Gas content prediction method and device, storage medium and program product

Abstract

The invention provides a gas content prediction method and device, a storage medium and a program product. The method comprises the following steps: acquiring multi-source data information of a rock stratum to be predicted; performing data integration processing and feature extraction processing on the multi-source data information to obtain a plurality of input features; performing scale division processing on the plurality of input features to obtain input features corresponding to each scale type, the scale types including a large scale type, a medium scale type and a small scale type; according to the scale type, inputting the plurality of input features to a trained gas-bearing prediction model step by step, the gas-bearing prediction model being obtained by training a plurality of training samples, and each training sample comprising a plurality of corresponding input feature samples and gas-bearing label samples; and according to the gas-bearing prediction label of the trained gas-bearing prediction model, carrying out lithology inversion to obtain a gas-bearing prediction result of the rock stratum to be predicted, so as to achieve the effect of improving the gas-bearing prediction accuracy and reliability.

Description

Technical Field

[0001] This application relates to the technical fields of artificial intelligence and geophysical exploration, and particularly to a gas-bearing property prediction method, device, storage medium, and program product. Background Art

[0002] With the growth of global energy demand and the depletion of conventional oil and gas resources, unconventional oil and gas resources have become a new focus. Among them, tight sandstone gas resources are rich, have stable production, are clean and environmentally friendly, and the development technology is mature, which is crucial for the global energy transition. However, its exploration and development face technical challenges, especially in differentiating water layers and gas layers in high-water-bearing formations. Due to the small difference in physical properties of tight sandstone gas reservoirs and the similar seismic response characteristics of water layers and gas layers, it is more difficult to predict the enrichment area. Therefore, for the tight sandstone gas enrichment area, developing a more efficient and accurate gas-bearing property prediction method has become a promising direction.

[0003] In the prior art, the gas-bearing property prediction method introduced seismic reflection exploration technology, analyzed the underground structure through seismic wave reflection data, and initially judged the gas-bearing property by using the influence of the gas layer on wave velocity and amplitude to achieve three-dimensional imaging of the area. Seismic attribute analysis (amplitude, frequency, phase, attenuation) and amplitude variation with offset (AVO) analysis further improved the prediction reliability;

[0004] However, in the prior art, a single seismic attribute is easily affected by noise and lithology changes, and it is difficult to comprehensively reflect the complexity of the reservoir, and it is easy to ignore the heterogeneity and multi-scale effects, resulting in the technical problems of poor accuracy and low reliability in gas-bearing property prediction. Summary of the Invention

[0005] The gas-bearing property prediction method, device, storage medium, and program product provided by this application are used to achieve the effect of improving the accuracy and reliability of gas-bearing property prediction.

[0006] In a first aspect, this application provides a gas-bearing property prediction method, including:

[0007] Obtain multi-source data information of the rock formation to be predicted;

[0008] Perform data integration processing and feature extraction processing on the multi-source data information to obtain multiple input features;

[0009] Perform scale division processing on the multiple input features to obtain the input features corresponding to each scale type, where the scale types include large-scale type, medium-scale type, and small-scale type;

[0010] According to the scale type, multiple input features are sequentially input into the trained gas-bearing property prediction model, where the gas-bearing property prediction model is trained by multiple training samples, and each training sample includes corresponding multiple input feature samples and gas-bearing property label samples;

[0011] According to the gas-bearing property prediction label of the trained gas-bearing property prediction model, lithology inversion is performed to obtain the gas-bearing property prediction result of the formation to be predicted.

[0012] In a possible implementation manner, the multi-source data information includes logging information, source rock information, fault information, and seismic information;

[0013] Among them, the logging information includes gamma inversion result and P-wave impedance inversion result;

[0014] The source rock information includes coal seam thickness and gas content;

[0015] The fault information includes seismic dip attribute;

[0016] The seismic information includes frequency gradient attribute and relative geological time.

[0017] In a possible implementation manner, scale division processing is performed on multiple input features to obtain input features corresponding to each scale type, including:

[0018] The input features corresponding to the source rock information and the fault information are divided into large scale types;

[0019] The input features corresponding to the seismic information are divided into medium scale types;

[0020] The input features corresponding to the logging information are divided into small scale types.

[0021] In a possible implementation manner, before sequentially inputting multiple input features into the trained gas-bearing property prediction model according to the scale type, it further includes:

[0022] Obtain a sample data set, where the sample data set includes multi-source data information of sample formations;

[0023] According to the sample data set, determine a training data set and a test data set;

[0024] Determine an initial gas-bearing property prediction model;

[0025] According to the training data set, perform training processing on the initial gas-bearing property prediction model to obtain the trained gas-bearing property prediction model.

[0026] In a possible implementation manner, determining the initial gas-bearing property prediction model includes:

[0027] Divide the training data set according to the preset sample types to obtain the sample ratio corresponding to each sample type, where the sample types include mudstone samples, water layer samples and gas layer samples;

[0028] Determine the sample weights corresponding to the loss function according to the sample ratio corresponding to each sample type to obtain the target loss function;

[0029] Determine the initial gas-bearing prediction model according to the target loss function.

[0030] In one possible implementation, after training the initial gas-bearing prediction model according to the training data set to obtain the trained gas-bearing prediction model, it further includes:

[0031] Test the initial gas-bearing prediction model according to the test data set to verify the prediction accuracy of the trained gas-bearing prediction model.

[0032] In one possible implementation, the initial gas-bearing prediction model is a transformer model, and the transformer model includes multiple network layers composed of a multi-head self-attention mechanism and a fully connected feed-forward network;

[0033] The transformer model is used to perform scale fusion on multiple input features.

[0034] In a second aspect, the present application provides a gas-bearing prediction device, including:

[0035] An acquisition module, configured to acquire multi-source data information of the formation to be predicted;

[0036] A first processing module, configured to perform data integration processing and feature extraction processing on the multi-source data information to obtain multiple input features;

[0037] A second processing module, configured to perform scale division processing on the multiple input features to obtain the input features corresponding to each scale type, where the scale types include large scale type, medium scale type and small scale type;

[0038] A third processing module, configured to sequentially input the multiple input features into the trained gas-bearing prediction model according to the scale type, where the gas-bearing prediction model is trained by multiple training samples, and each training sample includes corresponding multiple input feature samples and gas-bearing label samples;

[0039] A prediction module, configured to perform lithology inversion according to the gas-bearing prediction label of the trained gas-bearing prediction model to obtain the gas-bearing prediction result of the formation to be predicted.

