Method for evaluating hydraulic fracturing effect of coal seam under low-brightness background based on AI image recognition

Abstract

The invention relates to a coal seam hydraulic fracturing effect evaluation method under a low-brightness background based on AI image recognition, and belongs to the technical field of coal hydraulic fracturing. According to the method, a coal seam image in the fracturing process is collected through high-speed camera equipment, a low-brightness image is enhanced by utilizing a histogram equalization technology of Matlab, and graying processing is performed by adopting a weighted average method. On the basis of a convolutional neural network (CNN) model, the length, width, number, density, distribution and connectivity characteristics of cracks are automatically extracted, a multi-index evaluation system is established in combination with a fuzzy comprehensive evaluation method, a comprehensive score is calculated, and the anti-reflection effect grade is divided. The method solves the problems that a traditional method is low in image recognition precision and single in evaluation dimension in a low-brightness environment, quantitative and multi-dimensional scientific evaluation of the hydraulic fracturing effect is achieved, and reliable technical support is provided for efficient mining of coal bed gas.

Description

Technical field:

[0001] The present invention relates to the technical field of coal body hydraulic fracturing, and in particular to a method for evaluating the effect of coal seam hydraulic fracturing under a low-brightness background based on AI image recognition. Background technology:

[0002] my country's coalbed methane resources are abundant but their occurrence conditions are complex. The poor gas mobility within the coal bodies hinders efficient gas extraction. Hydraulic fracturing is an important technology for increasing the yield of gas-bearing coal seams. Due to its ease of use and low cost, it is used to increase the permeability of coal bodies in underground coal mines, thereby achieving the goal of co-mining coal and gas and reducing the risk of gas outburst accidents. However, when conducting hydraulic fracturing experiments based on coal seam characteristics, researchers often need to conduct corresponding permeability evaluations to verify the permeability-enhancing effect of a particular technology.

[0003] Currently, there are many methods for evaluating the effect of hydraulic fracturing in coal seams, but most of them measure the permeability before and after fracturing. This method often requires a large number of related technical operations, consumes a lot of manpower, material resources, and financial resources, and cannot accurately evaluate the effect of coal seam permeability enhancement. In addition, some researchers have established different hydraulic fracturing effect evaluation models based on the coal seam permeability enhancement process. This method can effectively evaluate the permeability enhancement effect, but it is not universal and its model cannot be applied to different coal seam characteristics and structures. With the development of computer technology, many technologies and methods for analyzing the permeability enhancement effect of coal bodies based on images have emerged. The hydraulic fracturing permeability enhancement effect is judged by the size or area of ​​the coal crack opening. However, the brightness of the coal mine environment is low, and the coal seam crack image obtained is poor. Therefore, the effect of hydraulic fracturing evaluation with the help of images is not satisfactory.

[0004] Based on this, it is urgent to propose a method to evaluate the permeability enhancement effect of hydraulic fracturing in coal seams under low-brightness background to solve the current problems. Summary of the invention:

[0005] The present invention aims to provide a method for evaluating the hydraulic fracturing effect of coal seams under low-light conditions based on AI image recognition, aiming to address the shortcomings of the aforementioned prior art. This method can effectively extract hydraulic fracturing cracks in coal seams under low-light conditions, effectively solving the problem of coal seam image processing under low-light conditions, and further evaluating the hydraulic fracturing effect.

[0006] To achieve the above objectives, the present invention provides a method for evaluating the hydraulic fracturing effect of coal seams under a low-brightness background based on AI image recognition. The specific steps include image acquisition, image preprocessing, image crack extraction, and hydraulic fracturing effect evaluation.

[0007] The method for evaluating the hydraulic fracturing effect of coal seams under low-light background based on AI image recognition refers to using high-speed camera image acquisition equipment to record the initiation and expansion images of cracks in the coal seam during the hydraulic fracturing process. The collected images are preprocessed with the help of Matlab software. To address the low-light background problem, the overall brightness of the image is enhanced by the histeq function in Matlab, making the crack information in the image more clearly visible. The image after brightness enhancement is further grayscaled to convert the color image into a grayscale image. With the help of convolutional neural network (CNN) data set training, the cracks in the image are effectively separated from the background, achieving accurate extraction of the cracks. Based on the extracted crack information, including the length, width, number and distribution of the cracks, a fuzzy comprehensive evaluation method is used to establish a hydraulic fracturing effect evaluation index based on Darcy's law, fracture mechanics and multiphase flow theory, so as to scientifically and accurately evaluate the hydraulic fracturing effect of the coal seam.

