Model training method, stratum lithology identification method, medium and product

Abstract

The invention provides a model training method, a stratum lithology identification method, a medium and a product. The method comprises the steps that a sample object set is determined, the sample object set comprises sample objects and attribute combinations and stratum lithology of the sample objects, and the attribute combinations at least comprise two different geophysical attributes; constructing coordinate systems of different attribute combinations according to the types of the geophysical attributes in the attribute combinations; according to the positions of the sample objects in coordinate systems of different attribute combinations, a target coordinate system and a target attribute combination corresponding to the target coordinate system are determined, and the sample objects of the same stratum lithology in the target coordinate system meet a preset aggregation requirement; and according to the target attribute combination of the sample object and the stratum lithology corresponding to the target attribute combination, training the to-be-trained model to obtain a target model. According to the method, the accuracy of stratum lithology prediction is improved.

Description

Technical Field

[0001] This application relates to the field of geophysical exploration, and particularly to a model training method, a formation lithology identification method, a medium and a product. Background Art

[0002] When extracting coalbed methane, it is necessary to identify the formation lithology in the geological structure to find the coal seam location. Identifying the formation lithology can start from geophysical properties, analyze the relationship between geophysical properties and formation lithology, and thus determine the formation lithology of the geological structure.

[0003] Currently, when analyzing the relationship between geophysical properties and formation lithology, usually geophysical properties are simply selected as the input of the identification model for training the model, so as to identify the formation lithology.

[0004] However, the existing identification model has the problem that the selection criteria of the input attributes during training are not clear, resulting in low prediction accuracy of the formation lithology. Summary of the Invention

[0005] This application provides a model training method, a formation lithology identification method, a medium and a product, which are used to solve the problem that the existing identification model has unclear selection criteria of input attributes during training, resulting in low prediction accuracy of the formation lithology.

[0006] In a first aspect, an embodiment of this application provides a model training method, which is applied to a geological model and includes:

[0007] Determine a sample object set, where the sample object set includes sample objects, as well as the attribute combinations and formation lithologies of the sample objects. The attribute combinations include at least two different geophysical properties;

[0008] According to the types of geophysical properties in the attribute combinations, construct coordinate systems for different attribute combinations;

[0009] According to the positions of the sample objects in the coordinate systems of different attribute combinations, determine a target coordinate system and a target attribute combination corresponding to the target coordinate system. Among the sample objects with the same formation lithology in the target coordinate system, the preset aggregation requirements are satisfied;

[0010] Train the model to be trained according to the target attribute combination of the sample objects and the formation lithology corresponding to the target attribute combination to obtain a target model.

[0011] In a possible implementation manner, determining the sample object set includes:

[0012] Obtain the geophysical information of the sample objects and the formation lithology data corresponding to the geophysical information, where the geophysical information includes at least one of seismic data, logging data, and drilling data;

[0013] Combined processing is performed on at least two geophysical attributes in seismic data, logging data, and drilling data to obtain an attribute combination of a sample object;

[0014] Based on the formation lithology data, the formation lithology corresponding to the attribute combination of the sample object is determined;

[0015] Based on the attribute combination of the sample object and the formation lithology, a sample object set is determined.

[0016] In a possible implementation manner, when the attribute combination includes two different geophysical attributes,

[0017] According to the types of geophysical attributes in the attribute combination, coordinate systems for different attribute combinations are constructed, including:

[0018] Taking the type of one geophysical attribute in the attribute combination as the x-axis of the coordinate system and the type of the other geophysical attribute in the attribute combination as the y-axis of the coordinate system, coordinate systems for different attribute combinations are constructed.

[0019] In a possible implementation manner, according to the positions of the sample object in the coordinate systems of different attribute combinations, a target coordinate system and a target attribute combination corresponding to the target coordinate system are determined, including:

[0020] According to the positions of the sample object in the coordinate systems of different attribute combinations, the aggregation degree of sample objects of different formation lithologies in the coordinate system is determined;

[0021] According to the aggregation degree of sample objects of different formation lithologies in the coordinate system, sample objects that meet the aggregation requirements are determined;

[0022] According to the sample objects that meet the aggregation requirements, a target coordinate system and a target attribute combination corresponding to the target coordinate system are determined.

[0023] In a possible implementation manner, according to the positions of the sample object in the coordinate systems of different attribute combinations, the aggregation degree of sample objects of different formation lithologies in the coordinate system is determined, including:

[0024] According to the positions of the sample object in the coordinate systems of different attribute combinations, the distances between sample objects in the coordinate systems of different attribute combinations are determined;

[0025] According to the distances between sample objects in the coordinate systems of different attribute combinations, the distance levels of sample objects on the coordinate systems of different attribute combinations are determined;

[0026] According to the distance levels and the positions of the sample object in the coordinate systems of different attribute combinations, the aggregation degree of sample objects of different formation lithologies in the coordinate system is determined.

[0027] In a possible implementation manner, training a model to be trained according to the target attribute combination of a sample object and the formation lithology corresponding to the target attribute combination to obtain a target model, including:

[0028] Inputting all attributes in the target attribute combination of the sample object into the model to be trained to obtain a predicted formation lithology;

[0029] Obtaining a cross-entropy loss and a similarity difference loss according to the predicted formation lithology and the formation lithology;

[0030] Determining a target loss according to the cross-entropy loss and the similarity loss;

[0031] Adjusting the model to be trained according to the target loss to obtain a target model.

[0032] In a possible implementation manner, determining a target loss according to the cross-entropy loss and the similarity loss, including:

[0033] Obtaining a target similarity loss according to the similarity loss and a preset coefficient, where the preset coefficient is determined according to the proportion of the number of sample objects with different formation lithologies in the sample object set, and different preset coefficients correspond to different proportions;

[0034] Determining a target loss according to the cross-entropy loss and the target similarity loss.

[0035] In a second aspect, an embodiment of the present application provides a method for identifying formation lithology, including:

[0036] Obtaining geophysical attributes of a sample in a target area;

[0037] Inputting the geophysical attributes into the target model to obtain the target formation lithology of the sample in the target area.

