Quantitative grading evaluation method and system for development potential of coal bed gas reservoir area

Abstract

The invention relates to the technical field of geological analysis, in particular to a coal bed gas reservoir area development potential quantitative grading evaluation method and system, and the method comprises the steps: abstracting each coal bed gas reservoir area into each mathematical sample through geological and engineering parameters, and carrying out the unsupervised clustering analysis through an improved adaptive fuzzy C-means clustering algorithm, obtaining a plurality of clustering and grading results of the mathematical sample; determining the development potential grade of each coal bed gas reservoir area based on the plurality of clustering grading results; and constructing an end-to-end coal bed gas reservoir regional development potential quantification model. According to the invention, unsupervised clustering analysis is carried out through the improved adaptive fuzzy C-means clustering algorithm, the problem that the clustering result is unstable due to the influence of algorithm sample outliers is overcome, and the GAN is introduced to overcome the defect that the adaptive fuzzy C-means clustering algorithm is easy to fall into a local optimal solution. Quantitative grading evaluation of the coal bed gas reservoir area development potential is achieved, and the grading evaluation accuracy and efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of geological analysis technology, specifically to a method and system for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs. Background Technology

[0002] The evaluation of the development potential of coalbed methane reservoirs is mainly based on qualitative analysis of geological parameters and reservoir characteristics, such as macroscopic judgments using static indicators like reservoir thickness, permeability, porosity, and gas content. Currently, quantitative evaluation methods for development potential include the analytic hierarchy process (AHP) and fuzzy comprehensive evaluation, but these rely heavily on human subjectivity and are difficult to implement accurately and objectively. For example, in the evaluation of low-rank coalbed methane, the weighting of parameters such as gas content, permeability, and burial depth lacks scientific basis, leading to significant differences in results from different studies.

[0003] Therefore, existing technologies require each coalbed methane reservoir area to be graded and evaluated individually, which is not only inefficient but also makes it difficult to achieve objective and accurate grading and evaluation results due to human subjectivity. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for quantitatively classifying and evaluating the development potential of coalbed methane reservoir areas, so as to solve the technical problems of the existing technology, which classifies and evaluates each coalbed methane reservoir area one by one, resulting in low evaluation efficiency and difficulty in achieving objective and accurate classification and evaluation results due to human subjectivity.

[0005] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution: A method for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs includes the following steps: Select geological and engineering parameters for quantifying the regional development potential of coalbed methane reservoirs; Each coalbed methane reservoir region is abstracted into a mathematical sample based on the geological and engineering parameters, and an improved adaptive fuzzy C-means clustering algorithm is used to perform unsupervised clustering analysis to obtain multiple clustering classification results for the mathematical samples. The development potential level of each coalbed methane reservoir area was determined based on multiple clustering and grading results. Based on the geological and engineering parameters and development potential levels of each coalbed methane reservoir area, an end-to-end quantitative model of the development potential of coalbed methane reservoir areas is constructed.

[0006] As a preferred embodiment of the present invention, the geological and engineering parameters include coal seam thickness, gas content, permeability, reservoir pressure, porosity, burial depth, and structural curvature.

[0007] As a preferred embodiment of the present invention, the mathematical sample corresponding to the coalbed methane reservoir region is: In the formula, This is a mathematical sample of the k-th coalbed methane reservoir region. These are identifiers for coal seam thickness, gas content, permeability, reservoir pressure, porosity, burial depth, and structural curvature, respectively, where n is the total number of coalbed methane reservoir regions.

