly identify rock types, existing technologies have developed direct and indirect geophysical measurement methods. The former involves laboratory measurements of borehole core samples, while the latter involves manual interpretation of rock lithology using multiple well logging curves. However, these methods suffer from drawbacks such as high cost, insufficient core samples, and predictive bias. Currently, neural network algorithms based on deep learning show promise for efficient reservoir lithology identification.

[0003] However, most current neural network algorithms based on deep learning for lithology identification rely on the inherent characteristics of the input parameters, rather than utilizing geological constraints to construct the deep learning network. For example, Ma et al. used principal component analysis to identify lithology in their paper "Lithology identification based on principal component analysis and fuzzy recognition." Although Anazi et al. used deep learning methods for lithology identification in their paper "A support vector machine algorithm to classify lithofacies and model permeability in heterogeneous reservoirs," they assumed that the relationship between lithology and well logging parameters was independent, neglecting the spatial sequence correlation of rocks and diagenesis during deposition. Summary of the Invention

[0004] The purpose of this invention is to solve the technical problem that existing deep learning methods for lithology identification are not based on geological constraints, and to provide a lithology identification method and related equipment based on geological constraint deep learning.

[0005] To achieve the above objectives, the present invention employs the following technical solution:

[0006] In a first aspect, the present invention provides a lithology identification method based on geologically constrained deep learning, characterized by comprising the following steps:

[0007] Obtain the logging curves of several wells within the same work area, and divide the logging curves into a training set and a test set;

[0008] Geological constraint operations are performed on the training set to obtain training samples; geological constraint operations are performed on the test set to obtain test samples.

[0009] A long short-term memory network was trained using training samples to obtain a geologically constrained long short-term memory network.

[0010] The accuracy of the geologically constrained long short-term memory network is tested using test samples. If the accuracy meets the requirements, the training is complete; otherwise, the process returns to the step of performing geological constraint operations on the training set.

[0011] The lithology of well logging is identified using a trained geologically constrained long short-term memory network.

[0012] A further improvement of the present invention is that:

[0013] The logging curves include compensated neutron logging curves, density logging curves, gamma ray logging curves, sonic logging curves, and deep transverse resistivity logging curves.

[0014] The geological constraint operations include rectangular window constraints and Gaussian window constraints; the geological constraint-based long short-term memory network includes a rectangular window-constrained long short-term memory network and a Gaussian window-constrained long short-term memory network.

[0015] The rectangular window-constrained long short-term memory network is established through the following steps:

[0016] A rectangular window constraint operation is performed on the training set, adding a rectangular window to the target depth of the training set to take into account neighbor information, thus obtaining rectangular window training samples; a rectangular window constraint operation is performed on the test set, adding a rectangular window to the target depth of the test set to take into account neighbor information, thus obtaining rectangular window test samples.

[0017] The Long Short-Term Memory (LSTM) network is trained using rectangular window training samples to obtain a rectangular window-constrained LSTM network.

[0018] In the rectangular window training samples, each input sample contains five sampling points with the same weights.

[0019] The Gaussian window-constrained long short-term memory network is established through the following steps:

[0020] Gaussian window constraint operation is performed on the training set to obtain Gaussian window training samples; Gaussian window constraint operation is performed on the test set to obtain Gaussian window test samples.

[0021] Different weight values ​​are assigned to the five sampling points that form the cluster of input samples in the Gaussian window training samples.

[0022] The specific steps for testing the accuracy of the geologically constrained long short-term memory network using test samples are as follows:

[0023] The test samples serve as test data, and a mapping relationship is obtained using a geologically constrained long short-term memory network. The test samples are used to test the accuracy of the geologically constrained long short-term memory network.

[0024] The training set is further subjected to unconstrained operations to obtain unconstrained training samples; the test set is further subjected to unconstrained operations to obtain unconstrained test samples.

[0025] The Long Short-Term Memory (LSTM) network is trained using unconstrained training samples to obtain an unconstrained LSM network.

[0026] Secondly, the present invention provides a lithology identification system based on geologically constrained deep learning, characterized in that it includes:

[0027] The input module acquires the logging curves of several wells within the same work area, and divides the logging curves into a training set and a test set;

[0028] The geological constraint module performs geological constraint operations on the training set to obtain training samples; and performs geological constraint operations on the test set to obtain test samples.