[0038] In a third aspect, an embodiment of the present application provides a model training device, including:

[0039] A first determination module, configured to determine a sample object set, where the sample object set includes a sample object, as well as an attribute combination and a formation lithology of the sample object, and the attribute combination includes at least two different geophysical attributes;

[0040] A processing module, configured to construct coordinate systems of different attribute combinations according to the types of geophysical attributes in the attribute combination;

[0041] A second determination module, configured to determine a target coordinate system and a target attribute combination corresponding to the target coordinate system according to the position of the sample object in the coordinate systems of different attribute combinations;

[0042] A training module for training a model to be trained based on the target attribute combination of a sample object and the formation lithology corresponding to the target attribute combination to obtain a target model.

[0043] In a fourth aspect, an embodiment of the present application provides a formation lithology identification device, including:

[0044] An acquisition module for acquiring geophysical attributes of a sample in a target area;

[0045] A obtaining module for inputting the geophysical attributes into the target model to obtain the target formation lithology of the sample in the target area.

[0046] In a fifth aspect, an embodiment of the present application provides an electronic device, including: a memory, a processor;

[0047] The memory stores computer-executable instructions;

[0048] The processor executes the computer-executable instructions stored in the memory to implement the method of the present application.

[0049] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored, and when the computer-executable instructions are executed by a processor, they are used to implement the first aspect and / or various possible implementation manners of the first aspect as described above.

[0050] In a seventh aspect, an embodiment of the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the method of the embodiment of the present application.

[0051] A model training method, a formation lithology identification method, a medium and a product provided by an embodiment of the present application. By determining a sample object set including sample objects, as well as the attribute combinations of the sample objects and the formation lithology, and using at least two different types of geophysical attributes included in the attribute combinations as the coordinate axes for constructing a coordinate system, a coordinate system corresponding to different attribute combinations is constructed. Then, according to the positions of the sample objects in each coordinate system and the formation lithology corresponding to the sample objects, sample objects that meet the aggregation requirements for the same formation lithology in each coordinate system are determined. Based on this, the target coordinate system in the coordinate system and the coordinate axes representing the target geophysical attribute types on the target coordinate system are determined. The target geophysical attributes and the corresponding formation lithology are used as the training data for the model to be trained, and the model to be trained is trained, and then the target model is obtained by technical means. By analyzing the positions of the sample objects in different coordinate systems, the target geophysical attributes related to the formation lithology can be quickly and accurately determined, so that the recognition effect of the target model trained using the target geophysical attributes and the formation lithology is better, and when the target model executes the formation lithology identification method, the target formation lithology of the sample in the target area is more accurate. Therefore, the accuracy of the prediction of the formation lithology is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0053] Figure 1 A schematic diagram of the scenario of a model training method and a formation lithology identification method provided by an embodiment of the present application;

[0054] Figure 2 A flowchart of a model training method provided by an embodiment of the present application Figure 1 ;

[0055] Figure 3 A flowchart of a model training method provided by an embodiment of the present application Figure 2 ;

[0056] Figure 4 An overall framework diagram of a model training method provided by an embodiment of the present application;

[0057] Figure 5 A result diagram of a model training method provided by an embodiment of the present application;

[0058] Figure 6 A flowchart of the formation lithology identification method provided by an embodiment of the present application;

[0059] Figure 6a A result diagram of the formation lithology identification method provided by an embodiment of the present applicationFigure 1 ;

[0060] Figure 6b Schematic diagram of the result of the formation lithology identification method provided by the embodiment of the present application Figure 2 ;

[0061] Figure 6c Schematic diagram of the result of the formation lithology identification method provided by the embodiment of the present application Figure 3 ;

[0062] Figure 6d Schematic diagram of the result of the formation lithology identification method provided by the embodiment of the present application Figure 4 ;

[0063] Figure 6e Schematic diagram of the result of the formation lithology identification method provided by the embodiment of the present application Figure 5 ;

[0064] Figure 6f Schematic diagram of the result of the formation lithology identification method provided by the embodiment of the present application Figure 6 ;

[0065] Figure 7 Schematic diagram of the structure of the model training device provided by the present application;

[0066] Figure 8 Schematic diagram of the structure of the formation lithology identification device provided by the present application;

[0067] Figure 9 Schematic diagram of the structure of the electronic device provided by the present application.

[0068] Through the above-mentioned drawings, specific embodiments of the present application have been shown, and more detailed descriptions will be given later. These drawings and written descriptions are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. Detailed Description of the Invention

[0069] Here, exemplary embodiments will be described in detail, and examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0070] First, the terms involved in the present application are explained:

[0071] Lithology: Refers to the physical and chemical properties of rocks, which describe characteristics such as the composition, structure, and texture of rocks. Lithology information is very important for determining rock types, sedimentary environments, diagenetic processes, and their potential uses;

[0072] Seismic data: Refers to data obtained through artificial seismic exploration technology. This technology generates seismic waves by setting up seismic sources on the ground or under the sea, and then uses receivers distributed at different positions to record the reflection or refraction of these seismic waves when they propagate underground. Seismic data mainly includes parameters such as seismic profiles, velocity models, amplitudes, and phases;

[0073] Well logging data: Refers to a series of information about the physical properties of rocks around the wellbore measured by specialized instruments in boreholes. Well logging technology can provide various types of curves, and each curve represents different physical properties, such as resistivity, natural gamma ray, acoustic travel time, density, neutron porosity, etc.;

[0074] Drilling data: Refers to various operating parameters and technical indicators collected during the drilling process, including but not limited to drilling rate, weight on bit, torque, drilling fluid properties (such as density, viscosity), temperature, pressure, gas detection, cuttings analysis, etc.;

[0075] Natural gamma: A method for measuring the radioactivity of natural radioactive elements (mainly potassium-40, uranium-238, and thorium-232) in formations. These radioactive elements emit gamma rays during the decay process. By measuring the intensity of these gamma rays, the natural gamma value of the formation can be obtained.