[0008] As a preferred embodiment of the present invention, the method for unsupervised clustering analysis using the improved adaptive fuzzy C-means clustering algorithm includes: An adaptive fuzzy index related to the local density of samples is introduced into the basic fuzzy C-means clustering algorithm. And introduce sample weights related to outliers. The improved objective function is obtained. ,in: ; In the formula, Let be the adaptive fuzzy index for the k-th mathematical sample. This is the base value for the adaptive fuzzy index. For adjustment coefficients, Let k be the sample density of the k-th mathematical sample. The minimum value of the sample density for all mathematical samples. This represents the maximum sample density of all mathematical samples; ; In the formula, Let k be the sample weight of the k-th mathematical sample. The LOF value obtained by the Local Outlier Factor (LOF) algorithm for the kth mathematical sample; ; In the formula, For the improved objective function, Let be the membership degree of the k-th mathematical sample belonging to the i-th class. As the category center of the i-th class, is the regularization coefficient, and C is the total number of categories.

[0009] As a preferred embodiment of the present invention, the improved objective function The constraints are: , In the formula, Let C be the membership degree of the k-th mathematical sample belonging to the i-th class, and C be the total number of classes.

[0010] As a preferred embodiment of the present invention, the Lagrange method is used based on an improved objective function. Determine the membership degree based on the constraints. Category Center ,in: ; ; In the formula, Let be the membership degree of the k-th mathematical sample belonging to the i-th class. As the category center of the i-th class, As the category center of class j, Let k be the sample weight of the k-th mathematical sample. Let n be the adaptive fuzzy index of the k-th mathematical sample, and n be the total number of mathematical samples.

[0011] As a preferred embodiment of the present invention, for C category centers Introducing C category centers generated based on GAN network The C category centers are corrected. The correction function is: ; In the formula, As the category center of the i-th class, For the category center of the i-th class generated based on the GAN network, For the improved objective function, The identifier for the category center.

[0012] As a preferred embodiment of the present invention, the structural expression of the GAN network is as follows: ; In the formula, , Let C be the class centers generated by the GAN network, z be random noise, and G be the generator identifier. Let C be the category center of the i-th class generated by the GAN network, where C is the total number of classes and n is the total number of mathematical samples. The training loss function of the GAN network is: ; ; ; In the formula, For the total loss, For generator loss, For discriminator loss, Here are the hyperparameters: D is the generator identifier, and E is the expectation identifier. According to Calculated , The objective function is to be improved.

[0013] As a preferred embodiment of the present invention, the method for constructing the quantitative model of regional development potential of coalbed methane reservoirs includes: The geological and engineering parameters are used as input terms of the convolutional neural network, and the development potential level is used as the output term of the convolutional neural network. The improved adaptive fuzzy C-means clustering algorithm is used to construct a dataset of each coalbed methane reservoir region, and a convolutional neural network is trained to obtain a quantitative model of the development potential of the coalbed methane reservoir region. The structural expression of the quantitative model for the regional development potential of coalbed methane reservoirs is as follows: ; In the formula, To develop potential level, These are identifiers for coal seam thickness, gas content, permeability, reservoir pressure, porosity, burial depth, and structural curvature, respectively. It is a convolutional neural network.

[0014] As a preferred embodiment of the present invention, the present invention provides a quantitative grading and evaluation system for the regional development potential of coalbed methane reservoirs, which is applied to the aforementioned method for quantitative grading and evaluation of the regional development potential of coalbed methane reservoirs. The system includes:

[0015] The feature processing unit is used to select geological and engineering parameters for quantifying the regional development potential of coalbed methane reservoirs; The clustering analysis unit is used to abstract each coalbed methane reservoir area into mathematical samples through the geological and engineering parameters, and to perform unsupervised clustering analysis using an improved adaptive fuzzy C-means clustering algorithm to obtain multiple clustering classification results of the mathematical samples. The graded quantification unit is used to determine the development potential grade of each coalbed methane reservoir area based on multiple clustering and grading results. The model building unit is used to construct an end-to-end quantitative model of the development potential of coalbed methane reservoir areas based on the geological and engineering parameters and development potential levels of each coalbed methane reservoir area.