[0008] The image preprocessing refers to using the histeq function in Matlab to perform histogram equalization on the image to enhance the contrast and brightness of the image, and using the weighted average method to grayscale the color image after brightness enhancement. The calculation formula is: Gray = 0.299×R+0.587×G+0.114×B, where R, G, and B are the values ​​of the red, green, and blue channels of the color image, respectively, and Gray is the converted grayscale value.

[0009] The crack extraction refers to using Python to establish a convolutional neural network (CNN) model, using the model to automatically extract crack information from preprocessed images, establishing a data set by collecting a large number of downhole hydraulic fracturing crack images, preprocessing the acquired images using the histeq function in Matlab software, and dividing the preprocessed images in the data set into a training set, a validation set, and a test set. The training set is used to train the CNN model, and the cracks in the preprocessed images are manually labeled. During the training process, the validation set is used to monitor the performance of the model to prevent overfitting, and the parameters of the model are continuously adjusted with the help of the back-propagation algorithm to make the model's prediction results as close as possible to the labeled crack positions. Finally, the test set is used to evaluate the generalization ability of the model, that is, the model's recognition performance on new low-brightness coal seam crack images.

[0010] The hydraulic fracturing effect evaluation refers to establishing a comprehensive evaluation index system and setting the weight of each factor using a fuzzy comprehensive evaluation method. The evaluation process of this method is to determine the evaluation index and level, construct a fuzzy membership function, construct a fuzzy relationship matrix, determine the weight vector, calculate the fuzzy comprehensive evaluation result, and defuzzify. The weight value is determined based on the defuzzification result. The weight of each factor is set according to the fuzzy comprehensive evaluation result. According to the fuzzy comprehensive evaluation result, the weight of fracture width is set to 0.3. Wider fractures can provide better gas diversion channels and play a key role in the rapid release of coalbed methane. The weight of fracture length is 0.25, which is of great significance in expanding the range of gas migration. The weight of fracture number is also 0.15. A sufficient number of fractures can increase the gas diversion path. The calculated weight of fracture density is 0.1. In terms of fracture distribution, based on the requirement of uniform gas flow in the coal seam, uniformly distributed fractures are conducive to improving the overall diversion effect. After analysis and calculation, the weight of fracture density is set to 0.1. The better the fracture connectivity, the more conducive it is to gas diffusion in the coal seam. Based on theory and calculation, the weight of fracture density is set to 0.1.

[0011] Assume that the crack length score is L (calculated based on the ratio of actual length to expected length, such as when the actual length reaches 80% of the expected length, L = 0.8), the crack width score is W (also calculated based on the ratio of actual width to expected width), and the crack number score is N (calculated based on the ratio of actual number to expected number).

[0012] The crack density score is calculated by first setting an ideal crack density value Dideal and the actual crack density Dactual. The score is calculated using the formula SD = Dactual / Dideal. If SD ≥ 1, it means that the actual crack density meets or exceeds expectations and the score is 1. If SD < 1, the score is based on the actual proportion.

[0013] The crack distribution score extracts the coordinates of the crack center points and uses the K-Means clustering algorithm to group the cracks into clusters. Each cluster represents an area where cracks are concentrated. The crack distribution is described by calculating the number of clusters, the number of cracks within each cluster, and the location of the cluster center.

[0014] The crack connectivity score is obtained by processing the crack image using the cv2.connectedComponents function of OpenCV in Python, marking different connected regions, and then calculating the number and size of connected regions. The connected crack ratio can be defined as the ratio of the length or area of ​​connected cracks to the total crack length or area.

[0015] Comprehensive score S = L × length weight + W × width weight + N × number weight + D × density weight + U × distribution weight + C × connectivity weight.