[0076] P-wave impedance: A measure of the resistance encountered by seismic waves when propagating in a medium. It is the product of the medium density and the P-wave velocity. P-wave impedance reflects the degree of obstruction of the medium to the transmission of acoustic wave energy;

[0077] Seismic amplitude: Refers to the maximum deviation value recorded during the propagation of seismic waves, usually captured by seismic detectors (such as seismographs); Waveform inversion: A technique that uses complete seismic waveform information to estimate the physical parameters of underground media. It is not only based on amplitude but also includes waveform characteristics such as phase and frequency to more accurately reconstruct the underground structure;

[0078] Waveform simulation: The process of generating the propagation path and waveform of seismic waves in underground media through numerical simulation technology. It is used to verify and improve the results of waveform inversion and can also be used as an independent tool for predicting seismic responses;

[0079] Gradient descent method: An optimization algorithm used to minimize the loss function, widely applied in machine learning and deep learning. By iteratively adjusting the model parameters, the value of the loss function gradually decreases until a local or global optimal solution is reached;

[0080] Feature number: In machine learning and data analysis, it refers to the number of features contained in each sample object in the input dataset, which is used to describe different dimensions or variables of the sample object and provides the information required by the model for prediction, classification, or other tasks;

[0081] Learning rate: It is an important hyperparameter in the gradient descent algorithm, which determines the step size of parameter update in each iteration and controls the speed and amplitude of model parameter adjustment.

[0082] In the prior art, for coalbed methane extraction, it is necessary to judge the formation lithology of the complex geological structure in the reservoir, find the coal seam position through the coal seam calibration in the formation lithology, and the formation lithology judgment technology deduces the formation lithology in a larger range through the local information of the wellbore. It is necessary to input various attributes as training samples into the model to be trained, and there is no clear standard for the selection of attributes. Too few attributes cannot correctly represent the formation lithology, while too many attributes are prone to data redundancy, reducing the prediction accuracy. Moreover, only the model design of linear weighting is adopted, ignoring the relationship between complex attributes, resulting in a problem of low prediction accuracy of formation lithology due to poor model training effect.

[0083] A model training method and a formation lithology identification method provided by an embodiment of the present application. Since the accuracy of judging the formation lithology through a single attribute of the sample object is relatively low, a variety of attributes of the sample object are combined, and the degree of sample aggregation in the attribute combination is observed to judge whether the formation lithology can be distinguished by the attribute combination, so as to select the geophysical attributes associated with the formation lithology. In addition, the cross-entropy loss and the similarity difference loss are used to adjust the parameters of the non-linear model, so that the non-linear model can be trained by continuously changing the loss value, thereby avoiding the complex relationship between attributes from affecting the non-linear model, and enabling the trained target model to more accurately identify the formation lithology based on a variety of geophysical attributes.

[0084] Figure 1 The scenario schematic diagram of a model training method and a formation lithology identification method provided by an embodiment of the present application is as Figure 1 shown. The execution subject of this method can be a server, and the server includes a first server and a second server.

[0085] The first server includes a model training system for performing the following operations: determining a set of sample objects, which includes sample objects, as well as the attribute combinations and formation lithologies of the sample objects, where the attribute combinations include at least two different geophysical attributes; constructing coordinate systems for different attribute combinations according to the types of geophysical attributes in the attribute combinations; determining a target coordinate system and a corresponding target attribute combination according to the positions of the sample objects in the coordinate systems of different attribute combinations, where the sample objects with the same formation lithology in the target coordinate system meet a preset aggregation requirement; and training a model to be trained according to the target attribute combinations of the sample objects and the formation lithologies corresponding to the target attribute combinations to obtain a target model.

[0086] The second server includes a formation lithology identification system for performing the following operations: obtaining the geophysical attributes of samples in a target area, and inputting the geophysical attributes into the target model to obtain the target formation lithology of the samples in the target area.

[0087] In the embodiments of the present application, the first server and the second server may be devices such as a computer, a notebook, a tablet, or a mobile phone. The embodiments of the present application do not particularly limit the implementation manner of the execution entity, as long as the execution entity can execute the model training method and the formation lithology identification method in the present application.

[0088] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below with specific embodiments. These specific embodiments below can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below with reference to the accompanying drawings.

[0089] Figure 2 Schematic flow of a model training method provided by the present application Figure 1 As Figure 2 shown, the method includes:

[0090] S201. Determine a set of sample objects, which includes sample objects, as well as the attribute combinations and formation lithologies of the sample objects, where the attribute combinations include at least two different geophysical attributes.

[0091] Among them, the set of sample objects may refer to the data set used for model training. The data set stores the sample data of multiple sample objects. Each sample object may be an actual physical object, event, or entity. In the embodiments of the present application, the set of sample objects is used to store multiple sample objects in the geological structure. The sample object is a specific geological sample unit. The sample objects at different positions may have the same geophysical attributes and formation lithologies, or may have different geophysical attributes and formation lithologies.

[0092] The sample data may include the combination of attributes of the sample object and the formation lithology. Among them, the combination of attributes of the sample object may refer to the combination of various geophysical attributes of the sample object, and the combination of attributes may include at least two different geophysical attributes. Among them, there may be two different attributes, or there may be three, at least including two different attributes.

[0093] Geophysical attributes may refer to various parameters or characteristics of the material properties of a certain sample point in the geological structure. The material properties may represent physical properties, chemical properties or structural properties. Physical properties may include color, hardness, density, etc. Chemical properties may include organic matter content, elemental composition, pH value, etc. Structural properties may include texture, fault, porosity, etc.

[0094] The formation lithology may refer to the identification result of the sample object. For example, when the sample object represents a sample point at a certain place in the geological structure, the formation lithology may refer to the category of the sample point at that place. For example, the category may include coal seams, and for another example, sandstone and other categories, used to distinguish different categories.

[0095] Among them, in the embodiments of the present application, determining the sample object set includes:

[0096] Obtain the geophysical information of the sample object and the formation lithology data corresponding to the geophysical information. Among them, the geophysical information includes at least one of seismic data, logging data, and drilling data;

[0097] Perform combination processing on at least two geophysical attributes in the seismic data, logging data, and drilling data to obtain the combination of attributes of the sample object;

[0098] Determine the formation lithology corresponding to the combination of attributes of the sample object according to the formation lithology data;

[0099] Determine the sample object set according to the combination of attributes of the sample object and the formation lithology.

[0100] Among them, geophysical information may refer to data on the earth's materials and their properties collected through various geological survey and exploration techniques. The earth's materials may refer to rocks, minerals, soils, etc. Geophysical information may be directly obtained from the field or the result obtained through laboratory analysis of samples, used to understand the crust structure, resource distribution, environmental conditions, etc.

[0101] In the embodiments of the present application, the geophysical information may include at least one of seismic data, logging data, and drilling data, specifically depending on the geological sample situation extracted from the on-site geological structure. Among them, the sample volumes of the three geophysical informations may be different or the same.

