Coal seam gas content prediction method, device, equipment, storage medium and product

Abstract

Embodiments of the present application provide a coal seam gas content prediction method, device, equipment, storage medium and product. The method comprises obtaining a corresponding relationship between different source multi-scale data of a target area and coal seam gas content, and preparing a plurality of cross-plot diagrams according to the corresponding relationship; determining main control factor data of the coal seam gas content according to the plurality of cross-plot diagrams; performing data preprocessing on the main control factor data to obtain parameter data of the same scale; inputting the parameter data of the same scale into a pre-trained single-scale coal seam gas content prediction model to obtain target seismic scale coal seam gas content prediction results, well logging scale coal seam gas content prediction results and geological scale coal seam gas content prediction results; and performing nonlinear fusion on all the prediction results to obtain a coal seam gas content prediction value; and outputting a coal seam gas content planar distribution map of the target area according to the coal seam gas content prediction value. The method is used to improve the prediction accuracy of the coal seam gas content.

Description

Technical Field

[0001] The present application relates to the field of geophysical exploration technology, and in particular to a method, device, equipment, storage medium and product for predicting coal seam gas content. Background Art

[0002] With the widespread application of multi-source, multi-scale data from seismic, well logging, and geological sources, the comprehensive utilization of these data can effectively enhance the understanding and assessment of coalbed methane resources and improve the accuracy of coalbed methane content prediction. In practical applications, different data sources exhibit significant differences in scale, information characteristics, and noise levels. Therefore, effectively integrating these data, eliminating redundant features, and retaining key characteristics has become a key issue in coalbed methane content prediction.

[0003] In the existing technology, coal seam gas content is predicted based on a simple data linear weighted fusion strategy. However, this method fails to fully exploit the complementarity between multi-source and multi-scale data, resulting in low coal seam gas content prediction accuracy. Summary of the Invention

[0004] The embodiments of the present application provide a coal seam gas content prediction method, apparatus, equipment, storage medium and product to achieve the effect of improving the accuracy of coal seam gas content prediction.

[0005] In a first aspect, an embodiment of the present application provides a coal seam gas content prediction method, which is applied to a computer device, comprising: obtaining multi-scale data from different sources in a target area; obtaining a correspondence between the coal seam gas content and the multi-scale data from different sources, and producing a plurality of intersection diagrams based on the correspondence; determining the main controlling factor data of the coal seam gas content based on the plurality of intersection diagrams; performing data preprocessing on the main controlling factor data to obtain parameter data of the same scale; inputting the parameter data of the same scale into a pre-trained seismic-scale coal seam gas content prediction model to obtain a seismic-scale coal seam gas content prediction result for the target area; and inputting the parameter data of the same scale into a pre-trained seismic-scale coal seam gas content prediction model to obtain a seismic-scale coal seam gas content prediction result for the target area. The parameter data of the target area are input into the pre-trained well logging scale coal seam gas content prediction model to obtain the well logging scale coal seam gas content prediction result of the target area; the parameter data of the same scale are input into the pre-trained geological scale coal seam gas content prediction model to obtain the geological scale coal seam gas content prediction result of the target area; the seismic scale coal seam gas content prediction result, the well logging scale coal seam gas content prediction result and the geological scale coal seam gas content prediction result are nonlinearly fused by a nonlinear integration method to obtain a coal seam gas content prediction value; according to the coal seam gas content prediction value, a planar distribution map of the coal seam gas content of the target area is output.

[0006] In one possible embodiment, the parameter data includes seismic scale data; accordingly, the training process of the seismic scale coal seam gas content prediction model includes: using different machine learning algorithms to respectively construct initial seismic scale coal seam gas content prediction models for the seismic scale data; for any machine learning algorithm, according to each training result, using the Bayesian hyperparameter optimization method to optimize the parameters of the seismic scale coal seam gas content prediction model, and calculating the seismic scale coal seam gas content prediction error; among the different machine learning algorithms, selecting the machine learning algorithm with the smallest seismic scale coal seam gas content prediction error to establish a trained seismic scale coal seam gas content prediction model.

[0007] In one possible embodiment, the machine learning algorithm is a random forest algorithm; accordingly, the training process of the seismic-scale coal seam gas content prediction model includes: randomly extracting samples with replacement to generate multiple different training subsets, wherein each training subset includes parameter data in the sample well; for each data subset, constructing a decision tree, and randomly selecting multiple feature data for subset data modeling; selecting the optimal feature value from the multiple feature data as a split point according to a preset criterion, and dividing the coal seam gas content data; continuing to split the two data areas obtained by the division according to the preset criterion until the preset stop condition is met, thereby completing the establishment of the seismic-scale coal seam gas content prediction model corresponding to the data subset; the seismic-scale coal seam gas content prediction models corresponding to all data subsets constitute the seismic-scale coal seam gas content prediction model.

[0008] In a possible implementation, the formula for dividing the coal seam gas content data is:

[0009]

[0010] Where y i The distribution is the gas content value of coal seams in different areas; Represents the region R1(j,s)={x|x j The average value of coal seam gas content in ≤s}; Represents the region R2(j,s)={x|x j The average gas content of coal seams ≥s}; x j represents the jth feature in the subset; s represents the split point; N1 represents the number of samples in the R1 region, and N2 represents the number of samples in the R2 region; MSE(j,i) represents traversing the randomly selected multiple feature data and taking the feature value with the smallest MSE as the optimal feature value.

[0011] In a possible implementation, the calculation formula for the prediction result of the seismic scale coal seam gas content prediction model is:

[0012]

[0013]

[0014] Where, T b represents the coal seam gas content prediction model of each subset; x test represents each training subset; represents the coal seam gas content prediction result given by the coal seam gas content prediction model of each subset; B represents the number of subsets; It represents the prediction result of the seismic scale coal seam gas content prediction model.

[0015] In one possible embodiment, the parameter data also includes well logging scale data; accordingly, the training process of the well logging scale coal seam gas content prediction model includes: using different machine learning algorithms to respectively construct initial well logging scale coal seam gas content prediction models for the well logging scale data; for any machine learning algorithm, according to each training result, using the Bayesian hyperparameter optimization method to optimize the parameters of the well logging scale coal seam gas content prediction model, and calculating the well logging scale coal seam gas content prediction error; among the different machine learning algorithms, selecting the machine learning algorithm with the smallest well logging scale coal seam gas content prediction error to establish a trained well logging scale coal seam gas content prediction model.

[0016] In one possible embodiment, the parameter data also includes geological scale data; accordingly, the training process of the geological scale coal seam gas content prediction model includes: using different machine learning algorithms to respectively construct initial geological scale coal seam gas content prediction models for the geological scale data; for any machine learning algorithm, according to each training result, using the Bayesian hyperparameter optimization method to optimize the parameters of the geological scale coal seam gas content prediction model, and calculating the geological scale coal seam gas content prediction error; among the different machine learning algorithms, selecting the machine learning algorithm with the smallest geological scale coal seam gas content prediction error to establish a trained geological scale coal seam gas content prediction model.