[0016] Furthermore, by setting dynamic benchmark values ​​(a, b, c, d), the comprehensive score results are compared with the dynamic benchmark values ​​to judge the hydraulic fracturing effect. The physical meaning and determination method of the dynamic benchmark values ​​are as follows:

[0017] a: critical value of poor air permeability, corresponding to the lowest acceptable threshold of the comprehensive score;

[0018] b: critical value of medium air permeability;

[0019] c: critical value for good air permeability;

[0020] d: Critical value for excellent air permeability.

[0021] The method for determining the dynamic benchmark value is to measure the permeability coefficient of the coal seam after hydraulic fracturing with the radial flow method based on the coal industry standard MT / T 1173-2019 of the People's Republic of China, determine the permeability level based on the size of the permeability coefficient, divide the grades (poor, medium, good, excellent) according to the permeability test results, and calculate the comprehensive score of the corresponding distribution range. By testing the hydraulic fracturing permeability coefficients of different coal seams multiple times, the comprehensive score S corresponding to different permeability levels is calculated respectively, and the values ​​of a, b, c, and d are determined by the average of the comprehensive scores under a certain permeability.

[0022] The permeability of coal can be measured by the permeability coefficient, which is divided into the following categories:

[0023] Excellent air permeability: air permeability coefficient is greater than 10m2 / (mbar-day);

[0024] Good air permeability: air permeability coefficient is 5-10 m2 / (mbar day);

[0025] Medium air permeability: air permeability coefficient is 1 to 5 square meters / (mbar day);

[0026] Poor air permeability: the air permeability coefficient is 0.1 to 1 square meters / (mbar day);

[0027] Very poor air permeability: air permeability coefficient is less than 0.1 m2 / (mbar day).

[0028] Based on the calculated comprehensive scores, the classification standards for hydraulic fracturing permeability enhancement effects are given, as shown in Table 1:

[0029] Table 1 Classification standard of hydraulic fracturing permeability enhancement effect

[0030]

[0031] The present invention adopts the above technical solution and has the following beneficial effects:

[0032] The present invention proposes a method for evaluating the hydraulic fracturing effect of coal seams under a low-brightness background based on AI image recognition. This method can better evaluate the hydraulic fracturing effect of coal seams under low-brightness conditions. The image is histogram-equalized using the histeq function in Matlab, and the contrast and brightness of the image are enhanced with the help of an algorithm. Then, a convolutional neural network is established using Python. After model training, the crack information is distinguished from the background, and parameter information such as the length, width, and number of cracks is further calculated. The hydraulic fracturing effect is evaluated according to the proposed hydraulic fracturing effect evaluation formula and grading standard. This method can quantitatively characterize the hydraulic fracturing permeability-enhancing effect, and the multi-dimensional evaluation method can more accurately reflect the actual fracturing effect, providing a more valuable reference for subsequent mining operations. Description of the drawings:

[0033] Figure 1 It is a schematic flow diagram of the present invention. Implementation method:

[0034] like Figure 1 The present invention aims to provide a method for evaluating the hydraulic fracturing effect of coal seams under low-light backgrounds based on AI image recognition, aiming to overcome the shortcomings of existing technologies. This method can effectively extract hydraulic fracturing cracks in coal seams under low-light backgrounds, effectively solving the problem of coal seam image processing under low-light backgrounds, and further evaluating the hydraulic fracturing effect.

[0035] To achieve the above objectives, the present invention provides a method for evaluating the hydraulic fracturing effect of coal seams under a low-brightness background based on AI image recognition. The specific steps include image acquisition, image preprocessing, image crack extraction, and hydraulic fracturing effect evaluation.

[0036] The method for evaluating the hydraulic fracturing effect of coal seams under a low-light background based on AI image recognition refers to using a high-speed camera image acquisition device to record the initiation and expansion images of cracks in the coal seam during the hydraulic fracturing process. The collected images are preprocessed with the help of Matlab software. To address the low-light background problem, the overall brightness of the image is enhanced by the histeq function in Matlab, making the crack information in the image more clearly visible. The image after brightness enhancement is further grayscaled to convert the color image into a grayscale image. With the help of convolutional neural network (CNN) data set training, the cracks in the image are effectively separated from the background, achieving accurate extraction of the cracks. Based on the extracted crack information, including the length, width, number and distribution of the cracks, a hydraulic fracturing effect evaluation index is established through a fuzzy comprehensive evaluation method to scientifically and accurately evaluate the hydraulic fracturing effect of the coal seam.