[0016] Compared with the prior art, the present invention has the following advantages: This invention uses an improved adaptive fuzzy C-means clustering algorithm for unsupervised clustering analysis, overcoming the instability of clustering results due to outliers in the algorithm samples. Furthermore, it introduces a GAN network to overcome the defect of the adaptive fuzzy C-means clustering algorithm being prone to getting trapped in local optima. This enables quantitative and graded evaluation of the development potential of coalbed methane reservoirs, improving the accuracy and efficiency of the evaluation. It can effectively guide the deployment or densification of irregular well networks in the "sweet spot" areas of coalbed methane reservoirs. Attached Figure Description

[0017] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0018] Figure 1 A flowchart of a method for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs provided in an embodiment of the present invention; Figure 2 A block diagram of a quantitative grading and evaluation system for the regional development potential of coalbed methane reservoirs provided in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, this invention provides a method for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs, comprising the following steps: Select geological and engineering parameters for quantifying the regional development potential of coalbed methane reservoirs; Each coalbed methane reservoir region is abstracted into mathematical samples through geological and engineering parameters, and an improved adaptive fuzzy C-means clustering algorithm is used to perform unsupervised clustering analysis to obtain multiple clustering and hierarchical results of the mathematical samples. The development potential level of each coalbed methane reservoir area was determined based on multiple clustering and grading results. Based on the geological and engineering parameters and development potential levels of each coalbed methane reservoir area, an end-to-end quantitative model of the development potential of coalbed methane reservoir areas is constructed.

[0021] This invention utilizes the fuzzy C-means clustering algorithm framework to perform unsupervised clustering analysis on various coalbed methane reservoir areas. This enables the classification of areas belonging to the same development potential level into the same cluster, thereby directly performing a unified quantitative assessment of all coalbed methane reservoir areas without the need to quantify the development potential of each coalbed methane reservoir area individually, achieving efficient quantification of development potential.

[0022] This invention introduces sample weights to control outliers within the framework of the basic (or traditional) fuzzy C-means clustering algorithm. This overcomes the instability of clustering results caused by the influence of outliers. The Local Outlier Factor (LOF) algorithm is used to identify outliers and assign weights to them. Specifically, the more severe the outlier, the closer the weight value is to 0, thus reducing its influence on the cluster centers (category centers). Non-outliers are assigned a weight of 1.

[0023] This invention also introduces an adaptive fuzzy index, which can dynamically adjust the fuzzy index of the basic (or traditional) fuzzy C-means clustering algorithm according to the sample distribution density, in areas of dense sample distribution. Reduced clustering (more rigid clusters, clearer boundaries); sparse sample distribution regions. Increase the size (making clustering "softer" and allowing fuzzy membership) to improve clustering adaptability.

[0024] This invention improves the objective function in the basic (or traditional) fuzzy C-means clustering algorithm framework by controlling the sample weights of outliers and introducing an adaptive fuzzy exponent, resulting in an improved objective function that can achieve both clustering accuracy and adaptability.

[0025] This invention also introduces a GAN network to correct the determination of the category centers in the basic (or traditional) fuzzy C-means clustering algorithm framework. In other words, it provides a way to correct the category centers obtained by traditional numerical calculation based on the Lagrange method by generating category centers through a GAN network, thereby avoiding the fuzzy C-means clustering algorithm framework from getting stuck in local optima, improving the ability to escape local optima, and correspondingly enhancing the global search capability.

[0026] Geological and engineering parameters, including coal seam thickness, gas content, permeability, reservoir pressure, porosity, burial depth, and structural curvature, can be modified as needed.

[0027] The mathematical sample corresponding to the coalbed methane reservoir area is In the formula, This is a mathematical sample of the k-th coalbed methane reservoir region. These are identifiers for coal seam thickness, gas content, permeability, reservoir pressure, porosity, burial depth, and structural curvature, respectively, where n is the total number of coalbed methane reservoir regions.