[0040] In one possible implementation, the acquisition module is further configured to:

[0041] Multi-source data information;

[0042] The multi-source data information includes logging information, source rock information, fault information, and seismic information;

[0043] Among them, the logging information includes gamma inversion results and P-wave impedance inversion results;

[0044] The source rock information includes coal seam thickness and gas content;

[0045] The fault information includes seismic dip attributes;

[0046] The seismic information includes frequency gradient attributes and relative geological time.

[0047] In a possible implementation manner, the second processing module is further configured to:

[0048] Divide the input features corresponding to the source rock information and the fault information into large-scale types;

[0049] Divide the input features corresponding to the seismic information into medium-scale types;

[0050] Divide the input features corresponding to the logging information into small-scale types.

[0051] In a possible implementation manner, the third processing module is further configured to:

[0052] Obtain a sample data set, where the sample data set includes multi-source data information of sample rock layers;

[0053] Determine a training data set and a test data set according to the sample data set;

[0054] Determine an initial gas-bearing prediction model;

[0055] Train the initial gas-bearing prediction model according to the training data set to obtain a trained gas-bearing prediction model.

[0056] In a possible implementation manner, the third processing module is further configured to:

[0057] Divide the training data set according to the preset sample types to obtain the sample ratio corresponding to each sample type, where the sample types include mudstone samples, water layer samples, and gas layer samples;

[0058] Determine the sample weights corresponding to the loss function according to the sample ratio corresponding to each sample type to obtain a target loss function;

[0059] Determine the initial gas-bearing prediction model according to the target loss function.

[0060] In a possible implementation manner, the third processing module is further configured to:

[0061] According to the test data set, the initial gas-bearing prediction model is tested to verify the prediction accuracy of the trained gas-bearing prediction model.

[0062] In a possible implementation manner, the third processing module is further configured to:

[0063] The initial gas-bearing prediction model;

[0064] The initial gas-bearing prediction model is a transformer model, and the transformer model includes a plurality of network layers composed of a multi-head self-attention mechanism and a fully connected feed-forward network;

[0065] The transformer model is used for scale fusion of a plurality of input features.

[0066] In a third aspect, the present application provides a gas-bearing prediction device, including: a memory, a processor;

[0067] The memory stores computer-executable instructions;

[0068] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementation manners of the first aspect.

[0069] In a fourth aspect, 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 above first aspect and / or various possible implementation manners of the first aspect.

[0070] In a fifth aspect, 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 above first aspect and / or various possible implementation manners of the first aspect.

[0071] A gas content prediction method, device, storage medium, and program product provided by this application can obtain multi-source data information of the rock formation to be predicted and perform data integration processing and feature extraction. This method can make full use of data from different sources, thereby improving the comprehensiveness and accuracy of data information. Moreover, by classifying multiple input features into large-scale, medium-scale, and small-scale types, it reflects the detailed analysis and hierarchical consideration of rock formation characteristics. Features at different scales reflect the characteristics of the rock formation at different spatial and temporal scales, which helps to understand the gas content law of the rock formation more deeply. Further, the gas content prediction model is trained with multiple training samples, and each training sample includes corresponding multiple input feature samples and gas content label samples. This training method enables the model to learn the complex relationship between input features and gas content, so that when facing new data of the rock formation to be predicted, it can give more accurate prediction results. In addition, according to the scale type, multiple input features are input into the trained gas content prediction model step by step, which not only helps to reduce the computational burden of the model and improve the prediction efficiency, but also can make full use of the complementarity between different scale features to further improve the accuracy and stability of the prediction. Finally, lithology inversion is performed according to the gas content prediction label of the trained gas content prediction model to obtain the gas content prediction result of the rock formation to be predicted. This step converts the complex prediction process into an intuitive and easy-to-understand prediction result, further achieving the technical effect of improving the accuracy and reliability of gas content prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] The drawings herein are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with this application, and are used together with the specification to explain the principles of this application.

[0073] Figure 1 Schematic diagram of an application data processing system architecture provided by an embodiment of this application;

[0074] Figure 2 Flow chart of the gas content prediction method provided by an embodiment of this application Figure 1 ;

[0075] Figure 3 Schematic diagram for analyzing the relationship between sandstone particle size and gas content provided by an embodiment of this application;

[0076] Figure 4 Data diagram of source rock information provided by an embodiment of this application;

[0077] Figure 5 Data diagram of seismic information provided by an embodiment of this application;

[0078] Figure 6 Schematic diagram of the structure of the gas content prediction model provided by an embodiment of this application;

[0079] Figure 7 Schematic diagram of the gas content prediction method provided by the embodiment of the present application Figure 2 ;

[0080] Figure 8 Schematic diagram of the gas content prediction method provided by the embodiment of the present application Figure 3 ;

[0081] Figure 9 Data schematic diagram of the input features provided by the embodiment of the present application;

[0082] Figure 10 Schematic diagram of the division result of the training wells and test wells provided by the embodiment of the present application;

[0083] Figure 11 Schematic diagram of the network training curve provided by the embodiment of the present application;

[0084] Figure 12 Schematic diagram of the interpretation result of the logging reservoir labels in the study area provided by the embodiment of the present application;

[0085] Figure 13 Schematic diagram of the network structure of the transformer model provided by the embodiment of the present application;

[0086] Figure 14 Schematic diagram of the gas content classification result of the test blind well provided by the embodiment of the present application;

[0087] Figure 15 Schematic diagram of the test of the gas content classification result of the cross-well profile provided by the embodiment of the present application;

[0088] Figure 16 Schematic diagram of the model interpretable analysis and feature sensitivity analysis provided by the embodiment of the present application;

[0089] Figure 17 Schematic diagram of the gas reservoir plane characterization result provided by the embodiment of the present application;

[0090] Figure 18 Schematic diagram of the structure of the gas content prediction device provided by the embodiment of the present application;

[0091] Figure 19 Schematic diagram of the structure of the gas content prediction device provided by the embodiment of the present application.

[0092] Through the above-mentioned drawings, the clear embodiments of the present application have been shown, and there will be more detailed descriptions hereinafter. These drawings and the textual 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 implementation manners

[0093] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying 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.