[0102] Stratigraphic lithology data can refer to specific category information used to describe and classify rocks, minerals, or strata. Geological materials can be classified into different geological units based on factors such as lithological characteristics, sedimentary environment, and formation process. In the embodiments of the present application, the stratigraphic lithology of each geological unit can be obtained by acquiring stratigraphic lithology data.

[0103] Combined processing can refer to combining geophysical information from different sources or of different types to form an attribute combination. The sample object can be more comprehensively described through the complementarity between multiple data. The types of attributes in combined processing can be two, three, or more than three, and are specifically set according to usage requirements.

[0104] S202. Construct coordinate systems for different attribute combinations according to the types of geophysical attributes in the attribute combination.

[0105] Among them, different attribute combinations correspond to any combination of multiple attributes. In the embodiments of the present application, the geophysical attributes in seismic data and the geophysical attributes in logging data can be combined, or two or three geophysical attributes in drilling data can be combined. Different combination methods correspond to different coordinate systems.

[0106] In the embodiments of the present application, when the attribute combination includes two different geophysical attributes, constructing coordinate systems for different attribute combinations according to the types of geophysical attributes in the attribute combination includes:

[0107] Use the type of one geophysical attribute in the attribute combination as the x-axis of the coordinate system and the type of the other geophysical attribute in the attribute combination as the y-axis of the coordinate system to construct coordinate systems for different attribute combinations.

[0108] The x-axis of the coordinate system can refer to the horizontal axis of the coordinate system, and the y-axis of the coordinate system can refer to the vertical axis of the coordinate system. The unit division of the horizontal axis and the vertical axis can be set according to the units of specific attributes to ensure that the position of each sample object in the coordinate system accurately reflects its actual measured values on these attributes. When the attribute combination includes two different geophysical attributes, constructing the coordinate system can use one attribute as the x-axis of the two-dimensional coordinate system and the other attribute as the y-axis of the two-dimensional coordinate system, and display the attribute values within the coordinate system according to their respective unit divisions. For example, the horizontal axis stores the natural gamma, and the unit is usually API (American Petroleum Institute), and the scale can be set as 0, 50, 100, 150, while the vertical axis stores the longitudinal wave impedance, and the unit is usually the product of density and velocity, and the scale can be set as 1000, 2000, 3000, 4000, 5000, 6000.

[0109] In some embodiments, a variety of different geophysical properties can be included in the property combination. For different numbers of geophysical properties, different processing methods are adopted. When two different geophysical properties are included in the property combination, a two-dimensional coordinate system is used to store the geophysical property and the corresponding property value. When three different geophysical properties are included in the property combination, a three-dimensional coordinate system is used to store the geophysical property and the corresponding property value.

[0110] S203. Determine the target coordinate system and the target property combination corresponding to the target coordinate system according to the positions of the sample objects in the coordinate systems of different property combinations. The sample objects with the same formation lithology in the target coordinate system meet the preset aggregation requirements.

[0111] Among them, the target coordinate system can refer to one or more coordinate systems selected through analysis and optimization. In these coordinate systems, the position distribution of the sample objects can reflect specific geological features or patterns. In the embodiments of the present application, by analyzing the coordinate systems formed by different property combinations, a coordinate system that can make the sample objects aggregate or distinguish in the expected manner is found, and the finally selected coordinate system is the target coordinate system. The target coordinate system enables the sample objects of the same category to be closely aggregated together, or makes the boundaries between different categories of sample objects clearer.

[0112] The target property combination can refer to a set of geophysical properties that reveal the relationships between sample objects. The target property combination is composed of two or more geophysical properties. By using the target property combination, an obvious data representation form can be obtained in the target coordinate system, thereby providing a basis for subsequent model training and prediction. Among them, the target property combination corresponds to the target coordinate system one by one.

[0113] Among them, in the embodiments of the present application, determining the target coordinate system and the target property combination corresponding to the target coordinate system according to the positions of the sample objects in the coordinate systems of different property combinations includes:

[0114] Determine the aggregation degree of the sample objects of different formation lithologies in the coordinate system according to the positions of the sample objects in the coordinate systems of different property combinations;

[0115] Determine the sample objects that meet the aggregation requirements according to the aggregation degree of the sample objects of different formation lithologies in the coordinate system;

[0116] Determine the target coordinate system and the target property combination corresponding to the target coordinate system according to the sample objects that meet the aggregation requirements.

[0117] Among them, the aggregation degree may refer to the degree of closeness or similarity between sample objects in a specific coordinate system, which is used to reflect whether the sample objects tend to form dense groups. The aggregation degree of sample objects with different formation lithologies in the coordinate system may refer to that after the sample objects are classified into different formation lithologies according to geophysical attributes, it is checked whether the sample objects of each formation lithology form a group. If the sample objects of the same formation lithology are highly concentrated in the coordinate system and there is sufficient separation between different formation lithologies, it indicates that the coordinate system has a good discrimination effect on these types.

[0118] The aggregation requirement may refer to the set standards or conditions used to measure and judge whether a coordinate system and its corresponding attribute combination can effectively distinguish sample objects with different formation lithologies. The sample objects that meet the preset aggregation requirements may refer to the sample objects whose distribution conforms to the above aggregation requirements in a specific coordinate system.

[0119] Among them, in the embodiments of the present application, according to the positions of the sample objects in the coordinate systems with different attribute combinations, determining the aggregation degree of the sample objects with different formation lithologies in the coordinate system includes:

[0120] Determining the distances between the sample objects in the coordinate systems with different attribute combinations according to the positions of the sample objects in the coordinate systems with different attribute combinations;

[0121] Determining the distance levels of the sample objects on the coordinate systems with different attribute combinations according to the distances between the sample objects in the coordinate systems with different attribute combinations;

[0122] Determining the aggregation degree of the sample objects with different formation lithologies in the coordinate system according to the distance levels and the positions of the sample objects in the coordinate systems with different attribute combinations.

[0123] Among them, the distance between the sample objects in the coordinate systems with different attribute combinations may refer to the relative position difference between two or more sample objects in the multi-dimensional space defined by a specific attribute combination. The relative position difference can be calculated according to the distance formula, and the distance formula is not limited in the present application as long as the distance between the sample objects can be calculated.