[0028] Methods for unsupervised clustering analysis using the improved adaptive fuzzy C-means clustering algorithm include: An adaptive fuzzy index related to the local density of samples is introduced into the basic fuzzy C-means clustering algorithm. And introduce sample weights related to outliers. The improved objective function is obtained. ,in: ; In the formula, Let be the adaptive fuzzy index for the k-th mathematical sample. This is the base value for the adaptive fuzzy index. For adjustment coefficients, The sample density of the k-th mathematical sample (for each mathematical sample) The average distance of its neighborhood (e.g., the 5 nearest samples) is used as the statistical measure. , The smaller the value, the higher the density. The minimum value of the sample density for all mathematical samples. This represents the maximum sample density of all mathematical samples; in =1.5 (base value), a=0.5 (adjustment coefficient), the higher the density ( Small), The smaller the size (the more "rigid" the clustering); the lower the density ( (large) then Increase (allow fuzzy classification).

[0029] This invention also introduces an adaptive fuzzy index, which can dynamically adjust the fuzzy index of the basic (or traditional) fuzzy C-means clustering algorithm according to the sample distribution density, in areas of dense sample distribution. Reduced clustering (more rigid clusters, clearer boundaries); sparse sample distribution regions. Increase the size (making clustering "softer" and allowing fuzzy membership) to improve clustering adaptability.

[0030] ; In the formula, Let k be the sample weight of the k-th mathematical sample. The LOF value obtained by the Local Outlier Factor (LOF) algorithm for the kth mathematical sample; The LOF (Local Outlier Factor) algorithm is a density-based unsupervised outlier detection method that identifies outliers by comparing the density difference between a data point and its neighborhood. Its core steps include: 1. Calculating local density: Using the data point as the center, select k nearest neighbors (typically k=30-50) and calculate the local density based on the number of neighboring points or their distance. 2. Calculating reachability distance: Adjust the actual distance based on the local density (considering the influence of neighborhood density). 3. Calculating the LOF value: Compare the ratio of the local density of the target point to that of its neighbors; a higher ratio indicates a higher likelihood of an outlier. This invention introduces sample weights to control outliers within the framework of the basic (or traditional) fuzzy C-means clustering algorithm. This overcomes the instability of clustering results caused by the influence of outliers. The Local Outlier Factor (LOF) algorithm is used to identify outliers and assign weights to them. Specifically, the more severe the outlier, the closer the weight value is to 0, thus reducing its influence on the cluster centers (category centers). Non-outliers are assigned a weight of 1.

[0031] This invention improves the objective function of the basic (or traditional) fuzzy C-means clustering algorithm by controlling the sample weights of outliers and introducing an adaptive fuzzy exponent. The improved objective function achieves both clustering accuracy and adaptability, as follows: ; In the formula, For the improved objective function, Let be the membership degree of the k-th mathematical sample belonging to the i-th class. As the category center of the i-th class, is the regularization coefficient, and C is the total number of categories.

[0032] Improved objective function The constraints are: , In the formula, Let C be the membership degree of the k-th mathematical sample belonging to the i-th class, and C be the total number of classes.

[0033] Using the Lagrange method based on an improved objective function Determine the membership degree based on the constraints. Category Center ,in: ; ; In the formula, Let be the membership degree of the k-th mathematical sample belonging to the i-th class. As the category center of the i-th class, As the category center of the j-th class, Let k be the sample weight of the k-th mathematical sample. Let n be the adaptive fuzzy index of the k-th mathematical sample, and n be the total number of mathematical samples.

[0034] This invention utilizes the Lagrange method to solve the improved objective function. Construct the Lagrange function and introduce the Lagrange multiplier λ. k (Each data sample corresponds to one): ; For the Lagrange function with respect to Taking the partial derivative and setting it to 0 (extremum condition), we obtain... ; Organized .

[0035] For the Lagrange function with respect to Find the partial derivative and set it to 0, then we get... ; Organized ; By constraints , Substitute Finally obtained .