[0017] In a possible implementation, the calculation formula for the predicted value of coal seam gas content is:

[0018]

[0019] Where, Represents the predicted value of coal seam gas content; RF (1,2,3) represent the seismic scale coal seam gas content prediction model, the well logging scale coal seam gas content prediction model and the geological scale coal seam gas content prediction model respectively; W l (1,2,3) represents the weight of the coal seam gas content prediction model at different scales of the lth layer, bl F represents the bias of the lth layer, l = 1, 2, …, L. l F represents the activation function of the lth layer, l = 1, 2, …, L.

[0020] In a possible implementation, the master factor data includes seismic data, logging data and geological data; accordingly, the data preprocessing of the master factor data comprises: performing mathematical transformation on the seismic data to obtain seismic attributes related to the coal seam gas content; for the logging data, obtaining a three-dimensional data body of the same scale as the seismic data through a high-resolution inversion method; for the geological data, obtaining geological attributes of the same scale as the seismic data through indirect reflection of the seismic attributes and / or the three-dimensional data body.

[0021] In a possible implementation, the geological data includes coal body structure and fracture data, fault data and discrete factor data; accordingly, the indirect reflection of the geological data through the seismic attributes and / or the three-dimensional data body comprises: indirectly reflecting the coal body structure and fracture data through a density value, wherein the density value is determined through the logging data; indirectly reflecting the fault data through a dip attribute, wherein the dip attribute is determined through the seismic attributes; determining an estimated value of each discrete factor data through a Kriging interpolation method, wherein a determination formula of the estimated value of each discrete factor data is:

[0022]

[0023] In the formula, G(x0, y0) represents an estimated value of a point (x0, y0) in a space-time domain; G(x i ,y j ) is each discrete factor data; and k(i, j) is a weight coefficient of different spatial points.

[0024] In a second aspect, an embodiment of the present application provides a coal seam gas content prediction device, applied to a computer device, comprising:

[0025] A data acquisition module is configured to acquire different-source multi-scale data of a target region.

[0026] A crossplot drawing module is configured to acquire a corresponding relationship between the coal seam gas content and the different-source multi-scale data, and to draw a plurality of crossplots according to the corresponding relationship.

[0027] A master factor data determination module is configured to determine master factor data of the coal seam gas content according to the plurality of crossplots.

[0028] A data preprocessing module is configured to perform data preprocessing on the master factor data to obtain parameter data of the same scale.

[0029] The seismic scale coal seam gas content prediction module is used to input the parameter data of the same scale into the pre-trained seismic scale coal seam gas content prediction model to obtain the seismic scale coal seam gas content prediction result of the target area.

[0030] The well logging scale coal seam gas content prediction module is used to input the parameter data of the same scale into the pre-trained well logging scale coal seam gas content prediction model to obtain the well logging scale coal seam gas content prediction result of the target area.

[0031] The geological scale coal seam gas content prediction module is used to input the parameter data of the same scale into the pre-trained geological scale coal seam gas content prediction model to obtain the geological scale coal seam gas content prediction result of the target area.

[0032] The nonlinear fusion module is used to perform nonlinear fusion on the seismic scale coal seam gas content prediction result, the well logging scale coal seam gas content prediction result and the geological scale coal seam gas content prediction result by using a nonlinear integration method to obtain a coal seam gas content prediction value.

[0033] The output module is used to output a planar distribution map of the coal seam gas content in the target area according to the predicted value of the coal seam gas content.

[0034] In a third aspect, an embodiment of the present application provides a computer device, including: a memory, a processor;

[0035] The memory stores computer-executable instructions;

[0036] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementations of the first aspect.

[0037] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the first aspect above and / or various possible implementation methods of the first aspect.

[0038] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above first aspect and / or various possible implementation methods of the first aspect.

[0039] The coal seam gas content prediction method, device, equipment, storage medium and product provided by the embodiments of the present application can output a seismic scale coal seam gas content prediction result, a logging scale coal seam gas content prediction result and a geological scale coal seam gas content prediction result by using an artificial intelligence algorithm for multi-source and multi-scale data such as seismic data, logging data and geological data, and can realize fusion of data of different sources and different scales in a high-dimensional space by using a nonlinear fusion algorithm, thereby overcoming the limitations of conventional methods in terms of single data source and insufficient nonlinear processing capability, and improving the coal seam gas content prediction accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0040] The accompanying drawings, which are incorporated herein and form part of the specification, illustrate embodiments consistent with the present application and, together with the description, further serve to explain the principles of the application.

[0041] Figure 1 A scene schematic diagram of the coal seam gas content prediction method provided by the embodiments of the present application is shown in the following figure.

[0042] Figure 2 A flowchart of the coal seam gas content prediction method provided by the embodiments of the present application is shown in the following figure.

[0043] Figure 2a A relationship diagram between a seismic scale parameter and coal seam gas content provided by the embodiments of the present application is shown in the following figure.

[0044] Figure 2b A relationship diagram between a logging scale parameter and coal seam gas content provided by the embodiments of the present application is shown in the following figure.

[0045] Figure 2c A relationship diagram between a burial depth and coal seam gas content provided by the embodiments of the present application is shown in the following figure.

[0046] Figure 2d A relationship diagram between a coal seam thickness and coal seam gas content provided by the embodiments of the present application is shown in the following figure.

[0047] Figure 2e A relationship diagram between a density and coal seam gas content provided by the embodiments of the present application is shown in the following figure.

[0048] Figure 2f A relationship diagram between a coal seam roof sandstone thickness and coal seam gas content provided by the embodiments of the present application is shown in the following figure.

[0049] Figure 2g A relationship diagram between a dip angle and coal seam gas content provided by the embodiments of the present application is shown in the following figure.

[0050] Figure 3a A relative error diagram of a prediction value on a single well provided by the embodiments of the present application is shown in the following figure.

[0051] Figure 3bA comparison chart of the predicted results and actual results of coal seam gas content in a single well provided in the embodiment of the present application;

[0052] Figure 3c A schematic diagram of the prediction results on a single well when different training wells and test wells are randomly selected according to an embodiment of the present application;

[0053] Figure 3d A plane distribution diagram of the predicted coal seam gas content provided in an embodiment of the present application;

[0054] Figure 4 A schematic diagram of the structure of a coal seam gas content prediction device provided in an embodiment of the present application;

[0055] Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of the present application.

