A quantitative characterization and prediction method for coal seam microstructure

Abstract

The present invention discloses a method for quantitatively characterizing and predicting coal seam microstructures, comprising observing and counting the microstructure development on the surface of full-diameter coal core samples, counting the number of microstructures on the surface of the coal core within circles of different radii starting from the center of the full-diameter coal core, establishing a logarithmic coordinate system with radius as the independent variable and the number of microstructures as the dependent variable, performing linear regression on the data points in the logarithmic coordinate system, and obtaining the slope of the regression equation as the microstructure radius dimension, which serves as the quantitative characterization of the coal seam microstructure; counting the acid-soluble matter content and calibrating the well logging data; establishing a two-layer artificial neural network model; and dividing all coal core samples into training and test sets to establish a two-layer artificial neural network model capable of quantitatively predicting coal seam microstructures. The present invention can achieve quantitative characterization and prediction of coal seam microstructures and is applicable to coal samples of different environments and lithologies, and has important guiding significance for development of deep coalbed methane and other restricted areas.

Description

Technical Field

[0001] The present invention specifically relates to a method for quantitative characterization and prediction of coal seam microstructure, and belongs to the technical field of unconventional natural gas exploration and development. Background Art

[0002] Inspired by shale gas volume fracturing, deep coalbed methane development requires the formation of a fracture network, similar to shale gas development, to avoid the same path as shallow coalbed methane. Traditionally, the view was that coal seams could not form fracture networks like shale because shale is a typically brittle rock, while coal seams are typically plastic formations. According to the criteria for fracture network formation in shale fracturing, deep coal seams lack the objective conditions for this to occur. However, this is not the case. Some researchers have discovered that, because deep coal seams differ fundamentally from shale, fracture networks can be formed in deep coal seams under conditions such as great thickness and high gas content. Just as natural fractures are the most important characteristic of shale, the randomly distributed and nearly orthogonal microstructure within the coal body is the most important characteristic of deep coal seams. Microstructure distinguishes coal from other reservoirs and serves as the primary seepage pathway for gas and water, crucially influencing coalbed methane extraction. Most importantly, microstructure provides the objective basis for the formation and extension of fracture networks. Therefore, a detailed description of microstructural morphology is necessary. Fracture research is a hot topic both domestically and internationally, and is also highly sophisticated. In fracture research, fractures are characterized using five parameters: fracture aperture, aperture roughness, spatial correlation length of the aperture field, fluctuation coefficient of the fracture trend surface, and spatial correlation length of the trend surface. Obviously, microstructure can also be characterized using similar parameters, but deep coal seam microstructure is unique, and using too many parameters would be difficult to guide field practice. Therefore, conducting research on the quantitative characterization and prediction of microstructure can provide effective guidance for understanding and developing deep coalbed methane. Summary of the Invention

[0003] The present invention mainly overcomes the shortcomings of the existing technology. The present invention provides a quantitative characterization and prediction method for coal seam microstructure. After inputting conventional coal seam logging parameters, the present invention can effectively predict the development of coal seam microstructure, which has important guiding significance for understanding and developing deep coalbed methane and other development restricted areas.

[0004] The present invention solves the above technical problems and provides a technical solution: a method for quantitative characterization and prediction of coal seam microstructure, comprising the following steps:

[0005] Step 1: Acquire multiple full-diameter natural coal cores in the target area, and obtain the density of the coal seam, the density of the cap layer, the density of the bottom layer, the acoustic transit time of the coal seam, the acoustic transit time of the cap layer, the acoustic transit time of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, the natural gamma of the bottom layer, compensated neutrons, and the resistivity of each full-diameter natural coal core;

[0006] Step 2: Based on fractal theory, a series of circles of different radii are generated from the center of each full-diameter natural coal core, and the number of microstructures on the surface of the coal core within circles of different sizes is observed and counted;

[0007] Step 3: Using radius as the independent variable and the number of microstructures as the dependent variable, a logarithmic coordinate system is established to perform linear regression on the data points to obtain the microstructure radius dimension of each full-diameter natural coal core;