[0094] In the prior art, a single seismic attribute is vulnerable to noise and lithology changes, making it difficult to comprehensively reflect the reservoir complexity, and it is easy to overlook heterogeneity and multi-scale effects. There are technical problems such as poor accuracy and low reliability in gas-bearing prediction.

[0095] In view of the above problems, a gas-bearing prediction method, device, storage medium, and program product provided by the present application obtain multi-source data information of the rock formation to be predicted, and perform data integration processing and feature extraction. This method can make full use of data from different sources, thereby improving the comprehensiveness and accuracy of data information. Moreover, multiple input features are processed by scale division, divided into large-scale, medium-scale, and small-scale types, which reflects the detailed analysis and hierarchical consideration of the rock formation characteristics. Features of different scales reflect the characteristics of the rock formation at different spatial and temporal scales, helping to more deeply understand the gas-bearing law of the rock formation. Further, the gas-bearing prediction model is trained with multiple training samples, and each training sample includes corresponding multiple input feature samples and gas-bearing label samples. This training method enables the model to learn the complex relationship between the input features and the gas-bearing property, so that when facing new data of the rock formation to be predicted, it can give a more accurate prediction result. In addition, according to the scale type, multiple input features are sequentially input into the trained gas-bearing prediction model, which not only helps to reduce the computational burden of the model and improve the prediction efficiency, but also can make full use of the complementarity between different scale features to further improve the accuracy and stability of the prediction. Finally, lithology inversion is performed according to the gas-bearing prediction label of the trained gas-bearing prediction model to obtain the gas-bearing prediction result of the rock formation to be predicted. This step transforms the complex prediction process into an intuitive and easy-to-understand prediction result, further achieving the technical effect of improving the accuracy and reliability of gas-bearing prediction.

[0096] 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.

[0097] Figure 1This is a schematic diagram of an application data processing system architecture provided by an embodiment of the present application. The application data processing system is a computer device. As Figure 1 shown, the above architecture includes at least one of a data acquisition device 101, a processing device 102, and a display device 103.

[0098] It can be understood that the structure schematically shown in the embodiments of the present application does not constitute a specific limitation on the architecture of the application data processing system. In other feasible embodiments of the present application, the above architecture may include more or fewer components than those shown in the figure, or combine certain components, or split certain components, or have different component arrangements, which can be specifically determined according to the actual application scenario and will not be limited herein. Figure 1 The components shown can be implemented in hardware, software, or a combination of software and hardware.

[0099] In the specific implementation process, the data acquisition device 101 may include an input / output interface or a communication interface. The data acquisition device 101 can be connected to the processing device through the input / output interface or the communication interface.

[0100] The processing device 102 can obtain multi-source data information of the rock formation to be predicted; perform data integration processing and feature extraction processing on the multi-source data information to obtain multiple input features; perform scale division processing on the multiple input features to obtain the input features corresponding to each scale type, where the scale types include large-scale type, medium-scale type, and small-scale type; according to the scale type, input the multiple input features into the trained gas-bearing prediction model level by level, where the gas-bearing prediction model is trained by multiple training samples, and each training sample includes corresponding multiple input feature samples and gas-bearing label samples; according to the gas-bearing prediction label of the trained gas-bearing prediction model, perform lithology inversion to obtain the gas-bearing prediction result of the rock formation to be predicted.

[0101] The display device 103 can also be a touch display screen or the screen of a terminal device, and is used to receive user instructions while displaying the above content to achieve interaction with the user.

[0102] It should be understood that the above processing device can be implemented by a processor reading instructions in a memory and executing the instructions, or can also be implemented by a chip circuit.

[0103] In addition, the network architecture and service scenarios described in the embodiments of the present application are for more clearly explaining the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those of ordinary skill in the art know that with the evolution of the network architecture and the emergence of new service scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0104] Figure 2 Flow schematic of the gas content prediction method provided by the embodiment of the present application Figure 1 , such as Figure 2 shown, the gas content prediction method provided by this embodiment includes:

[0105] S201. Obtain multi-source data information of the formation to be predicted;

[0106] Optionally, the multi-source data information includes logging information, source rock information, fault information, and seismic information; among them, the logging information includes gamma inversion results and P-wave impedance inversion results; the source rock information includes coal seam thickness and gas content; the fault information includes seismic dip attributes; the seismic information includes frequency gradient attributes and relative geological time.

[0107] Specifically, obtain various geological and geophysical information of the formation to be predicted, namely logging information (gamma inversion results and P-wave impedance inversion results), source rock information (coal seam thickness and gas content), fault information (seismic dip attributes), and seismic information (frequency gradient attributes and relative geological time).

[0108] S202. Perform data integration processing and feature extraction processing on the multi-source data information to obtain multiple input features;

[0109] Only perform data integration and feature extraction on the above-mentioned logging information, source rock information, fault information, and seismic information, and use the extracted multiple feature data as input features.

[0110] Specifically, for the logging information, Figure 3 Schematic diagram for analyzing the relationship between sandstone particle size and gas content provided by the embodiment of the present application, where Figure 3 a is the distribution diagram of gamma inversion results (GR) values of sandstones with different particle sizes, Figure 3 b is the distribution diagram of gas layer probabilities of sandstones with different particle sizes, Figure 3 c is the single-well histogram of Well A1, as Figure 3 shown, Figure 3 a and Figure 3 b statistically show the distribution of GR values of sandstones with different particle sizes and the distribution of gas layer probabilities of sandstones with different particle sizes in the target interval. It can be seen from the figure that compared with relatively fine-silty sandstones, the medium-coarse sandstones have larger particle sizes, smaller gamma inversion result GR values, and higher gas-bearing probabilities. It is speculated that under stronger hydrodynamic conditions, the transported particles are relatively larger, and the deposited sandstone particle sizes are also larger. Strong hydrodynamic forces can carry fine particles such as mud away from the deposition area, thereby reducing the proportion of mud mixed into the sandstone. The decrease in mud content is manifested as a decrease in the GR value in geophysical records. At the same time, sandstones with lower mud content are more resistant to compaction and cementation during deposition and diagenesis, and the primary pores are better preserved. Therefore, the reservoir physical properties are relatively better. Taking Well A1 as an example, Figure 3The lithology distribution, porosity distribution, gamma curve and log interpretation results of Well A1 are shown in c. The target reservoir is located at 1560 - 1580 m, and the main lithologies are medium sandstone, argillaceous siltstone and mudstone. The physical properties of different lithologies are also different. It can be clearly seen that the particle size is positively correlated with the physical properties, and the medium sandstone is interpreted as a gas layer, which is consistent with the results of the previous sensitive attribute analysis. Therefore, the gamma inversion result (Gr) is selected as the input of the model.