[0056] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0057] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0058] In order to clearly understand the technical solutions of this application, we first introduce the solutions of the prior art in detail. In recent years, with the continuous growth of global energy demand, oil and gas resources have occupied an important position in the global energy structure. Conventional oil and gas exploration and development technologies have long been relatively mature, and underground oil and gas resources have been effectively developed through means such as seismic exploration, drilling technology, and logging technology. However, the depletion of conventional oil and gas resources and the increasing global demand for clean energy have made the exploration and development of unconventional oil and gas resources and deep oil and gas resources gradually become hot topics. Unconventional oil and gas resources include shale gas, tight sandstone gas, and coalbed methane. They are widely distributed and have abundant reserves, and are becoming an important direction for future energy development. Among unconventional oil and gas resources, coalbed methane, a natural gas mainly distributed in coal seams, has attracted widespread attention due to its high methane content and abundant resources. With the promotion of renewable energy and the gradual reduction of conventional oil and gas resources, the importance of coalbed methane as a clean energy source is becoming increasingly prominent. Therefore, accurately assessing the gas content of coal seams is of great practical significance for drilling deployment, the effective development and utilization of this resource, promoting energy transition, and reducing environmental pollution. Coalbed methane (CBM) is primarily formed and stored in coal seams during the coalification process through biogenic or thermogenic processes. Its occurrence mechanism is influenced by multiple factors, including the geological structure, burial depth, and coal quality. Due to the complex and diverse processes of CBM formation and storage, its occurrence is often highly nonlinear and uncertain. Specifically, the porosity, thickness, burial depth, and interaction between the coal seam and the surrounding rocks all influence the flow and accumulation of CBM. Traditional prediction methods often rely on single or limited data sources and fail to fully utilize multi-source and multi-scale data. This results in significant uncertainty in coalbed methane predictions. Under varying geological conditions, traditional methods can produce large deviations in prediction results, making accuracy difficult to guarantee, hindering the development and economic benefits of CBM. With the widespread application of multi-source, multi-scale data such as seismic, well logging, and geological data, the comprehensive utilization of these data can effectively enhance the understanding and assessment of coalbed methane resources and improve the accuracy of coalbed methane content prediction. For example, seismic data can provide the spatial distribution and geometric characteristics of coal seams, well logging data can reflect coal reservoir characteristics, and geological data can reveal the overall distribution patterns of coalbed methane. However, in practical applications, different data sources exhibit significant differences in scale, information characteristics, and noise levels. Therefore, how to effectively fuse these data, eliminate redundant features, and retain key characteristics becomes a key issue in coalbed methane content prediction. Most research methods rely on simple linear weighted data fusion strategies and fail to fully exploit the complementarity between multi-source and multi-scale data. Although some studies have used statistical models (empirical formulas) to predict coalbed methane content and have achieved some success in specific regions, these methods generally lack the ability to deeply explore the complex relationships between multi-source and multi-scale data and are difficult to adapt to the complex occurrence of coalbed methane under diverse geological conditions.While machine learning methods such as artificial neural networks can process large amounts of data and identify underlying patterns, most studies predict gas content based on different attributes of seismic data or its mathematical transformations. This fusion of multiple seismic attributes rather than multi-source, multi-scale data is limited by the resolution of seismic data, resulting in low coalbed gas content prediction accuracy. Secondly, some studies have combined machine learning with various well logging data sensitive to coalbed gas content to predict coalbed gas content. While these predictions have high vertical resolution, they are limited to gas content predictions at well locations. The accuracy of mathematical interpolation methods applied to non-well locations is limited, especially in areas with strong lateral heterogeneity in coalbed gas content. Furthermore, the prediction of gas content in thin coalbeds is affected by lithologic combinations and is subject to significant uncertainty.

[0059] In order to solve the above technical problems, the inventors thought of improving the prediction accuracy of deep coal seam gas content by integrating multi-source and multi-scale data such as seismic data, well logging data and geological data and machine learning algorithms. Specifically, for multi-source and multi-scale data such as seismic data, well logging data and geological data, artificial intelligence is used to output seismic-scale coal seam gas content prediction results, well logging-scale coal seam gas content prediction results and geological-scale coal seam gas content prediction results. The nonlinear fusion algorithm is used to realize the fusion of data from different sources and scales in high-dimensional space, overcoming the limitations of conventional methods in terms of the singleness of data source and insufficient nonlinear processing capabilities, thereby improving the prediction accuracy of coal seam gas content.

[0060] Based on the above creative findings, the inventor proposed the technical solution of this application.

[0061] The following specific embodiments describe in detail the technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems. The following specific embodiments 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 in conjunction with the accompanying drawings.

[0062] Figure 1 A schematic diagram of a scenario of a method for predicting coal seam gas content provided in an embodiment of the present application, such as Figure 1 As shown, the specific application scenario of this application includes: a receiving device 101, a processor 102 and a display device 103.

[0063] It is understood that the structure illustrated in the embodiment of this application does not constitute a specific limitation on the coal seam gas content prediction method. In other feasible embodiments of this application, the above architecture may include more or fewer components than shown, or combine or split certain components, or arrange the components differently. The specific configuration can be determined based on the actual application scenario and is not limited here. Figure 1 The components shown may be implemented in hardware, software, or a combination of software and hardware.

[0064] In a specific implementation process, the receiving device 101 can be an input / output interface or a communication interface, which is used to obtain different source multi-scale data of the target area and the corresponding relationship between the coal seam gas content and the different source multi-scale data.

[0065] Processor 102 can produce multiple intersection maps based on the corresponding relationship, and determine the main controlling factor data of the coal seam gas content based on the multiple intersection maps; perform data preprocessing on the main controlling factor data to obtain parameter data of the same scale; input the parameter data of the same scale into the pre-trained seismic scale coal seam gas content prediction model to obtain the seismic scale coal seam gas content prediction result of the target area; input the parameter data of the same scale into the pre-trained well logging scale coal seam gas content prediction model to obtain the well logging scale coal seam gas content prediction result of the target area; input the parameter data of the same scale into the pre-trained geological scale coal seam gas content prediction model to obtain the geological scale coal seam gas content prediction result of the target area; use a nonlinear integration method to nonlinearly fuse the seismic scale coal seam gas content prediction result, the well logging scale coal seam gas content prediction result and the geological scale coal seam gas content prediction result to obtain the coal seam gas content prediction value; output the coal seam gas content plane distribution map of the target area based on the coal seam gas content prediction value.

[0066] The display device 103 can be used to display a planar distribution map of coal seam gas content in a target area.

[0067] The display device may also be a touch screen display, which is used to receive user instructions while displaying the above content to achieve operational interaction with the user.

[0068] It should be understood that the above-mentioned processor can be implemented by the processor reading instructions in the memory and executing the instructions, or it can be implemented by a chip circuit.

[0069] Figure 2 A flow chart of the coal seam gas content prediction method provided in the embodiment of the present application is shown as follows: Figure 2 As shown, the method includes:

[0070] S201: Acquire multi-scale data from different sources in the target area.

[0071] Data is prepared from three perspectives: seismic, well logging, and geological. Seismic data related to coal seam gas content includes raw seismic data and its mathematical transformation results, such as instantaneous frequency, seismic amplitude, wave impedance, attenuation coefficient, inclination, and coherence. Well logging data refers to rock parameters related to coal seam gas content, including density, natural gamma, and porosity. Geological data refers to key factors controlling the occurrence of coal seam gas content, including thermal evolution, sedimentary environment, and structural characteristics.

[0072] Specifically, seismic data acquired through active-source seismic exploration can reveal fluctuations in underground strata structure and properties. Drilling measures various parameters within coal seams, such as resistivity, density, and acoustic transit time. Based on the theory of geodetic reservoir formation and storage control, the key factors influencing coalbed methane occurrence, such as the coal seam's depositional environment, structural characteristics, and degree of thermal evolution, are analyzed.

[0073] S202: Obtain the corresponding relationship between the coal seam gas content and the multi-scale data from different sources, and generate multiple cross-plots based on the corresponding relationship.