[0037] The image preprocessing refers to using the histeq function in Matlab to perform histogram equalization on the image to enhance the contrast and brightness of the image, and using the weighted average method to grayscale the color image after brightness enhancement. The calculation formula is: Gray = 0.299×R+0.587×G+0.114×B, where R, G, and B are the values ​​of the red, green, and blue channels of the color image, respectively, and Gray is the converted grayscale value.

[0038] The crack extraction refers to using Python to establish a convolutional neural network (CNN) model, using the model to automatically extract crack information from preprocessed images, establishing a data set by collecting a large number of downhole hydraulic fracturing crack images, preprocessing the acquired images using the histeq function in Matlab software, and dividing the preprocessed images in the data set into a training set, a validation set, and a test set. The training set is used to train the CNN model, and the cracks in the preprocessed images are manually labeled. During the training process, the validation set is used to monitor the performance of the model to prevent overfitting, and the parameters of the model are continuously adjusted with the help of the back-propagation algorithm to make the model's prediction results as close as possible to the labeled crack positions. Finally, the test set is used to evaluate the generalization ability of the model, that is, the model's recognition performance on new low-brightness coal seam crack images.

[0039] The hydraulic fracturing effect evaluation refers to establishing a comprehensive evaluation index system and setting the weight of each factor using a fuzzy comprehensive evaluation method. The evaluation process of this method is to determine the evaluation index and level, construct a fuzzy membership function, construct a fuzzy relationship matrix, determine the weight vector, calculate the fuzzy comprehensive evaluation result, and defuzzify. The weight value is determined based on the defuzzification result. The weight of each factor is set according to the fuzzy comprehensive evaluation result. According to the fuzzy comprehensive evaluation result, the weight of fracture width is set to 0.3. Wider fractures can provide better gas diversion channels and play a key role in the rapid release of coalbed methane. The weight of fracture length is 0.25, which is of great significance in expanding the range of gas migration. The weight of fracture number is also 0.15. A sufficient number of fractures can increase the gas diversion path. The calculated weight of fracture density is 0.1. In terms of fracture distribution, based on the requirement of uniform gas flow in the coal seam, uniformly distributed fractures are conducive to improving the overall diversion effect. After analysis and calculation, the weight of fracture density is set to 0.1. The better the fracture connectivity, the more conducive it is to gas diffusion in the coal seam. Based on theory and calculation, the weight of fracture density is set to 0.1.

[0040] Assume that the crack length score is L (calculated based on the ratio of actual length to expected length, such as when the actual length reaches 80% of the expected length, L = 0.8), the crack width score is W (also calculated based on the ratio of actual width to expected width), and the crack number score is N (calculated based on the ratio of actual number to expected number).

[0041] The crack density score is calculated by first setting an ideal crack density value Dideal and the actual crack density Dactual. The score is calculated using the formula SD = Dactual / Dideal. If SD ≥ 1, it means that the actual crack density meets or exceeds expectations and the score is 1. If SD < 1, the score is based on the actual proportion.

[0042] The crack distribution score extracts the coordinates of the crack center points and uses the K-Means clustering algorithm to group the cracks into clusters. Each cluster represents an area where cracks are concentrated. The crack distribution is described by calculating the number of clusters, the number of cracks within each cluster, and the location of the cluster center.

[0043] The crack connectivity score is obtained by processing the crack image using the cv2.connectedComponents function of OpenCV in Python, marking different connected regions, and then calculating the number and size of connected regions. The connected crack ratio can be defined as the ratio of the length or area of ​​connected cracks to the total crack length or area.

[0044] Comprehensive score S = L × length weight + W × width weight + N × number weight + D × density weight + U × distribution weight + C × connectivity weight.

[0045] Furthermore, by setting dynamic benchmark values ​​(a, b, c, d), the comprehensive score results are compared with the dynamic benchmark values ​​to judge the hydraulic fracturing effect. The physical meaning and determination method of the dynamic benchmark values ​​are as follows:

[0046] a: critical value of poor air permeability, corresponding to the lowest acceptable threshold of the comprehensive score;

[0047] b: critical value of medium air permeability;

[0048] c: critical value for good air permeability;

[0049] d: Critical value for excellent air permeability.