[0124] The distance level of sample objects on a coordinate system with different attribute combinations can refer to the result of segmenting or classifying the calculated distances, usually based on a certain threshold or rule. The distance level can help identify which sample objects are close to each other and which are far apart. For example, the distance level is divided into a low distance level, a medium distance level, and a high distance level, indicating that the distances between sample objects are very small, moderate, and large respectively. By dividing the distance level, the clustering degree of sample objects with different formation lithologies in each coordinate system can be evaluated more intuitively. If a coordinate system can make the sample objects of the same formation lithology mainly concentrated in the "low distance level", while maintaining a large "high distance level" between different types, then this coordinate system shows a good clustering effect.

[0125] By analyzing the distances and their levels between sample objects in a coordinate system with different attribute combinations, a coordinate system that meets the preset aggregation requirements can be selected as the target coordinate system. In the target coordinate system, sample objects of the same type are closely clustered to form clear clusters, while there is an obvious separation between sample objects of different types. At the same time, record the attribute combination used to construct this coordinate system as the target attribute combination.

[0126] In the embodiment of the present application, well logging and seismic surveys are carried out on rocks about 2 kilometers deep to obtain data of different scales under different observation means. Then, a large number of derived physical-meaningful attributes are obtained from the data of different scales, and 2D or 3D scatter statistics are used for analysis to find the target attribute combination related to the formation lithology.

[0127] S204. Train the model to be trained according to the target attribute combination of the sample objects and the formation lithology corresponding to the target attribute combination to obtain the target model.

[0128] Among them, the model to be trained can refer to a model that has defined the structure, algorithm, and parameter form but has not been trained with data yet. The model to be trained contains a hypothetical framework for the function to be achieved or the problem to be solved, but its internal parameters have not been adjusted with actual data to adapt to a specific task.

[0129] The target model can refer to a model that can make accurate predictions or classifications for new data after training. The target model is obtained by training the model to be trained, using a known data set (including input features and corresponding output labels) to adjust the parameters of the model so that the model can minimize the loss function or optimize a certain performance metric on these data.

[0130] In the embodiments of the present application, based on the integration of multiple geophysical attributes, the formation lithology is determined. Since there is a non-linear relationship between multiple geophysical attributes, the traditional linear weighting method cannot be used. Instead, a non-linear relationship between multiple attributes and the formation lithology is established using a model. Multiple geophysical attributes are used as the input features of the model, and the formation lithology is used as the output label of the model. Among them, the non-linear relationship model can select a random forest, which reduces overfitting through multiple decision trees, has strong ability to process high-dimensional data and missing values, and provides feature importance evaluation. It can also select a Bayesian optimization model to find the global optimal solution on a small number of data points and maximize the use of existing information.

[0131] In some embodiments, the geophysical attributes in multiple target attribute combinations input to the model to be trained are normalized. The normalization method can use min-max normalization, which maintains the proportional relationship of the original data, or other normalization methods such as decimal scaling normalization. As long as the weight polarization of the model to be trained can be avoided, the convergence can be accelerated, and the accuracy of the model can be improved after normalization. Among them, min-max normalization satisfies:

[0132]

[0133] where X' is the normalized data, X is the data to be normalized, X min is the minimum value in the data, and X max is the maximum value in the data.

[0134] Preferably, BiGRU (Bidirectional Gated Recurrent Unit) can be used as the main network architecture to design complex non-linear functions. Since geophysical information usually has strong context dependence, that is, a data point at a location is not only affected by its previous data but may also be related to subsequent data. BiGRU can establish deeper connections between different parts of the time series through bidirectional information flow, ensuring that the model fully understands the context of each data point.

[0135] A model training method provided by an embodiment of the present application collects geological samples as sample objects. Each sample object has at least two different geophysical attributes and known formation lithologies. Then, multiple coordinate systems are constructed according to the attribute combinations, and each coordinate system represents a way of attribute combination, so as to comprehensively evaluate the influence of different attributes on the sample distribution. Then, the positions of the sample objects are analyzed in these coordinate systems to find a target coordinate system and its corresponding target attribute combination that can make similar samples closely cluster and different types of samples clearly separate, thereby improving the clustering effect and ensuring high similarity between similar samples and low confusion between different types of samples. Finally, the geophysical attributes in the target attribute combination are used as the input features of the model to be trained, and the formation lithology corresponding to the target attribute combination is used as the output label to obtain a target model, enabling the model to show higher accuracy and reliability when processing new data, effectively improving the efficiency and accuracy of geological classification, and helping to more deeply understand and predict geological phenomena.

[0136] Figure 3 Flow schematic of a model training method provided by the present application Figure 2 As Figure 3 shown, based on the Figure 2 embodiment, a model training method is described in detail. According to the target attribute combination of the sample object and the formation lithology corresponding to the target attribute combination, the model to be trained is trained to obtain a target model, including:

[0137] S301. Input all the attributes in the target attribute combination of the sample object into the model to be trained to obtain a predicted formation lithology.

[0138] Among them, the predicted formation lithology can refer to the formation lithology that the sample object is most likely to belong to after the attribute combination of the given sample object is input through the model to be trained. This prediction is based on the patterns and rules learned by the model to be trained from existing data. The model to be trained will calculate the probability distribution of each formation lithology according to the input attribute values and select the formation lithology with the highest probability as the predicted formation lithology.

[0139] S302. Obtain a cross-entropy loss and a similarity difference loss according to the predicted formation lithology and the formation lithology.

[0140] Among them, the cross-entropy loss can be an index to measure the difference between the predicted formation lithology and the actual formation lithology, used to evaluate the matching degree between the probability distribution predicted by the model and the true label. If the prediction is accurate, the cross-entropy loss will be low, otherwise it will be high.

[0141] The similarity difference loss can refer to an index that measures the degree of similarity between the model's prediction results and the true labels. Even if the predicted formation lithology is not completely correct, if it is very similar to the actual type, the similarity difference loss can reduce the penalty, thereby encouraging the model to capture more subtle geological features.

[0142] S303. Determine the target loss according to the cross-entropy loss and the similarity loss;

[0143] S304. Adjust the model to be trained according to the target loss to obtain the target model.

[0144] Among them, the target loss can refer to a comprehensive index used to evaluate the difference between the model's prediction results and the true labels during the model training process, which can guide the adjustment of model parameters, enabling the model to gradually optimize its performance. By minimizing the target loss, the model can gradually adjust its internal parameters to improve its prediction ability. In each iteration, the model updates the weights and other parameters according to the current target loss value in order to obtain a lower loss value on future batches of data.