[0008] Step 4: Perform X-ray diffraction analysis on each full-diameter natural coal core, calculate the content of acid-soluble matter, and calibrate the logging data based on the acid-soluble matter content;

[0009] Step 5: Construct a data set based on the density of each coal column, the density of the cap layer, the density of the bottom layer, the acoustic transit time of the coal seam, the acoustic transit time of the cap layer, the acoustic transit time of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, the natural gamma of the bottom layer, compensated neutrons, resistivity, acid-soluble matter content, and microstructure radius dimension;

[0010] Step 6: Using the density of the coal seam, the density of the cap layer, the density of the bottom layer, the acoustic transit time of the coal seam, the acoustic transit time of the cap layer, the acoustic transit time of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, and the natural gamma of the bottom layer as inputs, and the acid-soluble matter content as output, a first-layer artificial neural network is constructed; then, using the acid-soluble matter content, compensated neutrons, and resistivity as inputs, and the microstructure radius dimension as output, a second-layer artificial neural network is constructed;

[0011] Step 7: Standardize the data set and divide it into a training set and a test set. Based on the training set, continuously adjust the structure and parameters of the two-layer artificial neural network. The network with the highest accuracy in the test set is the final quantitative prediction model for coal seam microstructure. The microstructure radius dimension is predicted through the quantitative prediction model for coal seam microstructure.

[0012] A further technical solution is to obtain full-diameter natural coal core samples with a diameter of 100 mm at different depths in the target area in step 1, and cut and polish the cross-section of the full-diameter natural coal core samples to make their surfaces flat and smooth for easy statistics.

[0013] A further technical solution is that in step 2, circles with radii of 5 mm, 10 mm, 15 mm, 20 mm, 25 mm, 30 mm, 35 mm, 40 mm, 45 mm, and 50 mm are generated from the center of each full-diameter natural coal core.

[0014] A further technical solution is that the specific process of step 3 is:

[0015] Step 31: Radius As the independent variable sequence, the number of microstructures As the dependent variable sequence, that is:

[0016]

[0017] Where: m is the number of samples collected; n is the number of circles with different radii;

[0018] According to fractal theory, the radius r and the number of microstructures within the circle under this radius It obeys a scale-invariant power law relationship, namely:

[0019]

[0020] Where: D is the dimension of the microstructure radius; C is a constant;

[0021] Step 32: Radius and the number of microstructures Taking the logarithm gives and , establish a logarithmic coordinate system;

[0022] Step 33: Based on the data scatter points in the logarithmic coordinate system Performing linear regression, we can get from fractal theory:

[0023]

[0024] Where: is the slope of the regression equation, is a series of constants.

[0025] A further technical solution is that the specific process of step 4 is:

[0026] Step 41: Take the filling material in the full-diameter natural coal core and perform an X-ray diffraction experiment to calculate the acid-soluble matter content of each full-diameter natural coal core;

[0027] Step 42: Calibrate the acid-soluble matter content of the corresponding full-diameter natural coal core and the three logging parameters of density, acoustic transit time, and natural gamma at the corresponding depth according to the coring depth of the full-diameter natural coal core.

[0028] A further technical solution is that the specific process of step 7 is:

[0029] Step 701: perform the following standardization processing on the data set and divide the data set into a training set and a test set;

[0030] Step 702: Taking the density of the coal seam, the density of the cap layer, the density of the bottom layer, the acoustic time difference of the coal seam, the acoustic time difference of the cap layer, the acoustic time difference of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, and the natural gamma of the bottom layer as inputs, and the acid soluble matter content as output, set the learning rate of the network to , the number of iterations is , the number of hidden layer nodes is , randomly initialize the weights from the input layer to the hidden layer in the network in the range of (0,1) and threshold , and the weights from the hidden layer to the output layer and threshold ;

[0031] Step 703: According to the activation function from the input layer to the hidden layer Calculate the output of hidden layer neurons ;