[0029] The training module uses training samples to train the Long Short-Term Memory (LSTM) network, resulting in a geologically constrained LSTM network.

[0030] The testing module uses test samples to test the accuracy of the geologically constrained long short-term memory network. If the accuracy meets the requirements, the training is complete; otherwise, it returns to the step of performing geological constraint operations on the training set.

[0031] The lithology identification module uses a trained geologically constrained long short-term memory network to identify the lithology of well logs.

[0032] Thirdly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described lithology identification method based on geologically constrained deep learning.

[0033] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described lithology identification method based on geologically constrained deep learning.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] This invention discloses a lithology identification method based on geologically constrained deep learning. Regarding the network structure, it proposes enhancing the robustness of the Long Short-Term Memory (LSM) network using geological constraints. When predicting the lithology of rocks at a central depth using a rectangular window-constrained LSM network, only a rectangle needs to be added to the target depth to consider neighborhood information. In the Gaussian window-constrained LSM network (GWLSTM), the five sampling points forming the input sample cluster are assigned different weight values. This is because surrounding sampling points have different influences on the lithology of the current prediction point. A mapping relationship is established between a single label and multiple sampling points. This results in a lithology identification accuracy higher than traditional deep learning-based lithology identification methods. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart of the lithology identification method based on geologically constrained deep learning in this invention;

[0038] Figure 2 This is a flowchart illustrating the principle of a lithology identification method based on geologically constrained deep learning.

[0039] Figure 3 To utilize well data rich in lithological samples as a training set, neural network models were trained, and training graphs for LSTM, RCLSTM, and GWLSTM networks were generated.

[0040] Figure 4 A diagram illustrating the network construction approach that incorporates geological constraints; Figure 4 (a) uses a single sampling point as input; Figure 4 (b) uses multiple sampling points as input; Figure 4 (c) shows Gaussian function plots with different parameter settings;

[0041] Figure 5 This is a lithology identification map of tight sandstone obtained using the lithology identification method based on geologically constrained deep learning in this invention;

[0042] Figure 6 Macro-F1 scores of different models in three well logging operations in this invention are shown.

[0043] Figure 7 This is a system diagram of the lithology identification method based on geologically constrained deep learning in this invention;

[0044] Figure 8 This is a block diagram of the electronic device in this invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0046] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0047] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0048] The present invention will now be described in further detail with reference to the accompanying drawings:

[0049] See Figure 1 This invention proposes a lithology identification method based on geologically constrained deep learning, comprising the following steps:

[0050] S1, Obtain the logging curves of several wells within the same work area, and divide the logging curves into a training set and a test set;

[0051] S2, Perform geological constraint operations on the training set to obtain training samples; perform geological constraint operations on the test set to obtain test samples;

[0052] S3, using training samples to train a long short-term memory network, resulting in a geologically constrained long short-term memory network;

[0053] S4. Use test samples to test the accuracy of the geologically constrained long short-term memory network. If the accuracy meets the requirements, the training is complete; if it does not meet the requirements, return to the step of performing geological constraint operations on the training set.

[0054] S5 uses a trained geologically constrained long short-term memory network to identify the lithology of well logs.

[0055] To address the network structure, a method is proposed to enhance the robustness of the Long Short-Term Memory (LSTM) network by utilizing geological constraints. When predicting rock lithology at a central depth, the rectangular window-constrained LSTM network simply adds a rectangle at the target depth to account for neighboring information. In the Gaussian window-constrained GWLSTM network, the five sampling points forming the input sample cluster are assigned different weight values. This is because surrounding sampling points have varying impacts on the lithology of the current prediction point. A mapping relationship is established between a single label and multiple sampling points. This results in a higher lithology identification accuracy than traditional deep learning-based lithology identification methods.