[0111] It should be noted that Figure 3 It is only for reference of the display effect, not an improvement point, and does not affect the protection scope of the embodiments of the present application.

[0112] For the source rock information and fault information, Figure 4 This is a data schematic diagram of the source rock information provided by the embodiments of the present application, where Figure 4 a represents the bottom map of the well-connected section, Figure 4 b represents the gas content distribution map of the source rock, Figure 4 c represents the dip angle attribute map of the target layer, Figure 4 d represents the seismic profile passing through Well A2 - Well A6, Figure 4 e represents the reservoir profile passing through Well A2 - Well A6. As Figure 4 shown, taking the data of a high water cut tight sandstone reservoir in a certain work area as an example, select a well-connected section as shown in Figure 4 a and Figure 4 d. This seismic data has 817 traces, and each trace has 200 time sampling points. The thickness of the target layer is 100 ms, and through log analysis, gas layers and gas-bearing water layers are more developed in this target layer, and a set of stable source rocks are developed in the lower part. At the same time, there are 5 wells (A2, A3, A4, A5, A6) on this seismic profile that all vertically pass through the target layer.

[0113] It should be noted that Figure 4 It is only for reference of the display effect, not an improvement point, and does not affect the protection scope of the embodiments of the present application.

[0114] The main controlling factors for the gas content difference are analyzed as the gas content and thickness of the lower source rock and whether there is a major fault near the tight sandstone reservoir that can communicate with the lower source rock. Figure 4 b shows the gas content distribution of the source rock below the target layer. The thickness of the lower coal seam of Well A4 is 14 m, and the gas content reaches 9.3 t / m³. And there is an obvious strike-slip fault zone near Well A4. The major fault communicates the lower source rock and the upper tight sandstone reservoir, providing a good channel for gas migration. Figure 4c), affected by its gas supply capacity and large fractures, significant gas production effects were achieved during the trial production of the overlying target layer, with the absolute open flow rate reaching 19,800 m³ / day. This indicates that under the gas supply conditions of relatively high gas content and thick coal seams, coupled with the migration effect of large fractures, the upper tight sandstone has good gas accumulation capacity and production potential.

[0115] In contrast, in Well A3, the thickness of the coal seam is only 8 m and the gas content is 8.2 t / m³, showing relatively weak gas supply capacity, and there are no fractures nearby that can communicate the lower source rock and reservoir. The corresponding overlying target layer showed a water layer during gas testing, and no effective gas flow was observed. This further verifies the importance of the influence of coal seam thickness, gas content, and fracture migration in controlling the gas content of tight sandstone. Therefore, the coal seam thickness and gas content in the source rock information and the seismic dip angle attribute in the fault information are selected as the inputs of the model.

[0116] Regarding the seismic information, Figure 5 is a data schematic diagram of the seismic information provided by the embodiment of the present application. Among them, Figure 5 a is the seismic data, Figure 5 b is the frequency gradient result. As Figure 5 shown, the profile of the prestack data is Figure 5 a. Among them, the in-well data represents the gas content indicated by the total hydrocarbon in gas logging. It can be determined that only the resolution of the seismic data at the gas layer position is insufficient, and the gas layer identification effect is poor. Figure 5 As shown in b, it represents the frequency gradient attribute obtained by introducing the high-precision time-frequency analysis method. The resolution is significantly improved and there is a certain correspondence with the gas layer. Therefore, the frequency gradient attribute and relative geological time in the seismic information are selected as the inputs of the model.

[0117] It should be noted that Figure 5 is only for reference of the display effect, not an improvement point, and does not affect the protection scope of the embodiment of the present application.

[0118] S203. Perform scale division processing on multiple input features to obtain the input features corresponding to each scale type, where the scale types include large scale type, medium scale type, and small scale type;

[0119] Since the input features contain information of different spaces and different scales, the input features are divided into large-scale features, medium-scale features, and small-scale features according to the scale.

[0120] S204. According to the scale type, input multiple input features into the trained gas-bearing prediction model step by step, where the gas-bearing prediction model is trained by multiple training samples, and each training sample includes corresponding multiple input feature samples and gas-bearing label samples;

[0121] Adopt a multi-scale step-by-step learning strategy. According to the scale types of each input feature, each input feature is gradually input into the trained gas-bearing property prediction model.

[0122] S205. Perform lithology inversion based on the gas-bearing property prediction labels of the trained gas-bearing property prediction model to obtain the gas-bearing property prediction results of the formation to be predicted.

[0123] In this embodiment, the gas-bearing property prediction label refers to the logging interpretation result.

[0124] According to the logging interpretation results of the trained gas-bearing property prediction model, using lithology information as a mask, perform lithology inversion, and then obtain the gas-bearing property prediction results of the formation to be predicted.

[0125] Specifically, Figure 6 is a schematic structural diagram of the gas-bearing property prediction model provided by the embodiment of the present application. As Figure 6 shown, where Transformer represents a converter model. After each input feature is gradually input into the trained gas-bearing property prediction model, the network first learns the feature information of large scales, introduces the information of medium-scale features through feature fusion, and fuses the gas-bearing property features under the constraint of seismic sequences on the basis of geological information indicating favorable gas-bearing areas to achieve complementarity between scales. On this basis, fuse the high-resolution inversion information of small scales to further achieve the fusion and complementarity of scales. Finally, the gas-bearing property results of sandstone output by the model are masked and constrained with the lithology inversion data results (mudstone prediction results and sandstone gas-bearing property results), and finally the final gas-bearing property prediction results, that is, the sandstone gas-bearing property results, are obtained by matching.

[0126] It should be noted that the lithology inversion data is pre-trained data, and the mudstone prediction results and sandstone gas-bearing property results are both obtained based on the results of the lithology inversion data. The lithology inversion results are obtained by using a gated recurrent unit (GRU) network according to gamma inversion results and impedance. The lithology inversion results altogether include mudstone, sandstone and coal seams. After analysis, the coal seam thickness as part of the hydrocarbon source rock is used as a constraint, and mudstone and sandstone are lithology masked and constrained in the final gas-bearing property prediction model. Finally, only the results of gas layers and water layers are retained in the sandstone part.