[0032] Step 704: Based on the activation function from the hidden layer to the output layer Calculate the The prediction results of the microstructure radius dimension of the training set samples ;

[0033] Step 705: Calculate the output layer neuron gradient term and the gradient term of the hidden layer neurons ;

[0034] Step 706: Update all values and thresholds in the network;

[0035] Step 707: Repeat steps 703 to 706 until the number of iterations reaches T , and obtain the first layer of artificial neural network with acid-soluble matter content as output;

[0036] Step 708: With the acid soluble content, compensation neutrons, and resistivity as inputs and the microstructure radius dimension as output, the learning rate of the network is set to , the number of iterations is , the number of hidden layer nodes is ,exist Randomly initialize the weights from the input layer to the hidden layer in the network within the range and threshold , and the weights from the hidden layer to the output layer and threshold ;

[0037] Step 709: Calculate and iterate in the same way as in the first layer of the artificial neural network until the number of iterations reaches T , we get the second layer artificial neural network with the dimension of microstructure radius as output;

[0038] Step 710: For the established two-layer artificial neural network, input the test set data to predict the microstructure radius dimension, perform denormalization on the prediction results and compare them with the true value of the test set. If the prediction effect is not ideal, return to step 702 to adjust the network learning rate and the number of hidden layer nodes until the prediction accuracy of the test set meets the requirements. The resulting two-layer artificial neural network is the coal seam microstructure quantitative prediction model;

[0039] Step 711: Finally, the microstructure radius dimension is predicted using the coal seam microstructure quantitative prediction model.

[0040] The beneficial effects of the present invention are as follows: the present invention can quantitatively characterize and predict the acid-soluble matter content and microstructural development of coal seams, and is also universally applicable to coal rock samples of different environments and lithologies. It is an effective tool for understanding and analyzing deep coal rock microstructures and complex hydraulic fracture networks, and has important guiding significance for the development of deep coalbed methane and other restricted areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 Research technology roadmap for this invention;

[0042] Figure 2 This is a full-diameter coal core sample diagram;

[0043] Figure 3 This is the microstructural distribution map of the coal core section;

[0044] Figure 4 Schematic diagram for solving the radius dimension of microstructure of a coal core sample;

[0045] Figure 5 This is the structural diagram of the coal seam microstructure prediction model;

[0046] Figure 6 This is the prediction result graph of the test set samples. DETAILED DESCRIPTION

[0047] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0048] like Figure 1 As shown, a method for quantitative characterization and prediction of coal seam microstructure of the present invention comprises the following steps:

[0049] Step 1: Acquire multiple full-diameter natural coal cores at different depths in the target area, and obtain parameters such as the density of the coal seam, the density of the cap layer, the density of the bottom layer, the acoustic time difference of the coal seam, the acoustic time difference of the cap layer, the acoustic time difference of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, the natural gamma of the bottom layer, compensated neutrons, and resistivity of each full-diameter natural coal core;

[0050] Step 2: Based on fractal theory, circles of 5 mm, 10 mm, 15 mm, 20 mm, 25 mm, 30 mm, 35 mm, 40 mm, 45 mm, and 50 mm were drawn from the center of each full-diameter natural coal core, and the number of microstructures on the surface of the coal core within the circles of different sizes was observed and counted;

[0051] Step 3: Using radius as the independent variable and the number of microstructures as the dependent variable, a logarithmic coordinate system is established to perform linear regression on the data points to obtain the microstructure radius dimension of each full-diameter natural coal core;

[0052] Step 31: Radius As the independent variable sequence, the number of microstructures As the dependent variable sequence, that is:

[0053] (1)

[0054] Where: m is the number of samples collected; n is the number of circles with different radii;

[0055] According to fractal theory, the radius r and the number of microstructures within the circle under this radius It obeys a scale-invariant power law relationship, namely:

[0056] (2)

[0057] Where: D is the dimension of the microstructure radius; C is a constant;

[0058] Step 32: Radius and the number of microstructures Taking the logarithm gives and , establish a logarithmic coordinate system;