[0056] The present invention will be described in detail below with reference to specific embodiments:

[0057] Step 1: Selection of logging curves: Based on data availability and input / output sensitivity analysis, five logging curves were selected as inputs for each well, including compensated neutron logging (CNL), density logging (DEN), gamma-ray logging (GR), sonic logging (DT), and deep lateral resistivity logging (RD). Logging curves from five wells within the same experimental area were statistically analyzed as input features. Based on the distribution of input features, the logging curves from two wells with uniform lithological distribution were used as the training set, and the logging curves from the remaining three wells were used as the test set. Figure 3 As shown, the logging curve of one of the training wells is displayed. The logging data is abundant, but there are many thin interlayers and the sample distribution is uneven.

[0058] Step 2: Apply rectangular window and Gaussian window constraints to the input features of the training set to obtain rectangular window training samples and Gaussian window training samples; apply rectangular window and Gaussian window constraints to the input features of the test set to obtain rectangular window test samples and Gaussian window test samples. Perform unconstrained operations on the training set to obtain unconstrained training samples; perform unconstrained operations on the test set to obtain unconstrained test samples.

[0059] Step 3: Train three types of long short-term memory networks, one with geological constraints and one without.

[0060] Step 3.1: Use the unconstrained training samples obtained in Step 2 to train a traditional Long Short-Term Memory (LSTM) network.

[0061] Step 3.2: Train the Long Short-Term Memory (LSTM) network using the rectangular window training samples to obtain the Rectangular Window Constrained LSTM (RCLSTM). During the construction of the RCLSTM, only one rectangular window needs to be added at the target depth to consider neighbor information. In the rectangular window test samples, each input sample contains 5 sampling points with the same weights. A mapping relationship between a single label and multiple sampling points is established.

[0062] Step 3.3 involves training a Long Short-Term Memory (LSTM) network using Gaussian window training samples, resulting in a Gaussian Window Constrained LSTM (GWLSTM). The GWLSTM considers the influence of distance on the target rock lithology. Specifically, the five sampling points forming the input sample cluster in the GWLSTM are assigned different weight values. This is because surrounding sampling points have varying impacts on the lithology of the current prediction point. The closer a point is to the prediction point, the greater its influence.

[0063] like Figure 4 (a) and Figure 4 As shown in (b), in addition to using a single point as input, we can also apply example clusters geologically as input. Figure 4 (c) shows Gaussian functions with different parameter settings. Here, an odd number of sampling points is chosen to maintain the symmetry of the Gaussian window. Furthermore, experimental results show that a small number of sampling points (e.g., 3 sampling points) leads to insufficient constraints. However, too many sampling points (e.g., 7 sampling points) complicate the network and incur expensive training time.

[0064] Step four involves testing the accuracy of the geologically constrained Long Short-Term Memory (LSTM) network using test samples. If the accuracy meets the requirements, training is complete; otherwise, the process returns to the step of applying geological constraints to the training set. The test samples serve as test data, and the geologically constrained LSTM network is used to obtain mapping relationships, verifying the accuracy and superiority of geologically constrained LSTM networks (such as RCLSTM and GWLSTM) in lithology identification. A comparison is then made with traditional LSTM networks.

[0065] Step 5: Use the trained geologically constrained long short-term memory network to identify the lithology of the well logs.

[0066] like Figure 5 As shown in the figure, the rock lithology indicates that the rock lithology column of this well is quite complex. Furthermore, thin layers are frequently observed in these five lithologies. Different geologically constrained long short-term memory (LSTM) models are used to illustrate the predicted lithology of this well and the corresponding differences between the actual and predicted lithologies. In general, the conventional LSTM, rectangular window constrained LSTM, and Gaussian window constrained LSTM models exhibit relatively high accuracy in rock lithology prediction. Compared to the conventional LSTM model, the geologically constrained rectangular window constrained LSTM and Gaussian window constrained LSTM networks achieve even higher accuracy.

[0067] As shown in Table 1, compared with the traditional Long Short-Term Memory (LSTM) network model, the Rectangular Window Constrained LSTM (RCLSTM) and Gaussian Window Constrained LSTM (GWLSTM) models achieve higher accuracy (Table 1). Furthermore, the Rectangular Window Constrained LSTM (RCLSTM) and Gaussian Window Constrained LSTM (GWLSTM) models can more accurately predict calcareous rocks (e.g., Figure 5 (Thin layers at depths of 3380 m and 3510 m). Furthermore, Table 1 reminds us that the lithology prediction accuracy of the Gaussian window-constrained long short-term memory network (GWLSTM) is superior to that of the rectangular window-constrained long short-term memory network (RCLSTM). Overall, geologically constrained long short-term memory networks (e.g., RCLSTM, GWLSTM) are more accurate in lithology prediction than unconstrained traditional long short-term memory networks (LSTM).