[0074] Specifically, the intersection analysis method is used to select sensitive parameters or main controlling factors that have a certain change pattern with the change of coal seam gas content. For example, in seismic terms, in gas-bearing coal seams, the density is relatively low and the sound wave propagation speed is slow. When seismic waves pass through coal seams with high gas content, strong reflection information will be generated, which is manifested as a strong amplitude feature on the seismic profile. In well logging, when the ash content is high, the pores and cracks in the coal are partially blocked, reducing the adsorption space and migration channels of coalbed methane, which may lead to a decrease in gas content. Ash is mainly composed of inorganic minerals in coal seams, and a higher ash content corresponds to a higher natural gamma value. Therefore, natural gamma can be used to indirectly reflect the gas content of coal seams. Geologically, the factors affecting coal seam gas content can be analyzed from three aspects: "generation, storage, and capping". For example, a high degree of thermal evolution is conducive to the formation of more coal seam gas; when the roof and floor plates are mudstone deposits, the low permeability of the mudstone acts as a seal. The thicker the roof mudstone, the more conducive it is to sealing the underlying coal seam gas, preventing gas escape and retaining more coal seam gas; when the geological structure is complex and faults are developed, the faults can be coal seam gas migration channels, and can also serve as side blockages to prevent coal seam gas migration; in addition, the stress near the fault development is more complex, which easily forms cracks in the coal seam, increasing the coal seam gas storage space.

[0075] Specifically, the data of different sources and scales obtained from the basic analysis are taken as the standard, and the coal seam gas content calculated by well logging or the gas content of coal seam core sampling is used to make intersection plots of different data with the coal seam gas content.

[0076] S203: Determine the main controlling factor data of the coal seam gas content based on the multiple intersection diagrams.

[0077] Specifically, the changing trends of different data with coal seam gas content are analyzed, and data with obvious changing rules and high correlation are selected as the main controlling factor data of coal seam gas content.

[0078] Specifically, Figure 2a The diagram of the relationship between the seismic scale parameters and the coal seam gas content provided in the embodiment of this application is as follows. Figure 2aAs shown in the figure below, the left figure shows the relationship between seismic amplitude and coal seam gas content, and the right figure shows the relationship between intercept minus gradient and coal seam gas content. In the seismic data, the absolute value of seismic amplitude and the absolute value of intercept minus gradient both increase with the increase of coal seam gas content. Figure 2b The diagram of the relationship between the logging scale parameters and the coal seam gas content provided in the embodiment of this application is as follows. Figure 2b As shown in the well logging data, the smaller the natural gamma ray, the lower the ash content and the fewer impurities in the coal, which is conducive to gas adsorption on the coal seam surface. Geologically, the factors affecting coal seam gas content are analyzed from the perspectives of "source, reservoir, and caprock." Figure 2c The relationship between the buried depth and the gas content of the coal seam is shown in the following example: Figure 2c As shown in the figure, within the critical depth, the deeper the coal seam is, the higher the gas content of the coal seam is. Figure 2d The schematic diagram of the relationship between coal seam thickness and coal seam gas content provided in the embodiment of this application is as follows: Figure 2d As shown, thick coal seams generally have larger reservoir volumes and more organic matter that can generate and store coalbed methane, increasing the total gas content of the reservoir. Therefore, the thicker the coal seam, the higher the gas content. Figure 2e The schematic diagram of the relationship between density and coal seam gas content provided in the embodiment of this application is as follows: Figure 2e As shown, the denser the coal structure, the higher the gas content, the better the crack development, and the easier the gas migration and dispersion. The coal structure and crack development can be indirectly reflected by density. The smaller the density, the higher the gas content of the coal seam. Figure 2f The following is a schematic diagram of the relationship between the thickness of the coal seam roof sandstone and the coal seam gas content provided in the embodiment of this application. Figure 2f As shown in the figure, from the perspective of caprock gas preservation, the lithology, thickness, continuity and geological structure of the caprock determine its sealing effect on coalbed methane. A good caprock can effectively seal coalbed methane and prevent gas escape. The thicker the roof sandstone, the weaker the sealing ability and the smaller the gas content of the coal seam. Figure 2g The schematic diagram of the relationship between the inclination angle and the coal seam gas content provided in the embodiment of this application is as follows: Figure 2g As shown, the dip angle can indirectly reflect the geological structure of a coal seam, such as faults. Faults can serve as both pathways for coalbed methane migration and lateral barriers to prevent it. Furthermore, stresses near faults are more complex, making it more likely for cracks to form in the coal seam, increasing the gas storage capacity. Within a certain range, the greater the dip angle of the coal seam roof, the higher the coal seam's gas content.

[0079] S204: Preprocess the main control factor data to obtain parameter data of the same scale.

[0080] Among them, the main controlling factor data include seismic data, well logging data and geological data.

[0081] Specifically, the data preprocessing process includes:

[0082] Sa1: performing mathematical transformation on the seismic data to obtain seismic attributes related to the gas content of the coal seam.

[0083] The mathematical transformation can include spectral analysis related transformation, wavelet transformation and attribute transformation.

[0084] Specifically, when the seismic wave propagates underground, its frequency component is affected by the gas content of the coal seam. Different gas contents of the coal seam result in different absorption and reflection of the seismic wave, leading to changes in the frequency component. Through spectral analysis after Fourier transformation, the dominant frequency and frequency width can be extracted. Wavelet transformation can capture the local frequency and time characteristic changes of the seismic wave when passing through the coal seam. When the gas content of the coal seam changes, causing the frequency of the seismic wave to change within a certain time period, wavelet transformation can accurately locate this time period and analyze the frequency change, thereby obtaining attributes related to the gas content of the coal seam. Mathematical transformation of the amplitude can obtain more sensitive attributes, and analysis of the change of amplitude-based attributes such as reflection coefficient can infer the gas content of the coal seam.

[0085] Sa2: obtaining a three-dimensional data volume of the same scale as the seismic data through a high-resolution inversion method for the logging data.

[0086] The propagation characteristics of the seismic wave, such as velocity, amplitude and phase, are closely related to the lithology and physical parameters of the stratum when it propagates underground. High-resolution inversion is to use this relationship to inversely deduce the detailed parameter distribution of the underground stratum through known seismic data.

[0087] Specifically, an initial model of the underground stratum structure and physical parameter distribution is built with the existing geological, logging and seismic data. The initial model is forward calculated using the mathematical model of seismic wave propagation to simulate the propagation process of the seismic wave in the model, and then the theoretical seismic response data is obtained. The seismic response data obtained by forward simulation is compared with the actually collected seismic data to find the differences between them. According to these differences, the parameters of the initial model are adjusted according to a specific inversion algorithm to make the simulation results closer to the actual seismic data. Since the logging data can accurately reflect the physical parameters of the stratum at the wellbore, it can be used as a calibration point to ensure that the model obtained by inversion is consistent with the actual measurement results at the wellbore, thereby guiding the inversion process to a more accurate direction. Through the high-resolution inversion method, a three-dimensional data volume is finally generated. This data volume is consistent with the seismic data in terms of spatial range, covering the three-dimensional space of the entire study area. At the same time, it also incorporates the high-resolution information of the logging data.

[0088] Sa3: obtaining geological attributes of the same scale as the seismic data indirectly through the seismic attributes and / or the three-dimensional data volume for the geological data.