[0036] This invention utilizes the Lagrange method to improve the objective function. The membership degree was calculated using numerical methods. Category Center The update process may encounter local optima, especially when the data distribution is complex or noisy, making it prone to getting stuck in local extrema. This algorithm determines cluster centers by iteratively optimizing the objective function (usually minimizing the sum of weighted distances), but when the dataset contains multiple regions of varying densities or noisy points, the algorithm may prematurely converge to a local optimum rather than the global optimum. Therefore, this invention introduces a GAN-based method for generating cluster centers, combined with the Lagrange multiplier method, to correct and adjust the cluster centers determined by the Lagrange multiplier method, overcoming the limitations of local extrema, avoiding local optima problems, and improving global search capabilities.

[0037] For C category centers Introducing C category centers generated based on GAN network Make corrections, C category centers The correction function is: ; In the formula, As the category center of the i-th class, For the category center of the i-th class generated based on the GAN network, For the improved objective function, The identifier for the category center.

[0038] When the cluster centers generated by the GAN network are better than those determined by the Lagrange method, this invention will replace the cluster centers determined by the Lagrange method with the cluster centers generated by the GAN network. If the cluster centers generated by the GAN network are not better than those determined by the Lagrange method, the cluster centers determined by the Lagrange method will continue to be used. This invention achieves the goal of introducing optimization and adjustment of cluster centers through a cluster center generation method, thereby escaping local optima.

[0039] The structural expression of a GAN network is: ; In the formula, , Let C be the class centers generated by the GAN network, z be random noise, and G be the generator identifier. Let C be the category center of the i-th class generated by the GAN network, where C is the total number of classes and n is the total number of mathematical samples. The training loss function for GAN networks is: ; ; ; In the formula, For the total loss, For generator loss, For discriminator loss, Here are the hyperparameters (set as needed), D is the generator identifier, and E is the expectation identifier. According to Calculated , The objective function is to be improved.

[0040] To ensure the accuracy of cluster center generation by the GAN network, this invention employs adversarial training between the generator G and the discriminator D to simulate data distribution, thereby generating suitable cluster centers based on the subject distribution. This indicates that the discriminator generates the category centers. The discriminant output, This indicates that the discriminator correctly identifies the true cluster centers. The discriminant output is a probability value representing the probability that the discriminator considers the input to be a true cluster center.

[0041] The generator aims to produce samples that can fool the discriminator, that is, it hopes... Approaching 1, thus This would result in a very large negative number (significant loss), so the generator needs to minimize this loss, and also incorporates... This is the objective function, which is to determine the generated category centers. The objective function of fuzzy clustering can be made To minimize the size, constraints are imposed through the objective function to avoid the generator's random and ineffective generation.

[0042] middle The goal is to enable the discriminator to accurately judge real data. Discriminant probability As close to 1 as possible The goal is to enable the discriminator to analyze the generated data. Discriminant probability Get as close to 0 as possible, so that the discriminator will judge the generated data as false.

[0043] This invention utilizes adversarial training between the generator and the discriminator to generate cluster centers that more closely resemble the distribution of real data.

[0044] In summary, the generator learns to generate cluster centers that can simultaneously deceive the discriminator (i.e., appear to be the real data distribution) and serve as effective cluster centers (i.e., minimize the clustering objective function).

[0045] Furthermore, this invention utilizes convolutional neural networks to construct an end-to-end development potential quantification model, directly predicting the development potential level based on geological and engineering parameters without the aforementioned numerical calculation process, thus simplifying the quantification process. Moreover, by training on large-scale data, it learns implicit feature combinations for complex scenarios. Faced with changes in data distribution, the end-to-end model can automatically adjust its feature learning method through retraining to adapt to the characteristics of new data. In contrast, manual features in step-by-step systems may fail in new scenarios and require redesign, making the end-to-end model far less flexible.