[0059] Step 33: Based on the data scatter points in the logarithmic coordinate system Performing linear regression, we can get from fractal theory:

[0060] (3)

[0061] Where: is the slope of the regression equation, is a series of constants;

[0062] Step 4: Perform X-ray diffraction analysis on each full-diameter natural coal core, calculate the content of acid-soluble matter, and calibrate the logging data based on the acid-soluble matter content;

[0063] Step 41: Take the filling material in the full-diameter natural coal core and perform an X-ray diffraction experiment to calculate the acid-soluble matter content of each full-diameter natural coal core;

[0064] Step 42: Calibrate the acid-soluble content of the full-diameter natural coal core and the three logging parameters of density, acoustic transit time, and natural gamma at the corresponding depth according to the coring depth of the full-diameter natural coal core;

[0065] Step 5: Construct a data set based on the density of each coal column, the density of the cap layer, the density of the bottom layer, the acoustic transit time of the coal seam, the acoustic transit time of the cap layer, the acoustic transit time of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, the natural gamma of the bottom layer, compensated neutrons, resistivity, acid-soluble matter content, and microstructure radius dimension;

[0066] Step 6: Using the density of the coal seam, the density of the cap layer, the density of the bottom layer, the acoustic transit time of the coal seam, the acoustic transit time of the cap layer, the acoustic transit time of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, and the natural gamma of the bottom layer as inputs, and the acid-soluble matter content as output, a first-layer artificial neural network is constructed; then, using the acid-soluble matter content, compensated neutrons, and resistivity as inputs, and the microstructure radius dimension as output, a second-layer artificial neural network is constructed;

[0067] Step 7: Standardize the data set and divide it into a training set and a test set. Based on the training set, continuously adjust the structure and parameters of the two-layer artificial neural network. The network with the highest accuracy in the test set is the final coal seam microstructure quantitative prediction model. The microstructure radius dimension is predicted using the coal seam microstructure quantitative prediction model.

[0068] Step 701: perform the following standardization processing on the data set and divide the data set into a training set and a test set;

[0069] (4)

[0070] Step 702: Taking the density of the coal seam, the density of the cap layer, the density of the bottom layer, the acoustic time difference of the coal seam, the acoustic time difference of the cap layer, the acoustic time difference of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, and the natural gamma of the bottom layer as inputs, and the acid soluble matter content as output, set the learning rate of the network to , the number of iterations is , the number of hidden layer nodes is , randomly initialize the weights from the input layer to the hidden layer in the network in the range of (0,1) and threshold , and the weights from the hidden layer to the output layer and threshold ;

[0071] Step 703: According to the activation function from the input layer to the hidden layer Calculate the output of hidden layer neurons ;

[0072] (5)

[0073] Step 704: Based on the activation function from the hidden layer to the output layer Calculate the The prediction results of the microstructure radius dimension of the training set samples ;

[0074] (6)

[0075] Step 705: Calculate the output layer neuron gradient term and the gradient term of the hidden layer neurons ;

[0076] (7)

[0077] (8)

[0078] Step 706: Update all values and thresholds in the network;

[0079] (9)

[0080] (10)

[0081] (11)

[0082] (12)

[0083] Step 707: Repeat steps 703 to 706 until the number of iterations reaches T , and obtain the first layer of artificial neural network with acid-soluble matter content as output;

[0084] Step 708: With the acid soluble content, compensation neutrons, and resistivity as inputs and the microstructure radius dimension as output, the learning rate of the network is set to , the number of iterations is , the number of hidden layer nodes is ,exist Randomly initialize the weights from the input layer to the hidden layer in the network within the range and threshold , and the weights from the hidden layer to the output layer and threshold ;

[0085] Step 709: Calculate and iterate in the same way as in the first layer of the artificial neural network until the number of iterations reaches T , we get the second layer artificial neural network with the dimension of microstructure radius as output;

[0086] Step 710: For the established two-layer artificial neural network, input the test set data to predict the microstructure radius dimension, perform denormalization on the prediction results and compare them with the true value of the test set. If the prediction effect is not ideal, return to step 702 to adjust the network learning rate and the number of hidden layer nodes until the prediction accuracy of the test set meets the requirements. The resulting two-layer artificial neural network is the coal seam microstructure quantitative prediction model;

[0087] Step 711: Finally, the microstructure radius dimension is predicted using the coal seam microstructure quantitative prediction model.