[0089] Among them, geological data include coal structure and crack data, fault data and discrete factor data.

[0090] Specifically, the geological data is indirectly reflected through the seismic attributes and / or the three-dimensional data volume, including:

[0091] Sb1: indirectly reflecting the coal body structure and fracture data through density values, wherein the density values ​​are determined through the well logging data.

[0092] Sb2: indirectly reflecting the fault data through dip attributes, wherein the dip attributes are determined by the seismic attributes.

[0093] Sb3: Determine the estimated value of each discrete factor data using the Kriging interpolation method. The formula for determining the estimated value of each discrete factor data is:

[0094]

[0095] In the formula, G(x0,y0) represents the estimated value of the point (x0,y0) in the spatiotemporal domain; G(x i ,y j ) are the discrete factor data; k(i,j) are the weight coefficients of different spatial points.

[0096] S205: Inputting parameter data of the same scale into a pre-trained seismic-scale coal seam gas content prediction model to obtain a seismic-scale coal seam gas content prediction result for the target area.

[0097] Among them, the parameter data includes earthquake scale data.

[0098] Specifically, different machine learning algorithms, such as random forest, are optimized for seismic scale data to adapt to the characteristics of data at different scales and establish a single-scale gas content prediction model, including:

[0099] Sc1: Using different machine learning algorithms on the seismic scale data, construct initial seismic scale coal seam gas content prediction models.

[0100] Sc2: For any machine learning algorithm, according to each training result, the parameters of the seismic-scale coalbed gas content prediction model are optimized using the Bayesian hyperparameter optimization method, and the seismic-scale coalbed gas content prediction error is calculated.

[0101] Hyperparameters are parameters that need to be set in advance in machine learning algorithms. They are not learned from training data, but instead influence the model's training process and performance. For example, for decision trees, the maximum depth and minimum number of sample splits are hyperparameters; for neural networks, the number of hidden layers, the number of neurons in each layer, and the learning rate are hyperparameters.

[0102] Specifically, Bayesian optimization establishes a substitute function based on the evaluation result of the objective function and finds the value that minimizes the objective function. The specific process is:

[0103] 1) Initialize a prior distribution of the proxy model, and often use Gaussian process as the proxy function to approximate the target function: f(x)~GP(μ(x),k(x,x ’ )), where μ(x) is the mathematical expectation, k(x,x') is the covariance function, and f(x) is the objective function.

[0104] 2) Select the next hyperparameter combination through the acquisition function, using the expected improvement as the acquisition function:

[0105]

[0106] Where x and x t+1 are the hyperparameter combinations of the current and next samples, X is the hyperparameter search space, and EI(·) is the expected function.

[0107] 3) Use hyperparameter combination x t+1 Update the proxy model.

[0108] 4) Repeat steps 2) and 3) until the iteration stop condition is reached.

[0109] 5) Returns the optimal hyperparameter combination among the hyperparameter combinations evaluated by the objective function.

[0110] Sc3: Among the different machine learning algorithms, select the machine learning algorithm with the smallest seismic-scale coal seam gas content prediction error to establish a trained seismic-scale coal seam gas content prediction model.

[0111] Specifically, for seismic scale data, the machine learning method with the smallest coal seam gas content prediction error is selected to establish the optimal coal seam gas content prediction model for seismic scale data.

[0112] S206: Inputting parameter data of the same scale into a pre-trained well logging scale coal seam gas content prediction model to obtain a well logging scale coal seam gas content prediction result of the target area.

[0113] Among them, parameter data includes well logging scale data.

[0114] Specifically, different machine learning algorithms, such as random forest, are optimized for well logging scale data to adapt to the characteristics of data at different scales and establish a single-scale gas content prediction model, including:

[0115] Sd1: Using different machine learning algorithms on the logging scale data, initial logging scale coal seam gas content prediction models are constructed respectively.

[0116] Sd2: For any machine learning algorithm, according to each training result, the parameters of the well logging scale coal seam gas content prediction model are optimized using the Bayesian hyperparameter optimization method, and the well logging scale coal seam gas content prediction error is calculated.

[0117] Sd3: Among the different machine learning algorithms, the machine learning algorithm with the smallest well logging scale coal seam gas content prediction error is selected to establish a trained well logging scale coal seam gas content prediction model.

[0118] S207: Inputting parameter data of the same scale into a pre-trained geological scale coal seam gas content prediction model to obtain a geological scale coal seam gas content prediction result of the target area.

[0119] Among them, parameter data includes geological scale data.

[0120] Specifically, different machine learning algorithms, such as random forest, are optimized for geological scale data to adapt to the characteristics of data at different scales and establish single-scale gas content prediction models, including:

[0121] Se1: Use different machine learning algorithms to construct initial geological scale coal seam gas content prediction models for the geological scale data.

[0122] Se2: For any machine learning algorithm, according to each training result, the parameters of the geological-scale coal seam gas content prediction model are optimized using the Bayesian hyperparameter optimization method, and the geological-scale coal seam gas content prediction error is calculated.

[0123] Se3: Among the different machine learning algorithms, select the machine learning algorithm with the smallest geological scale coal seam gas content prediction error to establish a trained geological scale coal seam gas content prediction model.

[0124] S208: Using a nonlinear integration method, nonlinearly fuse the seismic scale coal seam gas content prediction results, the well logging scale coal seam gas content prediction results, and the geological scale coal seam gas content prediction results to obtain a coal seam gas content prediction value.

[0125] Specifically, the well logging scale coal seam gas content prediction model RF1, the seismic scale coal seam gas content prediction model RF2 and the geological scale coal seam gas content prediction model RF3 were constructed, and the coal seam gas content prediction nonlinearity at multiple scales was integrated into a multi-scale composite model. The final output of the model is:

[0126]

[0127] Where, Represents the predicted value of coal seam gas content; RF (1,2,3) They represent the seismic scale coal seam gas content prediction model, the well logging scale coal seam gas content prediction model and the geological scale coal seam gas content prediction model respectively; W l (1,2,3) represents the weight of the coal seam gas content prediction model at different scales of the lth layer, b l represents the bias of the lth layer, F l represents the activation function of the lth layer, l = 1, 2, …, L.

[0128] S209: Outputting a planar distribution map of coal seam gas content in the target area based on the predicted value of coal seam gas content.

[0129] Specifically, based on the predicted value of coal seam gas content, first organize it and the corresponding spatial location information of the target area, select tools such as drawing library, and set the coordinate system and color mapping according to the actual target area. If the predicted value is a discrete point, you can first draw the point and assign color according to the color mapping, or you can use the interpolation algorithm to expand it to continuous surface data to draw contour maps or fill contour maps, and then add coordinate axis labels, legends, map names and other annotations. Finally, output the plane distribution map of coal seam gas content in the target area in the required format.

[0130] In summary, for multi-source and multi-scale data such as seismic data, well logging data and geological data, artificial intelligence algorithms are used to output seismic-scale coal seam gas content prediction results, well logging-scale coal seam gas content prediction results and geological-scale coal seam gas content prediction results. The nonlinear fusion algorithm is used to realize the fusion of data from different sources and scales in high-dimensional space, which overcomes the limitations of conventional methods in terms of the singleness of data source and insufficient nonlinear processing capabilities, and improves the accuracy of coal seam gas content prediction.