[0068] Table 1 shows the accuracy of predicting rock lithology in three wells using long short-term memory network models with different constraints.

[0069]

[0070] The F1 score is a metric used in statistics to measure the accuracy of binary classification models, particularly for imbalanced data. It also considers the accuracy and recall of the classification model. Many variations of the F1 score have been proposed for multi-class classification problems. The Macro-F1 score is widely used to address imbalanced multi-class classification problems because it considers the number of samples in each class in its calculation. Figure 6 As shown, the Macro-F1 scores in three test wells are displayed. The performance of the unconstrained traditional Long Short-Term Memory (LSTM) network model is significantly lower than that of the constrained LSTM network model. This demonstrates the superiority of the geologically constrained LSTM network model.

[0071] See Figure 7 This invention proposes a lithology identification system based on geologically constrained deep learning, comprising:

[0072] The input module acquires the logging curves of several wells within the same work area, and divides the logging curves into a training set and a test set;

[0073] The geological constraint module performs geological constraint operations on the training set to obtain training samples; and performs geological constraint operations on the test set to obtain test samples.

[0074] The training module uses training samples to train the Long Short-Term Memory (LSTM) network, resulting in a geologically constrained LSTM network.

[0075] The testing module uses test samples to test the accuracy of the geologically constrained long short-term memory network. If the accuracy meets the requirements, the training is complete; otherwise, it returns to the step of performing geological constraint operations on the training set.

[0076] The lithology identification module uses a trained geologically constrained long short-term memory network to identify the lithology of well logs.

[0077] See Figure 8 A third objective of this invention is to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the geologically constrained deep learning-based lithology identification method.

[0078] The lithology identification method based on geologically constrained deep learning includes the following steps:

[0079] Obtain the logging curves of several wells within the same work area, and divide the logging curves into a training set and a test set;

[0080] Geological constraint operations are performed on the training set to obtain training samples; geological constraint operations are performed on the test set to obtain test samples.

[0081] A long short-term memory network was trained using training samples to obtain a geologically constrained long short-term memory network.

[0082] The accuracy of the geologically constrained long short-term memory network is tested using test samples. If the accuracy meets the requirements, the training is complete; otherwise, the process returns to the step of performing geological constraint operations on the training set.

[0083] The lithology of well logging is identified using a trained geologically constrained long short-term memory network.

[0084] A fourth objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the lithology identification method based on geologically constrained deep learning.

[0085] The lithology identification method based on geologically constrained deep learning includes the following steps:

[0086] Obtain the logging curves of several wells within the same work area, and divide the logging curves into a training set and a test set;

[0087] Geological constraint operations are performed on the training set to obtain training samples; geological constraint operations are performed on the test set to obtain test samples.

[0088] A long short-term memory network was trained using training samples to obtain a geologically constrained long short-term memory network.

[0089] The accuracy of the geologically constrained long short-term memory network is tested using test samples. If the accuracy meets the requirements, the training is complete; otherwise, the process returns to the step of performing geological constraint operations on the training set.

[0090] The lithology of well logging is identified using a trained geologically constrained long short-term memory network.

[0091] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0093] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0094] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps for the functions specified in one or more boxes, such as Figure 8 As shown.