[0088] Example

[0089] Step 1: Obtain 10 full-diameter natural coal cores at different depths in the target area;

[0090] Step 2: Based on fractal theory, circles with radii of 5 mm, 10 mm, 15 mm, 20 mm, 25 mm, 30 mm, 35 mm, 40 mm, 45 mm, and 50 mm were drawn from the center of each full-diameter natural coal core. The number of microstructures on the surface of the coal core within the circles of different sizes was observed and counted (the results are shown in Table 1).

[0091] Table 1 Statistical results of microstructures in circles of different radii of a coal core sample

[0092]

[0093] Step 3: Using radius as the independent variable and the number of microstructures as the dependent variable, a logarithmic coordinate system is established to perform linear regression on the data points to obtain the microstructure radius dimension of each full-diameter natural coal core;

[0094] The data scatter points of the sample coal core in the logarithmic coordinate system are shown in Table 2;

[0095] Table 2 Data scatter results of sample coal core samples in logarithmic coordinate system

[0096]

[0097] The quantitative characterization results of the microstructural radius dimension of all coal core samples are shown in Table 3;

[0098] Table 3 Quantitative characterization results of microstructure of all coal core samples

[0099]

[0100] Step 4: Perform X-ray diffraction analysis on each full-diameter natural coal core (as shown in Table 4), calculate the content of acid-soluble matter, and calibrate the logging data based on the acid-soluble matter content;

[0101] Table 4 X-ray diffraction analysis results of fillers

[0102]

[0103] Step 5: Construct a data set based on the density of each coal column, the density of the cap layer, the density of the bottom layer, the acoustic transit time of the coal seam, the acoustic transit time of the cap layer, the acoustic transit time of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, the natural gamma of the bottom layer, compensated neutrons, resistivity, acid-soluble matter content, and microstructure radius dimension;

[0104] Step 6: Using the density of the coal seam, the density of the cap layer, the density of the bottom layer, the acoustic transit time of the coal seam, the acoustic transit time of the cap layer, the acoustic transit time of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, and the natural gamma of the bottom layer as inputs, and the acid-soluble matter content as output, a first-layer artificial neural network is constructed; then, using the acid-soluble matter content, compensated neutrons, and resistivity as inputs, and the microstructure radius dimension as output, a second-layer artificial neural network is constructed;

[0105] Step 7: Standardize the data set (as shown in Table 5) and divide it into a training set and a test set. Based on the training set, continuously adjust the structure and parameters of the two-layer artificial neural network. The network with the highest accuracy in the test set is the final coal seam microstructure quantitative prediction model. The microstructure radius dimension is predicted using the coal seam microstructure quantitative prediction model.

[0106] (1) The data set is standardized and the standardized data table is shown in Table 5;

[0107] Table 5 Standardized data table

[0108]

[0109] (2) Set the learning rate of the first layer of artificial neural network to 0.001, the number of iterations to 1000, and the number of hidden layer nodes to 10. Randomly initialize the weights from the input layer to the hidden layer in the network within the range and threshold , and the weights from the hidden layer to the output layer and threshold ;

[0110] (3) According to the activation function from the input layer to the hidden layer Calculate the output of hidden layer neurons ;

[0111] (4) According to the activation function from hidden layer to output layer Calculate the The prediction results of the microstructure radius dimension of the training set samples ;

[0112] (5) Calculate the gradient term of the output layer neurons and the gradient term of the hidden layer neurons ;

[0113] (6) Update all value and thresholds in the network;

[0114] (7) Repeat steps (3) to (7) until the number of iterations reaches 1000, and obtain the first layer of artificial neural network with acid-soluble matter content as output;

[0115] (8) Divide the standardized dataset into training set and test set. The divided training set and test set are shown in Table 6.