[0131] In another embodiment provided in the present application, in training a seismic-scale coal seam gas content prediction model, a random forest algorithm is used as an example to establish a model. The training process includes:

[0132] S301: Generate multiple different training subsets by randomly sampling samples with replacement, wherein each training subset includes parameter data in a sample well.

[0133] Specifically, assuming that the seismic scale data of the training input is X, that is, multiple parameter data within the same scale, and the corresponding label Y is the coal seam gas content, multiple different training subsets are generated by randomly sampling samples with replacement. The single subset is represented as: D = {(x1, y1), (x2, y2), ..., (x N ,y N )},x i =(x i 1 ,x i 2 ,…x i n ). Where x i For the same scale data as the input, y i is the corresponding coal seam gas content, n is the number of data features at this scale, and N is the number of samples in itself.

[0134] S302: For each data subset, a decision tree is constructed, and multiple feature data are randomly selected to perform subset data modeling.

[0135] Specifically, a decision tree is constructed for each data subset to train the subset data. K feature data are randomly selected from the data for subset data modeling, where k≤n.

[0136] S303: Selecting an optimal characteristic value from the plurality of characteristic data as a dividing point according to a preset criterion, and dividing the coal seam gas content data.

[0137] Specifically, according to a preset criterion such as minimizing the square error, the eigenvalue of the best feature among the k features is selected as the split point, such as the jth feature x in the subset j and its value s as the segmentation variable and segmentation point, and define two regions R1(j,s)={x|x j ≤s} and R2(j,s)={x|x j ≥s}, the data partitioning error corresponding to different j and s is:

[0138]

[0139] Where y i The distribution is the gas content value of coal seams in different areas; Represents the region R1(j,s)={x|x j The average value of coal seam gas content in ≤s}; Represents the region R2(j,s)={x|x j ≥s}; MSE(j,i) represents traversing the plurality of randomly selected feature data and taking the feature value with the smallest MSE as the optimal feature value.

[0140]

[0141] Where x j represents the jth feature in the subset; s represents the split point; N1 represents the number of samples in the R1 region, and N2 represents the number of samples in the R2 region.

[0142] S304: Continue to split the two divided data regions according to the preset criteria until a preset stop condition is met, thereby completing the establishment of a seismic-scale coal seam gas content prediction model corresponding to the data subset.

[0143] Specifically, the two regions are split according to the same criteria until the preset stopping conditions are met, completing the establishment of a gas content prediction model for a single data subset. Common stopping conditions include: reaching the maximum tree depth; the number of samples being less than the minimum number of samples; and the mean square error (MSE) or other error metric of the current node being less than a set threshold.

[0144] S305: The seismic scale coal seam gas content prediction models corresponding to all data subsets constitute the seismic scale coal seam gas content prediction model.

[0145] Specifically, the gas content prediction model T for each subset b The calculation formula for the gas content prediction result is given as:

[0146]

[0147] The calculation formula of the content prediction value of the seismic scale coal seam gas content prediction model constructed by B subset coal seam gas content prediction models is:

[0148]

[0149] Where, T b represents the coal seam gas content prediction model of each subset; x test represents each training subset; represents the coal seam gas content prediction result given by the coal seam gas content prediction model of each subset; B represents the number of subsets; It represents the prediction result of the seismic scale coal seam gas content prediction model.

[0150] In summary, the random forest algorithm enhances its anti-overfitting ability through random sampling and feature selection and can process high-dimensional data with a large number of features, and performs well when there is a lot of feature redundancy.

[0151] In the examples provided herein, 4 wells were randomly selected from 18 wells in the target area as test wells and 14 wells as training wells. A random forest approach was used to construct coalbed gas content prediction models at different scales, and Bayesian hyperparameter optimization was used to find the optimal parameter combinations for the models at different scales and to optimize the nonlinear fusion process of the models at different scales.

[0152] Figure 3a This is a schematic diagram of the relative error of the predicted value on a single well provided in the embodiment of the present application, as shown in FIG. Figure 3a As shown, the dotted box contains data from four test wells. Overall, the prediction error of coal seam gas content in the 14 training wells is below 6%, and the prediction error of coal seam gas content in the four test wells is below 10%.

[0153] Figure 3b The comparison chart of the coal seam gas content prediction results and actual results in a single well provided in the embodiment of this application is as follows: Figure 3b As shown in the figure, the coal seam gas content prediction results are very close to the actual results, and the absolute error of coal seam gas content prediction is within 1m 3 / t, and the overall predicted gas content change trend is basically consistent with the actual gas content change trend, with no major deviation.

[0154] Figure 3c This is a schematic diagram of the prediction results on a single well when different training wells and test wells are randomly selected according to the embodiment of the present application. Figure 3c As shown in the figure, under different random selection conditions, the difference between the predicted results and the actual gas content is small, and the change trends are basically consistent.

[0155] Figure 3d The predicted coal seam gas content plane distribution diagram provided in the embodiment of this application is as follows: Figure 3d As shown in the figure, the lower and northeastern parts of the target area are the main coalbed methane storage areas, and the coalbed gas content can reach 9m 3 / t, the gas content of coal seams in the western part of the target area is relatively low, and its gas content is basically around 8m 3 / t approximately.

[0156] It can be seen from the above actual data test results that the coal seam gas content prediction method provided in the embodiment of the present application has high accuracy, robustness and generalization in coal seam gas content prediction.

[0157] Figure 4 This is a schematic diagram of the structure of the coal seam gas content prediction device provided in the embodiment of the present application. Figure 4As shown, the device includes: a data acquisition module 401, an intersection diagram drawing module 402, a main control factor data determination module 403, a data preprocessing module 404, a seismic scale coal seam gas content prediction module 405, a well logging scale coal seam gas content prediction module 406, a geological scale coal seam gas content prediction module 407, a nonlinear fusion module 408 and an output module 409.

[0158] The data acquisition module 401 is used to acquire multi-scale data from different sources in the target area.

[0159] The cross-plot drawing module 402 is used to obtain the corresponding relationship between the coal seam gas content and the multi-scale data from different sources, and to generate multiple cross-plots based on the corresponding relationship.

[0160] The main controlling factor data determining module 403 is configured to determine the main controlling factor data of the coal seam gas content based on the multiple cross-graphs.

[0161] The data preprocessing module 404 is used to perform data preprocessing on the main control factor data to obtain parameter data of the same scale.

[0162] The seismic scale coal seam gas content prediction module 405 is used to input the parameter data of the same scale into the pre-trained seismic scale coal seam gas content prediction model to obtain the seismic scale coal seam gas content prediction result of the target area.

[0163] The well logging scale coal seam gas content prediction module 406 is used to input the parameter data of the same scale into the pre-trained well logging scale coal seam gas content prediction model to obtain the well logging scale coal seam gas content prediction result of the target area.

[0164] The geological scale coal seam gas content prediction module 407 is used to input the parameter data of the same scale into the pre-trained geological scale coal seam gas content prediction model to obtain the geological scale coal seam gas content prediction result of the target area.