[0127] Figure 6 It is only for reference of the display effect, not an improvement point, and does not affect the protection scope of the embodiment of the present application.

[0128] A gas content prediction method provided by this application, by obtaining multi-source data information of the formation to be predicted and performing data integration processing and feature extraction, this method can make full use of data from different sources, thereby improving the comprehensiveness and accuracy of data information; and, performing scale division processing on multiple input features, dividing them into large-scale, medium-scale and small-scale types, reflecting the detailed analysis and hierarchical consideration of formation characteristics, and the characteristics of different scales reflect the characteristics of the formation at different spatial and temporal scales, which helps to more deeply understand the gas content law of the formation; further, the gas content prediction model is trained through multiple training samples, and each training sample includes corresponding multiple input feature samples and gas content label samples. This training method enables the model to learn the complex relationship between the input features and the gas content, so that when facing new formation data to be predicted, it can give more accurate prediction results; in addition, according to the scale type, multiple input features are input into the trained gas content prediction model step by step, which not only helps to reduce the computational burden of the model and improve the prediction efficiency, but also can make full use of the complementarity between different scale features to further improve the accuracy and stability of the prediction. Finally, performing lithology inversion according to the gas content prediction label of the trained gas content prediction model to obtain the gas content prediction result of the formation to be predicted. This step converts the complex prediction process into an intuitive and easy-to-understand prediction result, further achieving the technical effect of improving the accuracy and reliability of gas content prediction.

[0129] Figure 7 Flow schematic of the gas content prediction method provided by the embodiments of this application Figure 2 , such as Figure 7 shown, on the basis of the above embodiments, this embodiment details the acquisition process of the input features corresponding to each scale type, including:

[0130] S701. Divide the input features corresponding to source rock information and fault information into large-scale types;

[0131] In this embodiment, the input features of the large-scale type mainly refer to the coal seam thickness, coalbed methane content and dip angle attributes obtained based on geological analysis, which mainly reflect the range of gas-bearing favorable areas indicated by hydrocarbon generation of source rocks and fracture channels.

[0132] Divide the input features including source rock information and fault information into large-scale types.

[0133] S702. Divide the input features corresponding to seismic information into medium-scale types;

[0134] In this embodiment, the input features of the medium-scale type include relative geological time attributes and frequency gradient attributes obtained from seismic data, and this feature reflects the gas-bearing favorable areas under sequence constraints.

[0135] Divide the input features containing seismic information into meso-scale types.

[0136] S703. Divide the input features corresponding to logging information into small-scale types.

[0137] In this embodiment, the input features of the small-scale type include impedance information and gamma information obtained by high-resolution seismic inversion, providing higher-precision gas-bearing information.

[0138] Divide the input features containing logging information into small-scale types.

[0139] The gas-bearing prediction method provided by the embodiments of this application can, by integrating source rock and fault information, more comprehensively understand the geological background of oil and gas reservoirs, improve the macroscopic accuracy of gas-bearing prediction. At the same time, the extraction of large-scale features helps to identify potential oil and gas enrichment areas, providing important guidance for subsequent exploration and development; the utilization of seismic information can improve the spatial resolution of gas-bearing prediction, help identify the boundaries and internal structures of oil and gas reservoirs, and through the extraction of meso-scale features, the characteristics of oil and gas reservoirs can be described more precisely, improving the accuracy and reliability of prediction; the utilization of logging information can further improve the accuracy of gas-bearing prediction, and through the extraction of small-scale features, the presence and properties of oil and gas layers can be identified more accurately, providing a direct basis for oil and gas exploration and development. At the same time, logging information can also be combined with features of other scales to form a multi-scale comprehensive prediction method, improving the overall prediction effect, thereby achieving the technical effect of improving the accuracy and reliability of gas-bearing prediction.

[0140] Figure 8 Schematic diagram of the process of the gas-bearing prediction method provided by the embodiments of this application Figure 3 , as Figure 8 shown, on the basis of the above embodiments, this embodiment details the training process of the initial gas-bearing prediction model, including:

[0141] S801. Obtain a sample data set;

[0142] In this embodiment, the sample data set includes multi-source data information of sample rock formations.

[0143] Specifically, Figure 9 is a schematic diagram of multi-source data information provided by the embodiments of this application. As Figure 9 shown, obtain a sample data set including multi-source data information of sample rock formations, where Figure 9 a is the gamma inversion result, Figure 9 b is the relative geological time, Figure 9 c is the lithology inversion data result for lithology constraint in the above embodiment, Figure 9 d is the longitudinal wave impedance inversion result, Figure 9 e is the seismic dip attribute,Figure 9 f is the coal seam thickness, Figure 9 g is the frequency gradient attribute, Figure 9 h is the gas content.

[0144] It should be noted that, Figure 9 it is only for reference of the display effect, not an improvement point, and does not affect the protection scope of the embodiments of this application.

[0145] S802. Determine the training data set and the test data set according to the sample data set;

[0146] Specifically, Figure 10 as shown in the schematic diagram of the division result of the training wells and test wells provided by the embodiments of this application, Figure 10 taking 20 wells in the study area corresponding to a certain sample data set as an example, randomly select 4 of them as test blind wells (A19, A16, A3, A6), determine the corresponding test data set, and the remaining wells are training wells to determine the corresponding training data set. Among them, the tick represents the test blind well.

[0147] It should be noted that, Figure 10 it is only for reference of the display effect, not an improvement point, and does not affect the protection scope of the embodiments of this application.

[0148] Furthermore, Figure 11 as shown in the schematic diagram of the network training curve provided by the embodiments of this application, by analyzing the logging data, it is determined that the sample of gas layer data is much smaller than the sample of water layer data, which will lead to the problem of unbalanced learning during the training of the model. Therefore, set the sample weight w c = [0.1, 0.1, 0.8], that is, the weight of mudstone is 10%, the weight of water layer is 10%, and the weight of gas layer is 80%. To balance the samples of gas layer and water layer, 60% of the data is set as the training set and 40% of the data is set as the validation set during network training. The number of network training times is 1500, the learning rate is 0.001, and the optimizer uses the weight decay correction (Adamw) optimizer. Its training curve is as Figure 11 shown, where the blue curve represents the training loss and the orange curve represents the validation loss.