[0050] The method for determining the dynamic benchmark value is to measure the permeability coefficient of the coal seam after hydraulic fracturing with the radial flow method based on the coal industry standard MT / T 1173-2019 of the People's Republic of China, determine the permeability level based on the size of the permeability coefficient, divide the grades (poor, medium, good, excellent) according to the permeability test results, and calculate the comprehensive score of the corresponding distribution range. By testing the hydraulic fracturing permeability coefficient of the coal seam multiple times, the comprehensive score S corresponding to different permeability levels is calculated respectively, and the values ​​of a, b, c, and d are determined by the average of the comprehensive scores under a certain permeability.

[0051] The permeability of coal can be measured by the permeability coefficient, which is divided into the following categories:

[0052] Excellent air permeability: air permeability coefficient is greater than 10m2 / (mbar-day);

[0053] Good air permeability: air permeability coefficient is 5-10 m2 / (mbar day);

[0054] Medium air permeability: air permeability coefficient is 1 to 5 square meters / (mbar day);

[0055] Poor air permeability: the air permeability coefficient is 0.1 to 1 square meters / (mbar day);

[0056] Very poor air permeability: air permeability coefficient is less than 0.1 m2 / (mbar day).

[0057] Based on the calculated comprehensive scores, the classification standards for hydraulic fracturing permeability enhancement effects are given, as shown in Table 1:

[0058] Table 1 Classification standard of hydraulic fracturing permeability enhancement effect

[0059]

Claims

1. A method for evaluating the hydraulic fracturing effect of coal seams under low-light background based on AI image recognition refers to the use of high-speed camera image acquisition equipment to record the initiation and expansion images of cracks in the coal seam during the hydraulic fracturing process, and preprocessing operations on the acquired images with the help of Matlab software. To address the low-light background problem, the overall brightness of the image is enhanced by the histeq function in Matlab, making the crack information in the image clearer and more visible. The image after brightness enhancement is further grayscaled to convert the color image into a grayscale image. With the help of convolutional neural network (CNN) data set training, the cracks in the image are effectively separated from the background, and the cracks are accurately extracted. Based on the extracted crack information, including the length, width, number and distribution of the cracks, a hydraulic fracturing effect evaluation index is established through the fuzzy comprehensive evaluation method based on Darcy's law, fracture mechanics and multiphase flow theory, so as to scientifically and accurately evaluate the hydraulic fracturing effect of coal seams.

2. The image preprocessing as described in claim 1 refers to using the histeq function in Matlab to perform histogram equalization on the image to enhance the contrast and brightness of the image, and using the weighted average method to grayscale the color image after brightness enhancement. The calculation formula is: Gray = 0.299×R+0.587×G+0.114×B, where R, G, and B are the values ​​of the red, green, and blue channels of the color image, respectively, and Gray is the grayscale value after conversion.

3. The crack extraction method according to claim 1 refers to using Python to establish a convolutional neural network (CNN) model, using the model to automatically extract crack information from the preprocessed image, and establishing a data set by collecting a large number of downhole hydraulic fracturing crack images. After preprocessing the acquired images using the histeq function in Matlab software, the preprocessed images in the data set are divided into a training set, a validation set, and a test set. The CNN model is trained using the training set, and the cracks in the preprocessed images are manually labeled. During the training process, the validation set is used to monitor the performance of the model to prevent overfitting, and the parameters of the model are continuously adjusted with the help of the back propagation algorithm to make the model's prediction results as close as possible to the labeled crack positions. Finally, the test set is used to evaluate the generalization ability of the model, that is, the model's recognition performance on new low-brightness coal seam crack images.