[0116] Table 6. Results of normalized training set and test set division

[0117]

[0118] (9) Input the normalized training data set;

[0119] (13)

[0120] Where: yes OK The input parameter matrix of columns; The table is Output column vector with 1 row and 1 column; is the number of input parameters; is the number of samples in the training set;

[0121] (10) Set the learning rate of the second layer artificial neural network to 0.001, the number of iterations to 1000, the number of hidden layer nodes to 5, and Randomly initialize the weights from the input layer to the hidden layer in the network within the range and threshold , and the weights from the hidden layer to the output layer and threshold ;

[0122] (11) Calculation and iteration are performed in the same manner as above until the number of iterations reaches 1000, and a second-layer artificial neural network with the coal seam microstructure radius dimension as output is obtained;

[0123] (12) The test set data is input to predict the dimension of the microstructure radius and perform denormalization, and the structure and parameters of the network are continuously adjusted to finally obtain a two-layer artificial neural network with good results, which is used as the final quantitative prediction model for coal seam microstructure.

[0124] (12) The prediction results of the final prediction model for 10 samples in the test set are shown in Table 7. The mean absolute percentage error is used as an evaluation indicator of the model prediction effect, which can provide effective support for understanding the microstructural development of coal seams.

[0125] Table 7. Prediction results of two-layer artificial neural network

[0126]

[0127] The above description does not limit the present invention in any form. Although the present invention has been disclosed through the above embodiments, it is not intended to limit the present invention. Any technician familiar with the profession can use the technical content disclosed above to make some changes or modifications to equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for quantitative characterization and prediction of coal seam microstructure, characterized in that: The following steps are involved: Step 1: Acquire multiple full-diameter natural coal cores in the target area, and obtain the density of the coal seam, the density of the cap layer, the density of the bottom layer, the acoustic transit time of the coal seam, the acoustic transit time of the cap layer, the acoustic transit time of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, the natural gamma of the bottom layer, compensated neutrons, and the resistivity of each full-diameter natural coal core; Step 2: Based on fractal theory, a series of circles of different radii are generated from the center of each full-diameter natural coal core, and the number of microstructures on the surface of the coal core within circles of different sizes is observed and counted; Step 3: Using radius as the independent variable and the number of microstructures as the dependent variable, a logarithmic coordinate system is established to perform linear regression on the data points to obtain the microstructure radius dimension of each full-diameter natural coal core; Step 4: Perform X-ray diffraction analysis on each full-diameter natural coal core, calculate the content of acid-soluble matter, and calibrate the logging data based on the acid-soluble matter content; Step 5: Construct a data set based on the density of each coal column, the density of the cap layer, the density of the bottom layer, the acoustic transit time of the coal seam, the acoustic transit time of the cap layer, the acoustic transit time of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, the natural gamma of the bottom layer, compensated neutrons, resistivity, acid-soluble matter content, and microstructure radius dimension; Step 6: Using the density of the coal seam, the density of the cap layer, the density of the bottom layer, the acoustic transit time of the coal seam, the acoustic transit time of the cap layer, the acoustic transit time of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, and the natural gamma of the bottom layer as inputs, and the acid-soluble matter content as output, a first-layer artificial neural network is constructed; then, using the acid-soluble matter content, compensated neutrons, and resistivity as inputs, and the microstructure radius dimension as output, a second-layer artificial neural network is constructed; Step 7: Standardize the data set and divide it into a training set and a test set. Based on the training set, continuously adjust the structure and parameters of the two-layer artificial neural network. The network with the highest accuracy in the test set is the final quantitative prediction model for coal seam microstructure. The microstructure radius dimension is predicted through the quantitative prediction model for coal seam microstructure.

2. A method for quantitative characterization and prediction of coal seam microstructure according to claim 1, characterized in that: In the step 1, full-diameter natural coal core samples with a diameter of 100 mm are obtained at different depths in the target area, and the cross-sections of the full-diameter natural coal core samples are cut and polished to make their surfaces flat and smooth for easy statistics.