[0149] S803. Divide the training data set according to the preset sample types to obtain the sample ratio corresponding to each sample type;

[0150] In this embodiment, the sample types include mudstone samples, water layer samples and gas layer samples.

[0151] Specifically, Figure 12Schematic diagram of the logging reservoir label interpretation results in the study area provided by the embodiments of the present application. Among them, the gas layer includes the first-class gas layer, the second-class gas layer and the water-bearing gas layer, and the water layer includes the water layer, the gas-water co-existing layer and the gas-bearing water layer, as Figure 12 shown. Based on the existing logging data and reservoir interpretation results in the study area, the labels are divided into three categories, namely: gas layer, water layer, and mudstone. According to the above three types, the sample types of the training data set are divided to obtain the sample ratio corresponding to each sample type.

[0152] It should be noted that Figure 12 only for reference of the display effect, not as an improvement point, and does not affect the protection scope of the embodiments of the present application.

[0153] S804. Determine the sample weights corresponding to the loss function according to the sample ratio corresponding to each sample type to obtain the target loss function;

[0154] According to the sample ratio corresponding to each sample type, further determine the sample weights corresponding to the loss function, and through the following formula, to obtain the target loss function:

[0155]

[0156] Among them, L weighted is the weighted loss function, that is, the target loss function. C identifies the categories of the gas-bearing labels, and there are three categories: water layer, gas layer and mudstone. y i,c represents the true label of the i-th sample belonging to category C. p i,c represents the predicted probability of the i-th sample belonging to category C. w c represents the category weight.

[0157] It should be noted that compared with the water layer and mudstone labels, the gas layer labels are too few and seriously unbalanced. Therefore, the weighted loss function is used to give a greater weight to the minority class samples, thereby improving the sensitivity of the model to the gas layer samples.

[0158] S805. Determine the initial gas-bearing prediction model according to the target loss function;

[0159] Since the target loss function belongs to the weighted loss function, according to the weighted loss function, the initial gas-bearing prediction model is determined as the transformer model.

[0160] Optionally, the initial gas-bearing prediction model is the transformer model. The transformer model includes multiple network layers composed of the multi-head self-attention mechanism and the fully connected feed-forward network; the transformer model is used to perform scale fusion on multiple input features.

[0161] Specifically, Figure 13 Schematic diagram of the network structure of the transformer model provided by the embodiments of the present application, asFigure 13 As shown, the Transformer model is a deep learning model that relies on the self-attention mechanism; the network layer consists of a multi-head self-attention mechanism and a fully connected feed-forward network. Residual connections are used around the two sub-layers, and then the activation values of each layer are normalized;

[0162] The input data is based on a position encoding module to extract the spatial position information of the embedded input data at one time, facilitating parallel computing of the model. The calculation is as follows:

[0163]

[0164] where pos represents the position; represents the dimension.

[0165] The attention function is a function that maps a query vector Q and some key vectors K into a weighted sum of an output V numerical vector. Among them, the weight is obtained from the similarity of the dot product of K and Q, and then obtained by the Softmax function. The specific calculation is as follows:

[0166]

[0167] The multi-head attention mechanism projects the original query vector Q, numerical vector V, and key vector K into low dimensions through a linear layer, and then matches the attention mechanism of h dot products to obtain similarity functions of different patterns. Finally, the output vectors are merged and projected back, enabling the model to extract global features. The specific formula is as follows:

[0168]

[0169]

[0170] where, , , are the parameter matrices of the projection respectively.

[0171] The output vector can be calculated by two linear transformations of the fully connected feed-forward network. The specific formula is as follows:

[0172]

[0173] By stacking multiple Transformer network layers, the mapping relationship between multi-attribute features and gas-bearing properties can be better learned.

[0174] It should be noted that, Figure 13 This is only for reference in showing the effect, not an improvement point, and does not affect the protection scope of the embodiments of this application.

[0175] S806. Train the initial gas-bearing property prediction model according to the training data set to obtain a trained gas-bearing property prediction model.

[0176] According to the training data set, perform training processing on the initial gas-bearing property prediction model through steps such as data preprocessing, feature extraction, and parameter optimization to obtain a trained gas-bearing property prediction model.

[0177] It should be noted that during the oil and gas exploration process, the interpretation of logging data is crucial for identifying reservoir types. The gas logging curves of gas layers have greater differences compared to water layers and gas-water layers. Therefore, it is often difficult to draw accurate conclusions solely based on gas logging curves. To obtain more reliable interpretation results, it is necessary to combine gas logging curves with other logging information for comprehensive analysis. These information includes but is not limited to parameters such as acoustic travel time, resistivity, density, and gamma ray, which can provide detailed data on the physical properties of rock formations. Based on this multi-dimensional data, an artificial intelligence model can effectively establish the relationship between multi-source information and reservoir types. Through machine learning algorithms, the model can learn the complex relationships between different logging parameters and reservoir types from a large amount of historical data, thereby obtaining an accurate and reliable gas-bearing property prediction model.

[0178] S807. Test the initial gas-bearing property prediction model according to the test data set to verify the prediction accuracy of the trained gas-bearing property prediction model.

[0179] Specifically, Figure 14 is a schematic diagram of the gas-bearing property classification results of the test blind wells provided in the embodiments of this application. As Figure 14 shown, according to the 4 blind wells (A19, A16, A3, A6) corresponding to the test data set, test the initial gas-bearing property prediction model. The gas layer classification accuracies are 93%, 82%, 83%, and 93% respectively.

[0180] Furthermore, Figure 15 is a schematic diagram of the test of the gas-bearing property classification results of the cross-well profile provided in the embodiments of this application. As Figure 15 shown, based on the single-well test, a cross-well profile test is carried out. Draw a cross-well profile according to the well positions in the study area, and verify the prediction accuracy of the trained gas-bearing property prediction model from the prediction results of the cross-well profile. The coincidence rate between the prediction results of the cross-well profile and the well data has reached more than 80%. The prediction results of the cross-well profile can accurately reflect the distribution, thickness, and geological characteristics of gas layers and water layers, and have a high degree of coincidence with the reservoir interpretation information in the actual well data. In addition to the high coincidence rate, the inversion results of the cross-well profile prediction also conform to geological laws and are consistent with the prior understanding of upper gas and lower water.