[0145] The target loss can combine multiple loss functions, and there can be multiple combination methods, depending on the requirements of the training task, such as linear weighting, weighted average, or dynamic weighting.

[0146] Among them, in the embodiments of the present application, determining the target loss according to the cross-entropy loss and the similarity loss includes:

[0147] Obtain the target similarity loss according to the similarity loss and a preset coefficient, where the preset coefficient is determined according to the proportion of the number of sample objects of different formation lithologies in the sample object set, and different preset coefficients correspond to different proportions;

[0148] Determine the target loss according to the cross-entropy loss and the target similarity loss.

[0149] Among them, the preset coefficient can refer to a weight factor used to adjust the importance of the similarity loss. The preset coefficient dynamically adjusts the proportion of the similarity loss in the entire target loss to adapt to the optimization requirements under different class distributions. When the formation lithology is severely unbalanced, a larger preset coefficient can be set to increase the weight of the similarity loss, which helps the model better process small-sample formation lithology data and prevent it from being overwhelmed by the majority class; when the formation lithology distribution is relatively uniform, the preset coefficient is appropriately reduced to make the cross-entropy loss play a greater role, thereby improving the overall formation lithology classification accuracy.

[0150] The target similarity loss can refer to the similarity loss weighted by a preset coefficient, which reflects the overlap or similarity between the model prediction result and the true label, and adjusts the contribution of the similarity loss to the overall target loss through the preset coefficient.

[0151] In the embodiments of the present application, the target loss satisfies:

[0152] y = y1 + λ × y2;

[0153] Wherein, y represents the target loss, y1 represents the cross-entropy loss, y2 represents the similarity loss, λ represents the preset coefficient, and λ × y2 represents the target similarity loss.

[0154] Preferably, the target loss is calculated by dynamically weighting the cross-entropy loss function and the Dice loss function, and the cross-entropy loss function satisfies:

[0155]

[0156] Wherein, y1 represents the cross-entropy loss, N represents the total number of sample objects, K represents the total number of formation lithologies, p i,k represents the predicted probability value that the sample object i belongs to the formation lithology k, and g i,k represents the true label value that the sample object i belongs to the formation lithology k.

[0157] The similarity loss satisfies:

[0158]

[0159] Wherein, y2 represents the similarity loss, K represents the total number of formation lithologies, p k,i represents the predicted probability value that the sample object i belongs to the formation lithology k, and g k,i represents the true label value that the sample object i belongs to the formation lithology k, and ε represents the preset coefficient to prevent the denominator from being 0 due to numerical instability.

[0160] In some embodiments, when there are only two formation lithologies, the similarity loss can be set as the application of the Dice loss in binary classification, and the similarity loss satisfies:

[0161]

[0162] Wherein, y2 represents the similarity loss, p represents the formation lithology prediction result, and g represents the true formation lithology label.

[0163] By calculating the gradient of the target loss with respect to the model parameters, the gradient descent method guides the adjustment direction of the model parameters. In each iteration, the model updates the weights and other parameters according to the current target loss value, aiming to obtain a lower loss value on future batches of data. Through continuous monitoring and feedback of the target loss, the learning rate can be dynamically adjusted to ensure that the model converges quickly in the initial stage of training and can finely adjust the parameters in the later stage to avoid overfitting or underfitting.

[0164] A model training method provided by an embodiment of this application inputs the target attribute combination of a sample object into the model to obtain a predicted formation lithology, and calculates the cross-entropy loss and the similarity difference loss according to the prediction result and the true label. Then, in combination with a preset coefficient determined based on the proportion of the number of categories, the similarity loss is adjusted to obtain the target similarity loss. Finally, the target loss is determined by comprehensively considering the cross-entropy loss and the target similarity loss to guide the optimization of the model parameters. This not only improves the model's ability to process class-imbalanced data and ensures that minority classes can also be effectively identified, but also enhances the overall classification accuracy and robustness, making the model perform more excellently in complex geological classification tasks.

[0165] Figure 4 This is a schematic diagram of the overall framework of a model training method provided by an embodiment of this application. As Figure 4 shown, a sample object set is determined, and through analysis based on the sample object set, corresponding lithology-sensitive attribute analysis is carried out to obtain all the attributes in the target attribute combination, namely natural gamma, longitudinal wave impedance, and seismic attributes. Using the target attribute combination and its corresponding formation lithology, model training based on a neural network can be carried out to obtain the predicted formation lithology, which is represented in Figure 4 as a three-dimensional distribution map of coal, mudstone, and gravel.

[0166] Figure 5 This is a schematic diagram of the result of a model training method provided by an embodiment of this application. As Figure 5 shown, the model architecture based on BiGRU is stacked by custom multi-layer GRUs, and the sample data is read from both the forward and backward directions to obtain more comprehensive data information. The result obtained through the BiGRU network passes through a fully connected layer to output the final formation lithology. The model is designed with 6 layers, that is, 6 layers of non-linear functions are nested into a complex non-linear composite function, and each layer has 50 feature numbers. The initial learning rate is set to 0.01, the number of training iterations is 500 times, and the AdamW (Adam with WeightDecay Fix, adaptive moment estimation optimizer with weight decay) is used to solve the weight coefficients of the non-linear function.

[0167] Figure 6 This is a schematic diagram of the process of the formation lithology identification method provided by this application. As Figure 6As shown, the method includes:

[0168] S601. Obtain the geophysical properties of the samples in the target area;

[0169] S602. Input the geophysical properties into the target model to obtain the target formation lithology of the samples in the target area.

[0170] Among them, the samples in the target area can refer to the samples applied to the model after model training, which can be used to verify the accuracy of the model and can also be used for the identification of formation lithology in the task area.

[0171] The formation lithology identification method provided in this embodiment can use the target model in the above model training method to output the target formation lithology of the samples in the target area, which will not be elaborated here in this embodiment.

[0172] Figure 6a It is a result schematic diagram of the formation lithology identification method provided in the embodiment of the present application Figure 1 , such as Figure 6a shown, the longitudinal wave impedance can better distinguish coal rock and surrounding rock, and the natural gamma can better distinguish sandstone and mudstone.