[0165] The nonlinear fusion module 408 is used to perform nonlinear fusion on the seismic scale coal seam gas content prediction result, the well logging scale coal seam gas content prediction result and the geological scale coal seam gas content prediction result using a nonlinear integration method to obtain a coal seam gas content prediction value.

[0166] The output module 409 is configured to output a planar distribution map of the coal seam gas content in the target area according to the predicted value of the coal seam gas content.

[0167] In one possible implementation, the device also includes a seismic-scale coal seam gas content prediction model training module, which is used to use different machine learning algorithms to construct initial seismic-scale coal seam gas content prediction models for the seismic-scale data; for any machine learning algorithm, based on each training result, the parameters of the seismic-scale coal seam gas content prediction model are optimized using the Bayesian hyperparameter optimization method, and the seismic-scale coal seam gas content prediction error is calculated; among the different machine learning algorithms, the machine learning algorithm with the smallest seismic-scale coal seam gas content prediction error is selected to establish a trained seismic-scale coal seam gas content prediction model.

[0168] In one possible implementation, the machine learning algorithm is a random forest algorithm, and the seismic-scale coal seam gas content prediction model training module is specifically used to generate multiple different training subsets by randomly extracting samples with replacement, wherein each training subset includes parameter data in the sample well; for each data subset, a decision tree is constructed, and multiple feature data are randomly selected for subset data modeling; according to a preset criterion, the optimal feature value is selected from the multiple feature data as a cutting point to divide the coal seam gas content data; the two data areas obtained by the division are continued to be split according to the preset criterion until the preset stop condition is met, thereby completing the establishment of the seismic-scale coal seam gas content prediction model corresponding to the data subset; the seismic-scale coal seam gas content prediction models corresponding to all data subsets constitute the seismic-scale coal seam gas content prediction model.

[0169] In a possible implementation, the formula for dividing the coal seam gas content data in the seismic scale coal seam gas content prediction model is:

[0170]

[0171] Where y i The distribution is the gas content value of coal seams in different areas; Represents the region R1(j,s)={x|x j The average value of coal seam gas content in ≤s}; Represents the region R2(j,s)={x|x j The average gas content of coal seams ≥s}; x j represents the jth feature in the subset; s represents the split point; N1 represents the number of samples in the R1 region, and N2 represents the number of samples in the R2 region; MSE(j,i) represents traversing the randomly selected multiple feature data and taking the feature value with the smallest MSE as the optimal feature value.

[0172] In a possible implementation, the calculation formula of the prediction result of the seismic scale coal seam gas content prediction model in the seismic scale coal seam gas content prediction model is:

[0173]

[0174]

[0175] Where, T b represents the coal seam gas content prediction model of each subset; x test represents each training subset; represents the coal seam gas content prediction result given by the coal seam gas content prediction model of each subset; B represents the number of subsets; It represents the prediction result of the seismic scale coal seam gas content prediction model.

[0176] In one possible implementation, the device also includes a well logging scale coal seam gas content prediction model training module, which is used to use different machine learning algorithms to construct initial well logging scale coal seam gas content prediction models for the well logging scale data; for any machine learning algorithm, according to each training result, the parameters of the well logging scale coal seam gas content prediction model are optimized using the Bayesian hyperparameter optimization method, and the well logging scale coal seam gas content prediction error is calculated; among the different machine learning algorithms, the machine learning algorithm with the smallest well logging scale coal seam gas content prediction error is selected to establish a trained well logging scale coal seam gas content prediction model.

[0177] In one possible implementation, the device also includes a geological-scale coal seam gas content prediction model training module, which is used to use different machine learning algorithms to construct initial geological-scale coal seam gas content prediction models for the geological-scale data; for any machine learning algorithm, according to each training result, the parameters of the geological-scale coal seam gas content prediction model are optimized using the Bayesian hyperparameter optimization method, and the geological-scale coal seam gas content prediction error is calculated; among the different machine learning algorithms, the machine learning algorithm with the smallest geological-scale coal seam gas content prediction error is selected to establish a trained geological-scale coal seam gas content prediction model.

[0178] In one possible implementation, the calculation formula for the coal seam gas content prediction value in the nonlinear fusion module 408 is:

[0179]

[0180] Where, Represents the predicted value of coal seam gas content; RF (1,2,3) represent the seismic scale coal seam gas content prediction model, the well logging scale coal seam gas content prediction model and the geological scale coal seam gas content prediction model respectively; W l(1,2,3) represents the weight of the coal seam gas content prediction model at different scales of the lth layer, b l represents the bias of the lth layer, F l represents the activation function of the lth layer, l = 1, 2, …, L.

[0181] In one possible implementation, the data preprocessing module 404 is specifically used to perform mathematical transformation on the seismic data to obtain seismic attributes related to the coal seam gas content; for the well logging data, a three-dimensional data volume of the same scale as the seismic data is obtained through a high-resolution inversion method; for the geological data, geological attributes of the same scale as the seismic data are obtained through indirect reflection through the seismic attributes and / or the three-dimensional data volume.

[0182] In one possible implementation, the geological data in the data preprocessing module 404 includes coal body structure and fracture data, fault data, and discrete factor data; accordingly, the geological data is indirectly reflected through the seismic attributes and / or the three-dimensional data volume, including: indirectly reflecting the coal body structure and fracture data through density values, wherein the density values ​​are determined through the well logging data; indirectly reflecting the fault data through dip attributes, wherein the dip attributes are determined through the seismic attributes; and determining estimated values ​​of each discrete factor data through the Kriging interpolation method, wherein the formula for determining the estimated values ​​of each discrete factor data is:

[0183]

[0184] In the formula, G(x0,y0) represents the estimated value of the point (x0,y0) in the spatiotemporal domain; G(x i ,y j ) are the discrete factor data; k(i,j) are the weight coefficients of different spatial points.

[0185] The coal seam gas content prediction device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effects are similar, and will not be described in detail in this embodiment.

[0186] Figure 5 This is a schematic diagram of the structure of the computer device provided in the embodiment of the present application. Figure 5 As shown, the computer device provided in this embodiment includes: at least one processor 501 and a memory 502. Optionally, the computer device also includes a communication component 503. The processor 501, the memory 502 and the communication component 503 are connected via a bus 504.

[0187] In a specific implementation process, at least one processor 501 executes the computer-executable instructions stored in the memory 502, so that the at least one processor 501 performs the above method.

[0188] The specific implementation process of the processor 501 can be found in the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here in this embodiment.

[0189] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the present invention may be directly implemented by a hardware processor or implemented by a combination of hardware and software modules in the processor.

[0190] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (NVM), such as at least one disk memory.

[0191] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be classified into address buses, data buses, and control buses. For ease of illustration, the buses in the drawings of this application are not limited to just one bus or just one type of bus.

[0192] An embodiment of the present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0193] An embodiment of the present application further provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above-mentioned method is implemented.

[0194] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile memory 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, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0195] An exemplary readable storage medium is coupled to a 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 an integral part 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 in the device as discrete components.

[0196] The division of units is merely a logical functional division; actual implementations may employ alternative divisions, such as combining or integrating multiple units or components into another system, or omitting or disabling certain features. Furthermore, any direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between devices or units, either through an interface, electrical, mechanical, or other means.