[0181] It should be noted that when predicting the gas content of an artificial intelligence model, it is also necessary to analyze the sensitivity of features and the interpretability of the model, that is, use the Shapley Additive Explanation model (SHAP model) to perform interpretability analysis on the Transformer model;

[0182] Figure 16 FIG. is a schematic diagram of model interpretability analysis and feature sensitivity analysis provided by an embodiment of the present application. As Figure 16 shown, the results of feature sensitivity analysis by the SHAP model indicate that geological information (faults, source rocks, and sequences) has an important impact on the prediction results. Figure 16 mainly includes two scatter plots of SHAP feature importance analysis and a bar chart of feature rankings. Among them, RGT represents the relative geological age body, TIC represents the thickness of the source rock, DIP represents the formation dip attribute, THL represents the low-frequency spatial trend of gas content, and FG represents the frequency gradient.

[0183] First, the scatter plots respectively show the distribution of SHAP values of different geological features in the prediction of water layers and gas layers. The horizontal axis represents the SHAP value, which represents the impact of the feature on the model output, and the vertical axis is the specific feature name (such as RGT, TIC, DIP, etc.). The color of the scatter points represents the high or low value of the feature, with red indicating high feature values and blue indicating low feature values. It can be determined from the SHAP diagrams of water layer prediction and gas layer prediction that features such as RGT, TIC, and DIP have a greater impact on the prediction results, and at the same time, the positive and negative impacts on the model output show certain regularity under different feature values.

[0184] Secondly, the bar chart on the right shows the F-value score ranking of each feature by the model, which is used to quantify the importance of the feature. The results show that features such as GR, RGT, and TIC have the highest importance scores and contribute the most to the prediction effect of the model, while features such as THL and FG have slightly lower importance but still have a significant impact.

[0185] Combining these analysis results illustrates the key role of geological information, especially when distinguishing different sedimentary environments or geological features. For water layer prediction and gas layer prediction, features such as DIP and RGT play a particularly obvious role in different models, and their importance is consistently reflected in the SHAP value and F-value rankings. This shows that the interpretability analysis of the model can provide a scientific basis for the selection and optimization of geological information, and at the same time verifies that the combined effect of multiple geological features can significantly improve the prediction accuracy.

[0186] Figure 17 FIG. is a schematic diagram of the gas layer plane characterization result provided by an embodiment of the present application. Among them, Figure 17 a is the plane characterization map of the gas content in the S7 section, Figure 17 b is the plane characterization map of the gas content in the S8 section. As Figure 17As shown in the figure, according to the accurate horizon results (S7 and S8 sections) in the study area, the planar distribution of the target gas layer in the study area can be obtained. By comparing the drilling statistics results (the bar chart in the figure, red represents gas layer and blue represents water layer) with the planar results, the planar coincidence rate of the S7 section is 89%, and that of the S8 section is 84%. Therefore, it is determined that the prediction results have good accuracy.

[0187] It should be noted that Figures 14 - 17 It is only for reference of the display effect, not an improvement point, and does not affect the protection scope of the embodiments of the present application.

[0188] The gas-bearing prediction method provided by the embodiments of the present application obtains a sample data set containing multi-source data information, makes full use of rock formation data from different sources, and improves the comprehensiveness and accuracy of the prediction model. In addition, the sample data set is divided into a training data set and a test data set, and the sample ratio is divided according to the sample type (such as mudstone, water layer, gas layer), which helps the model learn the characteristics of different sample types during the training process. By determining the sample weights corresponding to the loss function, the rarity or importance of different sample types can be adjusted to further optimize the prediction performance of the model. Further, a transformer model is used as the initial gas-bearing prediction model. This model has strong feature extraction and scale fusion capabilities. The multi-head self-attention mechanism and the fully connected feed-forward network in the transformer model can process complex input feature relationships, thereby improving the prediction accuracy and generalization ability of the model. At the same time, the model is trained with the training data set and verified with the test data set, which can ensure the stability and reliability of the model in practical applications. Therefore, through the above steps, through fine data processing, model selection and training strategies, the accuracy of gas-bearing prediction is significantly improved, and the technical effect of improving the accuracy and reliability of gas-bearing prediction is achieved.

[0189] Figure 18 It is a schematic structural diagram of the gas-bearing prediction device provided by the embodiments of the present application. The device of this embodiment can be in the form of software and / or hardware. As Figure 18 As shown, the gas-bearing prediction device 1800 provided by the embodiments of the present application includes: an acquisition module 1801, a first processing module 1802, a second processing module 1803, a third processing module 1804, and a prediction processing module 1805:

[0190] The acquisition module 1801 is used to acquire multi-source data information of the rock formation to be predicted;

[0191] The first processing module 1802 is used to perform data integration processing and feature extraction processing on the multi-source data information to obtain a plurality of input features;

[0192] The second processing module 1803 is configured to perform scale division processing on multiple input features to obtain input features corresponding to each scale type, where the scale types include large scale type, medium scale type, and small scale type;

[0193] The third processing module 1804 is configured to input multiple input features into the trained gas-bearing prediction model step by step according to the scale type, where the gas-bearing prediction model is trained by multiple training samples, and each training sample includes corresponding multiple input feature samples and gas-bearing label samples;

[0194] The prediction module 1805 is configured to perform lithology inversion according to the gas-bearing prediction label of the trained gas-bearing prediction model to obtain the gas-bearing prediction result of the formation to be predicted.

[0195] In a possible implementation manner, the acquisition module 1801 is further configured to:

[0196] Multi-source data information;

[0197] The multi-source data information includes logging information, source rock information, fault information, and seismic information;

[0198] Among them, the logging information includes gamma inversion result and P-wave impedance inversion result;

[0199] The source rock information includes coal seam thickness and gas content;

[0200] The fault information includes seismic dip attribute;

[0201] The seismic information includes frequency gradient attribute and relative geological time.

[0202] In a possible implementation manner, the second processing module 1803 is further configured to:

[0203] Divide the input features corresponding to the source rock information and the fault information into the large scale type;

[0204] Divide the input features corresponding to the seismic information into the medium scale type;

[0205] Divide the input features corresponding to the logging information into the small scale type.

[0206] In a possible implementation manner, the third processing module 1804 is further configured to:

[0207] Obtain a sample data set, where the sample data set includes multi-source data information of sample formations;

[0208] Determine a training data set and a test data set according to the sample data set;

[0209] Determine an initial gas-bearing prediction model;

[0210] According to the training data set, the initial gas-bearing prediction model is trained to obtain a trained gas-bearing prediction model.