[0173] Figure 6b It is a result schematic diagram of the formation lithology identification method provided in the embodiment of the present application Figure 2 , such as Figure 6b shown, and there is also an approximately linear relationship between the seismic amplitude data and the coal seam thickness. Therefore, the longitudinal wave impedance, natural gamma, and seismic amplitude data are used to predict the formation lithology. Among them, the longitudinal wave impedance and natural gamma are obtained through logging, and only at a certain point in the three-dimensional space, their three-dimensional volume data need to be obtained through waveform indication inversion and waveform indication simulation respectively.

[0174] Figure 6c It is a result schematic diagram of the formation lithology identification method provided in the embodiment of the present application Figure 3 , the samples are divided into 16 training well samples and 2 verification well samples. Among them, the 16 training well samples are used as the sample object set. After training the target model, the 2 verification well samples are used for verification. Among them, the conventional method is to linearly weight and analyze the formation lithology. It can be seen that the formation lithology in the prediction result of the present application basically corresponds to the actual formation lithology. The accuracy rate can be obtained by dividing the statistically matched data points by the total data points. As Figure 6c shown, it is a result schematic diagram of one of the verification well w1 samples. The accuracy rates of the conventional method and the lithology prediction of the present application for well w1 are 86% and 90% respectively, and the accuracy rate is increased by 6%.

[0175] Figure 6d It is a result schematic diagram of the formation lithology identification method provided in the embodiment of the present application Figure 4 , such as Figure 6dThe figure shows the result schematic diagram of one of the verification well w2 samples. The accuracy rate of the lithology prediction result by the conventional method is only 77%, while the accuracy rate of the lithology prediction result of this application is 87%, with an increase of 10% in accuracy rate.

[0176] Figure 6e This is the result schematic of the formation lithology identification method provided by the embodiment of this application Figure 5 , such as Figure 6e The figure shows the lithology prediction profile of the conventional method.

[0177] Figure 6f This is the result schematic of the formation lithology identification method provided by the embodiment of this application Figure 6 , such as Figure 6f The figure shows the lithology prediction profile of the method of this application. From the profile, although there are some burr phenomena in the lithology prediction profile of this application compared with the lithology prediction profile of the conventional method, the overall lithology resolution is significantly improved, the lithology continuity is good, there are many thin layers and they can correspond to the wellhead lithology.

[0178] Figure 7 This is the structural schematic diagram of the model training device provided by this application. As Figure 7 shown, the model training device 70 provided by this embodiment includes:

[0179] The first determination module 701 is used to determine a sample object set, and the sample object set includes sample objects, as well as the attribute combinations and formation lithologies of the sample objects. The attribute combinations include at least two different geophysical attributes;

[0180] The processing module 702 is used to construct coordinate systems with different attribute combinations according to the types of geophysical attributes in the attribute combinations;

[0181] The second determination module 703 is used to determine a target coordinate system and a target attribute combination corresponding to the target coordinate system according to the positions of the sample objects in the coordinate systems with different attribute combinations;

[0182] The training module 704 is used to train the model to be trained according to the target attribute combination of the sample object and the formation lithology corresponding to the target attribute combination to obtain a target model.

[0183] In the embodiment of this application, the first determination module 701 can also specifically be used for:

[0184] Obtain the geophysical information of the sample object and the formation lithology data corresponding to the geophysical information, where the geophysical information includes at least one of seismic data, logging data, and drilling data;

[0185] Perform combined processing on at least two geophysical attributes among seismic data, logging data, and drilling data to obtain the attribute combination of the sample object;

[0186] Determine the formation lithology corresponding to the attribute combination of the sample object according to the formation lithology data;

[0187] Determine the sample object set according to the attribute combination of the sample object and the formation lithology.

[0188] In the embodiment of the present application, the processing module 702 may further be specifically configured to:

[0189] When the attribute combination includes two different geophysical attributes,

[0190] Construct coordinate systems for different attribute combinations according to the types of geophysical attributes in the attribute combination, including:

[0191] Use the type of one geophysical attribute in the attribute combination as the x-axis of the coordinate system and the type of the other geophysical attribute in the attribute combination as the y-axis of the coordinate system to construct coordinate systems for different attribute combinations.

[0192] In the embodiment of the present application, the second determination module 703 may further be specifically configured to:

[0193] Determine the aggregation degree of sample objects with different formation lithologies in the coordinate system according to the positions of the sample objects in the coordinate systems of different attribute combinations;

[0194] Determine the sample objects that meet the aggregation requirements according to the aggregation degree of sample objects with different formation lithologies in the coordinate system;

[0195] Determine the target coordinate system and the target attribute combination corresponding to the target coordinate system according to the sample objects that meet the aggregation requirements.

[0196] In the embodiment of the present application, the second determination module 703 may further be specifically configured to:

[0197] Determine the distances between sample objects in the coordinate systems of different attribute combinations according to the positions of the sample objects in the coordinate systems of different attribute combinations;

[0198] Determine the distance levels of sample objects on the coordinate systems of different attribute combinations according to the distances between sample objects in the coordinate systems of different attribute combinations;

[0199] Determine the aggregation degree of sample objects with different formation lithologies in the coordinate system according to the distance levels and the positions of the sample objects in the coordinate systems of different attribute combinations.

[0200] In the embodiment of the present application, the training module 704 may further be specifically configured to:

[0201] Input all the attributes in the target attribute combination of the sample object into the model to be trained to obtain the predicted formation lithology;

[0202] Based on the predicted formation lithology and the formation lithology, obtain the cross-entropy loss and the similarity difference loss;

[0203] Based on the cross-entropy loss and the similarity loss, determine the target loss;

[0204] Based on the target loss, adjust the model to be trained to obtain the target model.

[0205] In the embodiment of the present application, the training module 704 may further be specifically configured to:

[0206] Based on the similarity loss and a preset coefficient, obtain the target similarity loss, where the preset coefficient is determined according to the proportion of the number of sample objects of different formation lithologies in the sample object set, and different preset coefficients correspond to different proportions;

[0207] Based on the cross-entropy loss and the target similarity loss, determine the target loss.

[0208] The model training device provided in this embodiment can execute the method provided in the above method embodiment, and its implementation principle and technical effect are similar, which will not be elaborated here in this embodiment.

[0209] Figure 8 The structural schematic diagram of the formation lithology identification device provided for the present application is as Figure 8 shown. The formation lithology identification device 80 provided in this embodiment includes:

[0210] An acquisition module 801, configured to acquire the geophysical attributes of the sample in the target area;

[0211] A obtaining module 802, configured to input the geophysical attributes into the target model to obtain the target formation lithology of the sample in the target area.