3. The method for quantitative characterization and prediction of coal seam microstructure according to claim 1, characterized in that: In step 2, circles with radii of 5 mm, 10 mm, 15 mm, 20 mm, 25 mm, 30 mm, 35 mm, 40 mm, 45 mm, and 50 mm are generated from the center of each full-diameter natural coal core.

4. A method for quantitative characterization and prediction of coal seam microstructure according to claim 3, characterized in that: The specific process of step 3 is as follows: Step 31: Radius As the independent variable sequence, the number of microstructures As the dependent variable sequence, that is: Where: m is the number of samples collected; n is the number of circles with different radii; According to fractal theory, the radius r and the number of microstructures within the circle under this radius It obeys a scale-invariant power law relationship, namely: Where: D is the dimension of the microstructure radius; C is a constant; Step 32: Radius and the number of microstructures Taking the logarithm gives and , establish a logarithmic coordinate system; Step 33: Based on the data scatter points in the logarithmic coordinate system Performing linear regression, we can get from fractal theory: Where: is the slope of the regression equation, is a series of constants.

5. The method for quantitative characterization and prediction of coal seam microstructure according to claim 1, characterized in that: The specific process of step 4 is as follows: Step 41: Take the filling material in the full-diameter natural coal core and perform an X-ray diffraction experiment to calculate the acid-soluble matter content of each full-diameter natural coal core; Step 42: Calibrate the acid-soluble matter content of the corresponding full-diameter natural coal core and the three logging parameters of density, acoustic transit time, and natural gamma at the corresponding depth according to the coring depth of the full-diameter natural coal core.

6. The method for quantitative characterization and prediction of coal seam microstructure according to claim 1, characterized in that: The specific process of step 7 is as follows: Step 701: perform the following standardization processing on the data set and divide the data set into a training set and a test set; Step 702: Taking the density of the coal seam, the density of the cap layer, the density of the bottom layer, the acoustic time difference of the coal seam, the acoustic time difference of the cap layer, the acoustic time difference of the bottom layer, the natural gamma of the coal seam, the natural gamma of the cap layer, and the natural gamma of the bottom layer as inputs, and the acid soluble matter content as output, set the learning rate of the network to , the number of iterations is , the number of hidden layer nodes is , randomly initialize the weights from the input layer to the hidden layer in the network in the range of (0,1) and threshold , and the weights from the hidden layer to the output layer and threshold ; Step 703: According to the activation function from the input layer to the hidden layer Calculate the output of hidden layer neurons ; Step 704: Based on the activation function from the hidden layer to the output layer Calculate the The prediction results of the microstructure radius dimension of the training set samples ; Step 705: Calculate the output layer neuron gradient term and the gradient term of the hidden layer neurons ; Step 706: Update all values and thresholds in the network; Step 707: Repeat steps 703 to 706 until the number of iterations reaches T , and obtain the first layer of artificial neural network with acid-soluble matter content as output; Step 708: With the acid soluble content, compensation neutrons, and resistivity as inputs and the microstructure radius dimension as output, the learning rate of the network is set to , the number of iterations is , the number of hidden layer nodes is ,exist Randomly initialize the weights from the input layer to the hidden layer in the network within the range and threshold , and the weights from the hidden layer to the output layer and threshold ; Step 709: Calculate and iterate in the same way as in the first layer of the artificial neural network until the number of iterations reaches T , we get the second layer artificial neural network with the dimension of microstructure radius as output; Step 710: For the established two-layer artificial neural network, input the test set data to predict the microstructure radius dimension, perform denormalization on the prediction results and compare them with the true value of the test set. If the prediction effect is not ideal, return to step 702 to adjust the network learning rate and the number of hidden layer nodes until the prediction accuracy of the test set meets the requirements. The resulting two-layer artificial neural network is the coal seam microstructure quantitative prediction model; Step 711: Finally, the microstructure radius dimension is predicted using the coal seam microstructure quantitative prediction model.

Images