[0211] In a possible implementation manner, the third processing module 1804 is further configured to:

[0212] According to the preset sample types, the training data set is divided into sample types to obtain the sample ratio corresponding to each sample type, where the sample types include shale samples, water layer samples, and gas layer samples;

[0213] According to the sample ratio corresponding to each sample type, the sample weights corresponding to the loss function are determined to obtain a target loss function;

[0214] According to the target loss function, the initial gas-bearing prediction model is determined.

[0215] In a possible implementation manner, the third processing module 1804 is further configured to:

[0216] According to the test data set, the initial gas-bearing prediction model is tested to verify the prediction accuracy of the trained gas-bearing prediction model.

[0217] In a possible implementation manner, the third processing module 1804 is further configured to:

[0218] The initial gas-bearing prediction model;

[0219] The initial gas-bearing prediction model is a transformer model, and the transformer model includes a plurality of network layers composed of a multi-head self-attention mechanism and a fully connected feed-forward network;

[0220] The transformer model is used for scale fusion of multiple input features.

[0221] The gas-bearing prediction device provided in this embodiment can execute the method provided in the above method embodiment, and its implementation principle and technical effects are similar, which will not be elaborated here in this embodiment.

[0222] Figure 19 It is a schematic structural diagram of a gas-bearing prediction device provided in an embodiment of the present application. As Figure 19 shown, the electronic device 1900 provided in this embodiment includes: at least one processor 1901 and a memory 1902. Optionally, the device 1900 further includes a communication component 1903. Among them, the processor 1901, the memory 1902, and the communication component 1903 are connected through a bus.

[0223] In a specific implementation process, at least one processor 1901 executes computer-executable instructions stored in a memory 1902, so that at least one processor 1901 executes the above-mentioned method.

[0224] For the specific implementation process of the processor 1901, reference may be made to the above method embodiment. The implementation principle and technical effects are similar, and will not be elaborated here in this embodiment.

[0225] In the above embodiment, it should be understood that the processor may be a central processing unit (Central Processing Unit, CPU for short), or may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP for short), application specific integrated circuits (Application Specific Integrated Circuit, ASIC for short), etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the invention can be directly embodied as being completed by a hardware processor, or can be completed by a combination of hardware and software modules in the processor.

[0226] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (Non-volatile Memory, NVM), such as at least one disk memory.

[0227] 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 sake of convenience in representation, the buses in the drawings of this application are not limited to only one bus or one type of bus.

[0228] The embodiment of the present application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the above-mentioned method is implemented.

[0229] The embodiment of the present application also provides a computer-readable storage medium, in which computer-executable instructions are stored, and when the processor executes the computer-executable instructions, the above-mentioned method is implemented.

[0230] 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.

[0231] An exemplary readable storage medium is coupled to the processor so that the processor can read information from the readable storage medium 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.

[0232] 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.

[0233] The units described as separate components may or may not be physically separated, and 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.

[0234] In addition, in each embodiment of the present invention, the functional units 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.

[0235] 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 of the 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.

[0236] 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.

[0237] 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 the common general knowledge or conventional technical means in the technical field not disclosed in the present invention. It is not limited to the exact structures 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 gas content prediction method, characterized in that, Comprising: Obtaining multi-source data information of a rock formation to be predicted; Performing data integration processing and feature extraction processing on the multi-source data information to obtain a plurality of input features; Performing scale division processing on the plurality of input features to obtain input features corresponding to each scale type, wherein the scale type includes a large-scale type, a medium-scale type, and a small-scale type; According to the scale type, sequentially inputting the plurality of input features into a trained gas-bearing property prediction model, wherein the gas-bearing property prediction model is obtained by training with a plurality of training samples, and each of the training samples includes corresponding multiple input feature samples and gas-bearing property label samples; Performing lithology inversion according to the gas-bearing property prediction label of the trained gas-bearing property prediction model to obtain a gas-bearing property prediction result of the rock formation to be predicted.

2. The method according to claim 1, wherein The multi-source data information includes logging information, hydrocarbon source rock information, fault information, and seismic information; Wherein, the logging information includes gamma inversion results and P-wave impedance inversion results; The hydrocarbon source rock information includes coal seam thickness and gas content; The fault information includes seismic dip attributes; The seismic information includes frequency gradient attributes and relative geological time.

3. The method according to claim 2, wherein The performing scale division processing on the plurality of input features to obtain input features corresponding to each scale type includes: Dividing the input features corresponding to the hydrocarbon source rock information and the fault information into the large-scale type; Dividing the input features corresponding to the seismic information into the medium-scale type; Dividing the input features corresponding to the logging information into the small-scale type.

4. The method according to any one of claims 1 to 3, characterized in that, Before sequentially inputting the plurality of input features into the trained gas-bearing property prediction model according to the scale type, it further includes: Obtaining a sample data set, wherein the sample data set includes multi-source data information of a sample rock formation; Determining a training data set and a test data set according to the sample data set; Determining an initial gas-bearing property prediction model; Performing training processing on the initial gas-bearing property prediction model according to the training data set to obtain a trained gas-bearing property prediction model.

5. The method according to claim 4, characterized in that, The determining the initial gas-bearing property prediction model includes: Performing sample type division on the training data set according to a preset sample type to obtain a sample ratio corresponding to each sample type, wherein the sample type includes mudstone samples, water layer samples, and gas layer samples; Determining sample weights corresponding to a loss function according to the sample ratio corresponding to each sample type to obtain a target loss function; Determining an initial gas-bearing property prediction model according to the target loss function.

6. The method according to claim 4, characterized in that, After performing training processing on the initial gas-bearing property prediction model according to the training data set to obtain a trained gas-bearing property prediction model, it further includes: Performing test processing on the initial gas-bearing property prediction model according to the test data set to verify the prediction accuracy of the trained gas-bearing property prediction model.

7. The method according to claim 4, characterized in that, The initial gas-bearing property prediction model is a transformer model, and the transformer model includes a plurality of network layers composed of a multi-head self-attention mechanism and a fully-connected feed-forward network; The transformer model is used for performing scale fusion on the plurality of input features.

8. An aeration property prediction device, characterized in that, Comprising: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the gas content prediction method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, Computer-executable instructions are stored in the computer-readable storage medium, and when the computer-executable instructions are executed by a processor, they are used to implement the gas content prediction method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the gas content prediction method according to any one of claims 1 to 7.

Images