[0212] The formation lithology identification device provided in this embodiment can execute the method provided in the above method embodiment, and its implementation principle and technical effect are similar, which will not be elaborated here in this embodiment.

[0213] Figure 9 The structural schematic diagram of the electronic device provided for the present application. As Figure 9 shown, the electronic device 90 provided in this embodiment includes: at least one processor 901 and a memory 902. Optionally, the device 90 further includes a communication component 903. Among them, the processor 901, the memory 902, and the communication component 903 are connected through a bus 904.

[0214] In the specific implementation process, at least one processor 901 executes the computer execution instructions stored in the memory 902, so that at least one processor 901 executes the above method.

[0215] For the specific implementation process of the processor 901, reference may be made to the above method embodiments. Their implementation principles and technical effects are similar, and will not be elaborated here in this embodiment.

[0216] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU for short), or other general-purpose processors, digital signal processors (DSP for short), application specific integrated circuits (ASIC for short), etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method disclosed in combination with the invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

[0217] The memory may include a high-speed random access memory (RAM), and may also include a non-volatile memory (NVM), such as at least one disk memory.

[0218] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, the buses in the drawings of this application are not limited to only one bus or one type of bus.

[0219] This application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0220] This application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When the processor executes the computer-executable instructions, the above method is implemented.

[0221] The above-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disk. The readable storage medium can be any available medium accessible by a general-purpose or special-purpose computer.

[0222] An exemplary readable storage medium is coupled to the processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium can also exist as discrete components in a device.

[0223] The division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be indirect couplings or communication connections through some interfaces, devices, or units, and can be in electrical, mechanical, or other forms.

[0224] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0225] In addition, the functional units in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0226] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, and other various media that can store program codes.

[0227] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When this program is executed, it executes the steps including the above method embodiments; and the aforementioned storage medium includes: ROMs, RAMs, magnetic disks, or optical discs, and other various media that can store program codes.

[0228] Finally, it should be noted that: after considering the specification and practicing the invention disclosed herein, those skilled in the art will easily think of other implementation manners of the present invention. The present invention is intended to cover any variations, uses, or adaptive changes of the present invention. These variations, uses, or adaptive changes follow the general principles of the present invention and include common general knowledge or conventional technical means in the technical field not disclosed in the present invention. It is not limited to the exact structure described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.

Claims

1. A model training method, applied to a geological model, characterized in that: include: Determine a sample object set, the sample object set including sample objects, attribute combinations of the sample objects and formation lithology, the attribute combination including at least two different geophysical attributes; According to the type of geophysical attributes in the attribute combination, constructing coordinate systems of different attribute combinations; Determine a target coordinate system and a target attribute combination corresponding to the target coordinate system according to the position of the sample object in the coordinate system of different attribute combinations, wherein the sample objects of the same stratum lithology in the target coordinate system meet a preset aggregation requirement; The model to be trained is trained according to the target attribute combination of the sample object and the stratum lithology corresponding to the target attribute combination to obtain a target model.

2. The method according to claim 1, characterized in that The step of determining the sample object set includes: Acquire geophysical information of the sample object and stratum lithology data corresponding to the geophysical information, wherein the geophysical information includes at least one of seismic data, well logging data, and drilling data; Combining and processing at least two geophysical attributes of the seismic data, the well logging data, and the drilling data to obtain an attribute combination of the sample object; Determining the stratum lithology corresponding to the attribute combination of the sample object according to the stratum lithology data; A sample object set is determined according to the attribute combination of the sample object and the stratum lithology.

3. The method according to claim 1, characterized in that When the attribute combination includes two different geophysical attributes, The step of constructing coordinate systems of different attribute combinations according to the types of geophysical attributes in the attribute combination includes: The type of one geophysical attribute in the attribute combination is used as the x-axis of the coordinate system, and the type of another geophysical attribute in the attribute combination is used as the y-axis of the coordinate system, so as to construct coordinate systems of different attribute combinations.

4. The method according to claim 1, characterized in that The determining of a target coordinate system and a target attribute combination corresponding to the target coordinate system according to the position of the sample object in the coordinate system of different attribute combinations includes: Determine the degree of aggregation of sample objects of different stratum lithology in the coordinate system according to the positions of the sample objects in the coordinate system of different attribute combinations; Determining sample objects that meet the aggregation requirements according to the aggregation degree of sample objects of different stratum lithology in the coordinate system; The target coordinate system and a target attribute combination corresponding to the target coordinate system are determined according to the sample objects that meet the aggregation requirement.

5. The method according to claim 4, characterized in that Determining the aggregation degree of sample objects of different stratum lithologies in the coordinate system according to the positions of the sample objects in the coordinate system of different attribute combinations includes: Determine the distance between the sample objects in the coordinate systems of different attribute combinations according to the positions of the sample objects in the coordinate systems of different attribute combinations; Determine the distance level of the sample objects in the coordinate systems of different attribute combinations according to the distances between the sample objects in the coordinate systems of different attribute combinations; The degree of aggregation of sample objects of different stratum lithology in the coordinate system is determined according to the distance level and the positions of the sample objects in the coordinate system of different attribute combinations.

6. The method according to claim 1, characterized in that The step of training the model to be trained according to the target attribute combination of the sample object and the stratum lithology corresponding to the target attribute combination to obtain the target model includes: Inputting all attributes in the target attribute combination of the sample object into the model to be trained to obtain predicted formation lithology; According to the predicted formation lithology and the formation lithology, a cross entropy loss and a similarity difference loss are obtained; Determining a target loss according to the cross entropy loss and the similarity loss; According to the target loss, the model to be trained is adjusted to obtain a target model.

7. The method according to claim 6, characterized in that The determining the target loss according to the cross entropy loss and the similarity loss includes: According to the similarity loss and the preset coefficient, a target similarity loss is obtained, wherein the preset coefficient is determined according to the proportion of the sample objects of different stratum lithologies in the sample object set, and different proportions of the number correspond to different preset coefficients; A target loss is determined according to the cross entropy loss and the target similarity loss.

8. A method for identifying stratum lithology, characterized in that: The method comprises: Obtain geophysical properties of samples from target areas; The geophysical attributes are input into the target model according to any one of claims 1 to 7 to obtain target formation lithology of samples in the target area.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 8 when executed by a processor.

10. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 8 when being executed by a processor.

Images