[0046] Methods for constructing quantitative models of regional development potential of coalbed methane reservoirs include: Geological and engineering parameters are used as input terms for the convolutional neural network, and development potential level is used as the output term for the convolutional neural network. The improved adaptive fuzzy C-means clustering algorithm was used to construct a dataset of each coalbed methane reservoir region, and a convolutional neural network was trained to obtain a quantitative model of the development potential of the coalbed methane reservoir region. The structural expression of the quantitative model for regional development potential of coalbed methane reservoirs is as follows: ; In the formula, To develop potential level, These are identifiers for coal seam thickness, gas content, permeability, reservoir pressure, porosity, burial depth, and structural curvature, respectively. It is a convolutional neural network.

[0047] like Figure 2 As shown, the present invention provides a quantitative grading and evaluation system for the regional development potential of coalbed methane reservoirs, which is applied to a method for quantitative grading and evaluation of the regional development potential of coalbed methane reservoirs. The system includes:

[0048] The feature processing unit is used to select geological and engineering parameters for quantifying the regional development potential of coalbed methane reservoirs; The clustering analysis unit is used to abstract each coalbed methane reservoir area into mathematical samples through geological and engineering parameters, and to perform unsupervised clustering analysis using an improved adaptive fuzzy C-means clustering algorithm to obtain multiple clustering hierarchical results of the mathematical samples. The graded quantification unit is used to determine the development potential grade of each coalbed methane reservoir area based on multiple clustering and grading results. The model building unit is used to construct an end-to-end quantitative model of the development potential of coalbed methane reservoir areas based on the geological and engineering parameters and development potential levels of each coalbed methane reservoir area.

[0049] This invention uses an improved adaptive fuzzy C-means clustering algorithm for unsupervised clustering analysis, overcoming the instability of clustering results due to outliers in the algorithm samples. Furthermore, it introduces a GAN network to overcome the defect of the adaptive fuzzy C-means clustering algorithm being prone to getting trapped in local optima. This enables quantitative and graded evaluation of the development potential of coalbed methane reservoirs, improving the accuracy and efficiency of the evaluation. It can effectively guide the deployment or densification of irregular well networks in the "sweet spot" areas of coalbed methane reservoirs.

[0050] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. A method for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs, characterized in that, Includes the following steps: Select geological and engineering parameters for quantifying the regional development potential of coalbed methane reservoirs; Each coalbed methane reservoir region is abstracted into a mathematical sample based on the geological and engineering parameters, and an improved adaptive fuzzy C-means clustering algorithm is used to perform unsupervised clustering analysis to obtain multiple clustering classification results for the mathematical samples. The development potential level of each coalbed methane reservoir area was determined based on multiple clustering and grading results. Based on the geological and engineering parameters and development potential levels of each coalbed methane reservoir area, an end-to-end quantitative model of the development potential of coalbed methane reservoir areas is constructed.

2. The method for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs according to claim 1, characterized in that: The geological and engineering parameters include coal seam thickness, gas content, permeability, reservoir pressure, porosity, burial depth, and structural curvature.

3. The method for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs according to claim 2, characterized in that: The mathematical sample corresponding to the coalbed methane reservoir region is: In the formula, This is a mathematical sample of the k-th coalbed methane reservoir region. These are identifiers for coal seam thickness, gas content, permeability, reservoir pressure, porosity, burial depth, and structural curvature, respectively, where n is the total number of coalbed methane reservoir regions.