[0197] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0198] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0199] If the 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, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0200] Those skilled in the art will appreciate that all or part of the steps in the above-described method embodiments can be implemented using hardware associated with program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0201] Finally, it should be noted that those skilled in the art will readily identify other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art not disclosed herein. The present invention is not limited to the precise structure described above and illustrated in the accompanying drawings, and various modifications and variations may be made without departing from the scope thereof. The scope of the present invention is limited solely by the appended claims.

Claims

1. A method for predicting coal seam gas content, characterized in that: Applicable to computer equipment, including: Acquire multi-scale data from different sources in the target area; Obtaining a correspondence between coal seam gas content and the multi-scale data from different sources, and producing a plurality of cross-plots based on the correspondence; Determining main controlling factor data of the coal seam gas content according to the plurality of cross-plots, wherein the main controlling factor data includes seismic data, well logging data and geological data; The main controlling factor data is subjected to data preprocessing to obtain parameter data of the same scale, wherein the data preprocessing of the main controlling factor data includes: performing mathematical transformation on the seismic data to obtain seismic attributes related to the gas content of the coal seam; for the well logging data, obtaining a three-dimensional data volume of the same scale as the seismic data through a high-resolution inversion method; for the geological data, obtaining geological attributes of the same scale as the seismic data through indirect reflection of the seismic attributes and / or the three-dimensional data volume; Inputting the parameter data of the same scale into a pre-trained seismic scale coal seam gas content prediction model to obtain a seismic scale coal seam gas content prediction result of the target area; Inputting the parameter data of the same scale into a pre-trained well logging scale coal seam gas content prediction model to obtain a well logging scale coal seam gas content prediction result of the target area; Inputting the parameter data of the same scale into a pre-trained geological scale coal seam gas content prediction model to obtain a geological scale coal seam gas content prediction result of the target area; Using a nonlinear integration method, nonlinearly fuse the seismic scale coal seam gas content prediction result, the well logging scale coal seam gas content prediction result, and the geological scale coal seam gas content prediction result to obtain a coal seam gas content prediction value; According to the predicted value of coal seam gas content, a planar distribution map of coal seam gas content in the target area is output.

2. The method according to claim 1, characterized in that The parameter data includes seismic scale data; Accordingly, the training process of the seismic-scale coal seam gas content prediction model includes: Using different machine learning algorithms on the seismic scale data to construct initial seismic scale coal seam gas content prediction models; For any machine learning algorithm, based on each training result, the parameters of the seismic-scale coal seam gas content prediction model are optimized using the Bayesian hyperparameter optimization method, and the seismic-scale coal seam gas content prediction error is calculated; Among the different machine learning algorithms, the machine learning algorithm with the smallest seismic-scale coal seam gas content prediction error is selected to establish a trained seismic-scale coal seam gas content prediction model.

3. The method according to claim 2, characterized in that The machine learning algorithm is a random forest algorithm; Accordingly, the training process of the seismic-scale coal seam gas content prediction model includes: Generating a plurality of different training subsets by randomly sampling samples with replacement, wherein each training subset includes parameter data in the sample well; For each data subset, a decision tree is constructed and multiple feature data are randomly selected for subset data modeling; Selecting an optimal characteristic value from the plurality of characteristic data as a dividing point according to a preset criterion, and dividing the coal seam gas content data; The two data regions obtained by the division are continuously split according to the preset criteria until a preset stopping condition is satisfied, thereby completing the establishment of a seismic-scale coal seam gas content prediction model corresponding to the data subset; The seismic scale coal seam gas content prediction models corresponding to all data subsets constitute the seismic scale coal seam gas content prediction model.

4. The method according to claim 3, characterized in that The formula for dividing the coal seam gas content data is: Where y i The distribution is the gas content value of coal seams in different areas; Represents the region R1(j,s)={x|x j The average value of coal seam gas content in ≤s}; Represents the region R2(j,s)={x|x j The average gas content of coal seams ≥s}; x j represents the jth feature in the subset; s represents the split point; N1 represents the number of samples in the R1 region, and N2 represents the number of samples in the R2 region; MSE(j,i) represents traversing the randomly selected multiple feature data and taking the feature value with the smallest MSE as the optimal feature value.

5. The method according to claim 3, characterized in that The calculation formula of the prediction result of the seismic scale coal seam gas content prediction model is: Where, T b represents the coal seam gas content prediction model of each subset; x test represents each training subset; represents the coal seam gas content prediction result given by the coal seam gas content prediction model of each subset; B represents the number of subsets; It represents the prediction result of the seismic scale coal seam gas content prediction model.

6. The method according to claim 2, characterized in that The parameter data also includes well logging scale data; Accordingly, the training process of the well logging scale coal seam gas content prediction model includes: Using different machine learning algorithms on the well logging scale data to respectively construct initial well logging scale coal seam gas content prediction models; For any machine learning algorithm, based on each training result, the parameters of the well logging scale coal seam gas content prediction model are optimized using the Bayesian hyperparameter optimization method, and the well logging scale coal seam gas content prediction error is calculated; Among the different machine learning algorithms, the machine learning algorithm with the smallest logging-scale coal seam gas content prediction error is selected to establish a trained logging-scale coal seam gas content prediction model.

7. The method according to claim 2, characterized in that The parameter data also includes geological scale data; Accordingly, the training process of the geological scale coal seam gas content prediction model includes: Using different machine learning algorithms on the geological scale data to construct initial geological scale coal seam gas content prediction models; For any machine learning algorithm, based on each training result, the parameters of the geological-scale coal seam gas content prediction model are optimized using the Bayesian hyperparameter optimization method, and the geological-scale coal seam gas content prediction error is calculated; Among the different machine learning algorithms, the machine learning algorithm with the smallest geological-scale coal seam gas content prediction error is selected to establish a trained geological-scale coal seam gas content prediction model.

8. The method according to claim 1, characterized in that The calculation formula for the predicted value of coal seam gas content is: Where, Represents the predicted value of coal seam gas content; RF (1,2,3) represent the seismic scale coal seam gas content prediction model, the well logging scale coal seam gas content prediction model and the geological scale coal seam gas content prediction model respectively; W l (1,2,3) represents the weight of the coal seam gas content prediction model at different scales of the lth layer, b l represents the bias of the lth layer, F l Represents the activation function of the lth layer, l=1,2,…,L.

9. The method according to claim 1, characterized in that The geological data includes coal body structure and fracture data, fault data and discrete factor data; Accordingly, the indirect reflection of the geological data through the seismic attributes and / or the three-dimensional data volume includes: indirectly reflecting the coal body structure and fracture data through density values, wherein the density values ​​are determined by the well logging data; indirectly reflecting the fault data through a dip attribute, wherein the dip attribute is determined by the seismic attribute; The estimated value of each discrete factor data is determined by the Kriging interpolation method, wherein the formula for determining the estimated value of each discrete factor data is: In the formula, G(x0,y0) represents the estimated value of the point (x0,y0) in the spatiotemporal domain; G(x i ,y j ) are the discrete factor data; k(i,j) are the weight coefficients of different spatial points.

Images