[0095] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A lithology identification method based on geologically constrained deep learning, characterized in that, Includes the following steps: Obtain the logging curves of several wells within the same work area, and divide the logging curves into a training set and a test set; Geological constraints are applied to the training set to obtain training samples; geological constraints are also applied to the test set to obtain test samples; the geological constraints include rectangular window constraints and Gaussian window constraints; the geologically constrained long short-term memory network includes a rectangular window-constrained long short-term memory network and a Gaussian window-constrained long short-term memory network; the rectangular window-constrained long short-term memory network is established through the following steps: A rectangular window constraint operation is performed on the training set, adding a rectangular window to the target depth of the training set to take into account neighbor information, thus obtaining rectangular window training samples; a rectangular window constraint operation is performed on the test set, adding a rectangular window to the target depth of the test set to take into account neighbor information, thus obtaining rectangular window test samples. The Long Short-Term Memory (LSTM) network is trained using rectangular window training samples to obtain a rectangular window-constrained LSTM network. In the rectangular window training samples, each input sample contains five sampling points with the same weights; The Gaussian window-constrained long short-term memory network is established through the following steps: Gaussian window constraint operation is performed on the training set to obtain Gaussian window training samples; Gaussian window constraint operation is performed on the test set to obtain Gaussian window test samples. Different weight values ​​are assigned to the five sampling points that form the cluster of input samples in the Gaussian window training samples; A long short-term memory network was trained using training samples to obtain a geologically constrained long short-term memory network. The accuracy of the geologically constrained long short-term memory network is tested using test samples. If the accuracy meets the requirements, the training is complete; otherwise, the process returns to the step of performing geological constraint operations on the training set. The lithology of well logging is identified using a trained geologically constrained long short-term memory network.

2. The lithology identification method based on geologically constrained deep learning according to claim 1, characterized in that, The logging curves include compensated neutron logging curves, density logging curves, gamma ray logging curves, sonic logging curves, and deep transverse resistivity logging curves.

3. The lithology identification method based on geologically constrained deep learning according to claim 1, characterized in that, The specific steps for testing the accuracy of the geologically constrained long short-term memory network using test samples are as follows: The test samples serve as test data, and a mapping relationship is obtained using a geologically constrained long short-term memory network. The test samples are used to test the accuracy of the geologically constrained long short-term memory network.

4. The lithology identification method based on geologically constrained deep learning according to claim 3, characterized in that, The training set is further subjected to unconstrained operations to obtain unconstrained training samples; the test set is further subjected to unconstrained operations to obtain unconstrained test samples. The Long Short-Term Memory (LSTM) network is trained using unconstrained training samples to obtain an unconstrained LSM network.

5. A lithology identification system based on geologically constrained deep learning, characterized in that, include: The input module acquires the logging curves of several wells within the same work area, and divides the logging curves into a training set and a test set; The geological constraint module performs geological constraint operations on the training set to obtain training samples; it also performs geological constraint operations on the test set to obtain test samples. The geological constraint operations include rectangular window constraints and Gaussian window constraints. The geological constraint-based Long Short-Term Memory (LSTM) network includes both rectangular window-constrained LSTM networks and Gaussian window-constrained LSTM networks. The rectangular window-constrained LSTM network is established through the following steps: A rectangular window constraint operation is performed on the training set, adding a rectangular window to the target depth of the training set to take into account neighbor information, thus obtaining rectangular window training samples; a rectangular window constraint operation is performed on the test set, adding a rectangular window to the target depth of the test set to take into account neighbor information, thus obtaining rectangular window test samples. The Long Short-Term Memory (LSTM) network is trained using rectangular window training samples to obtain a rectangular window-constrained LSTM network. In the rectangular window training samples, each input sample contains five sampling points with the same weights; The Gaussian window-constrained long short-term memory network is established through the following steps: Gaussian window constraint operation is performed on the training set to obtain Gaussian window training samples; Gaussian window constraint operation is performed on the test set to obtain Gaussian window test samples. Different weight values ​​are assigned to the five sampling points that form the cluster of input samples in the Gaussian window training samples; The training module uses training samples to train the Long Short-Term Memory (LSTM) network, resulting in a geologically constrained LSTM network. The testing module uses test samples to test the accuracy of the geologically constrained long short-term memory network. If the accuracy meets the requirements, the training is complete; otherwise, it returns to the step of performing geological constraint operations on the training set. The lithology identification module uses a trained geologically constrained long short-term memory network to identify the lithology of well logs.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the lithology identification method based on geologically constrained deep learning as described in any one of claims 1-4.

7. A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the lithology identification method based on geologically constrained deep learning as described in any one of claims 1-4.

Images