4. The method for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs according to claim 3, characterized in that: Methods for unsupervised clustering analysis using the improved adaptive fuzzy C-means clustering algorithm include: An adaptive fuzzy index related to the local density of samples is introduced into the basic fuzzy C-means clustering algorithm. And introduce sample weights related to outliers. The improved objective function is obtained. ,in: ； In the formula, Let be the adaptive fuzzy index for the k-th mathematical sample. This is the base value for the adaptive fuzzy index. For adjustment coefficients, Let k be the sample density of the k-th mathematical sample. The minimum value of the sample density for all mathematical samples. This represents the maximum sample density of all mathematical samples; ； In the formula, Let k be the sample weight of the k-th mathematical sample. The LOF value obtained by the Local Outlier Factor (LOF) algorithm for the kth mathematical sample; ； In the formula, For the improved objective function, Let be the membership degree of the k-th mathematical sample belonging to the i-th class. As the category center of the i-th class, is the regularization coefficient, and C is the total number of categories.

5. The quantitative grading and evaluation system for regional development potential of coalbed methane reservoirs according to claim 4, characterized in that: The improved objective function The constraints are: , In the formula, Let C be the membership degree of the k-th mathematical sample belonging to the i-th class, and C be the total number of classes.

6. The method for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs according to claim 5, characterized in that: Using the Lagrange method based on an improved objective function Determine the membership degree based on the constraints. Category Center ,in: ； ； In the formula, Let be the membership degree of the k-th mathematical sample belonging to the i-th class. As the category center of the i-th class, As the category center of the j-th class, Let k be the sample weight of the k-th mathematical sample. Let n be the adaptive fuzzy index of the k-th mathematical sample, and n be the total number of mathematical samples.

7. The method for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs according to claim 6, characterized in that: For C category centers Introducing C category centers generated based on GAN network The correction is made to the C category centers. The correction function is: ； In the formula, As the category center of the i-th class, For the category center of the i-th class generated by the GAN network, For the improved objective function, The identifier for the category center.

8. The method for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs according to claim 7, characterized in that: The structural expression of the GAN network is as follows: ； In the formula, , Let C be the class centers generated by the GAN network, z be random noise, and G be the generator identifier. Let C be the category center of the i-th class generated by the GAN network, where C is the total number of classes and n is the total number of mathematical samples. The training loss function of the GAN network is: ； ； ； In the formula, For the total loss, For generator loss, For discriminator loss, Here are the hyperparameters: D is the generator identifier, and E is the expectation identifier. According to Calculated , The objective function is to be improved.

9. The method for quantitatively classifying and evaluating the regional development potential of coalbed methane reservoirs according to claim 8, characterized in that: The method for constructing the quantitative model of regional development potential of coalbed methane reservoirs includes: The geological and engineering parameters are used as input terms of the convolutional neural network, and the development potential level is used as the output term of the convolutional neural network. The improved adaptive fuzzy C-means clustering algorithm is used to construct a dataset of each coalbed methane reservoir region, and a convolutional neural network is trained to obtain a quantitative model of the development potential of the coalbed methane reservoir region. The structural expression of the quantitative model for the regional development potential of coalbed methane reservoirs is as follows: ； In the formula, To develop potential level, These are identifiers for coal seam thickness, gas content, permeability, reservoir pressure, porosity, burial depth, and structural curvature, respectively. It is a convolutional neural network.

10. A quantitative grading and evaluation system for the regional development potential of coalbed methane reservoirs, characterized in that, The system, which is applied to the quantitative grading and evaluation method for regional development potential of coalbed methane reservoirs as described in any one of claims 1-9, comprises: The feature processing unit is used to select geological and engineering parameters for quantifying the regional development potential of coalbed methane reservoirs; The clustering analysis unit is used to abstract each coalbed methane reservoir area into mathematical samples through the geological and engineering parameters, and to perform unsupervised clustering analysis using an improved adaptive fuzzy C-means clustering algorithm to obtain multiple clustering classification results of the mathematical samples. The graded quantification unit is used to determine the development potential grade of each coalbed methane reservoir area based on multiple clustering and grading results. The model building unit is used to construct an end-to-end quantitative model of the development potential of coalbed methane reservoir areas based on the geological and engineering parameters and development potential levels of each coalbed methane reservoir area.

Images