4. The hydraulic fracturing effect evaluation as claimed in claim 1 refers to establishing a comprehensive evaluation index system, giving the calculation method of crack length, width, number, density, distribution, and connectivity scores respectively, and using the fuzzy comprehensive evaluation method to set the weight of each factor. The evaluation process of this method is to determine the evaluation index and grade, construct a fuzzy membership function, construct a fuzzy relationship matrix, determine the weight vector, calculate the fuzzy comprehensive evaluation result, defuzzify, determine the weight value based on the defuzzification result, and set the weight of each factor according to the fuzzy comprehensive evaluation result. According to the fuzzy evaluation result, the weight of the crack width is set to 0.3, which is relatively high. Wider cracks can provide better gas conduction channels and play a key role in the rapid release of coalbed methane. The weight of crack length is 0.25, which is of great significance in expanding the range of gas migration. The weight of crack number is also 0.

15. A sufficient number of cracks can increase the gas conduction path. The calculation of crack density shows that its weight is 0.

1. In terms of crack distribution, based on the requirement of uniform gas flow in coal seams, evenly distributed cracks are conducive to improving the overall conduction effect. After analysis and calculation, its weight is set to 0.

1. The better the crack connectivity, the more conducive it is to gas diffusion in the coal seam. Based on theory and calculation, its weight is set to 0.

1.

5. The crack length score of claim 4 is L (calculated based on the ratio of the actual length to the expected length, e.g., if the actual length reaches 80% of the expected length, L = 0.8), the crack width score is W (also calculated based on the ratio of the actual width to the expected width), and the crack number score is N (calculated based on the ratio of the actual number to the expected number). The crack density score is calculated by first setting an ideal crack density value D ideal , the actual crack density is D actual , using formula S D =D actual / D ideal Calculate the score, if S D ≥1, indicating that the actual crack density reaches or exceeds the expected value, with a score of 1. D If the value is less than 1, the score is calculated according to the actual proportion. The crack distribution score is obtained by extracting the coordinates of the crack center points and using the K-Means clustering algorithm to divide the cracks into different clusters. Each cluster represents an area where cracks are concentrated. The distribution of cracks is described by calculating the number of clusters, the number of cracks in each cluster, the center position of the cluster, and other information. The crack connectivity score is obtained by processing the crack image using the cv2.connectedComponents function of OpenCV in Python, marking different connected areas, and then calculating the number and size of connected areas. The connected crack ratio can be defined as the ratio of the length or area of ​​connected cracks to the total crack length or area.

6. The comprehensive score S according to claim 4 or 5 = L × length weight + W × width weight + N × number weight + D × density weight + U × distribution weight + C × connectivity weight. Furthermore, by setting dynamic reference values ​​(a, b, c, d), the comprehensive score result is compared with the dynamic reference values ​​to determine the hydraulic fracturing effect. The physical meaning and determination method of the dynamic reference values ​​are as follows: a: critical value of poor air permeability, corresponding to the lowest acceptable threshold of the comprehensive score; b: critical value of medium air permeability; c: critical value for good air permeability; d: Critical value for excellent air permeability.

7. The method for determining the dynamic reference value according to claim 6 is as follows: based on the coal industry standard MT / T1173-2019 of the People's Republic of China, the permeability coefficient of the coal seam after hydraulic fracturing is measured by the radial flow method, the permeability level is determined by the size of the permeability coefficient, and the grades (poor, medium, good, excellent) are divided according to the permeability test results, and the comprehensive scores of the corresponding distribution ranges are calculated. By repeatedly testing the hydraulic fracturing permeability coefficients of different coal seams, the comprehensive scores S corresponding to different permeability levels are calculated respectively, and the values ​​of a, b, c, and d are determined by the average of the comprehensive scores under a certain permeability. The standard of the good or bad permeability of coal can be measured by the permeability coefficient, and the specific classification is as follows: Excellent air permeability: air permeability coefficient is greater than 10m2 / (mbar-day); Good air permeability: air permeability coefficient is 5-10 m2 / (mbar day); Medium air permeability: air permeability coefficient is 1 to 5 square meters / (mbar day); Poor air permeability: the air permeability coefficient is 0.1 to 1 square meters / (mbar day); Very poor air permeability: air permeability coefficient is less than 0.1 m2 / (mbar day).

8. The calculated comprehensive score according to claim 6 or 7 is used to provide a grading standard for hydraulic fracturing permeability enhancement effects, as shown in Table 1: Table 1 Classification standard of hydraulic fracturing permeability enhancement effect

Images