Complex coal body structure horizontal well crack parameter matching relation optimization design method

Abstract

The invention discloses a complex coal body structure horizontal well crack parameter matching relation optimization design method. The method comprises the following steps that 1, a complex coal body structure coal seam fine numerical simulation model is established through a Petri-Re numerical simulator; secondly, main control factors of the coal bed gas fractured horizontal well development effect are screened through a numerical simulation method; according to the method, a plurality of test results of a coal rock core are combined, Petri-Re numerical simulation software is utilized to construct a numerical simulation model for characterizing the characteristics of the multi-layer stacked coal seam by considering coal seam gas adsorption-desorption-diffusion, Lagmuir isothermal adsorption, interlayer heterogeneity and the like. Meanwhile, a fractured horizontal well model is constructed by utilizing an artificial fracture, and main control factors influencing the productivity of the coal bed gas are screened by adopting a numerical simulation method. Meanwhile, a value simulator is called through Python to achieve self-generation of a value simulation sample, a yield prediction agent model is built with the help of an XGboost algorithm, and then the optimal solution of the coal bed gas fractured horizontal well fracture parameter combination is obtained.

Description

Technical Field

[0001] The invention relates to the field of coalbed gas pressure analysis, and in particular to a method for optimizing the design of fracture parameter matching relationships in a horizontal well with a complex coal body structure. Background Art

[0002] Coalbed methane resources are huge, with a total amount of about 80 trillion m3 of resources in different depths and regions. At present, my country is in the primary stage of coalbed methane development. Due to the characteristics of diverse structures, complex geology and low coal seam permeability in China, the average daily gas production of a single well is 994 m3, which is only 1 / 4 of that in the United States. In view of this feature, developing coal reservoir fracturing transformation technology suitable for low permeability conditions and optimizing fracturing fracture parameters are the inevitable ways to increase the production of coalbed methane single wells and development benefits.

[0003] Most of the existing coalbed methane reservoirs are developed using hydraulic fracturing technology, and the optimization of hydraulic fracture parameters often adopts numerical simulation. The current numerical simulation method is mainly based on the continuous medium model (dual medium, multiple medium and equivalent continuous medium model). The model considers Knudsen diffusion, molecular diffusion, and adsorbed gas analysis to predict the production capacity of coalbed methane fractured horizontal wells and optimize fracture parameters. The advantages of numerical simulation method are that it can be repeated, short time, low cost, and can simulate various heterogeneous conditions, but the numerical simulation grid construction and numerical calculation process considering the flow characteristics of fractured horizontal wells are complicated. In addition, numerical simulation technology always involves solving large-scale partial differential equations, which requires a lot of computing time, making the optimization of fracturing parameters in the entire area impractical.

[0004] At present, intelligent oil and gas exploration and development has become an industry hotspot and development trend. Using machine learning methods such as support vector machines and neural networks, the geological parameters and field production parameters of coalbed methane reservoirs are processed, and a capacity prediction model is established by analyzing the potential patterns in a large amount of data. Compared with traditional numerical simulation methods, machine learning methods not only shorten the modeling cycle and speed up the calculation, but also have better scalability. Therefore, on the basis of clarifying the main controlling factors affecting coalbed methane production capacity, establishing a coalbed methane reservoir capacity prediction model based on machine learning can further optimize fracturing parameters and improve the design level of fracturing technology. Thereby guiding the design of fracturing schemes and the production management of coalbed methane wells throughout their life cycle. . Summary of the invention

[0005] In order to solve the problems existing in the prior art, a method for optimizing the design of fracture parameter matching relationships in horizontal wells with complex coal structures is provided.

[0006] The present invention solves the above technical problems through the following technical solutions, and the present invention comprises the following steps:

[0007] Step 1: Establish a fine numerical simulation model of coal seams with complex coal structure through the Petrel-Re numerical simulator;

[0008] Step 2: Screen the main controlling factors of coalbed methane fracturing horizontal well development effect through numerical simulation method;

[0009] Step 3: Construct a self-generated model of numerical simulation samples;

[0010] Step 4: Establish a coalbed methane production capacity prediction agent model based on the XGBoost algorithm.

[0011] Furthermore, the specific process of step one is as follows:

[0012] Using Petrel-Re reservoir numerical simulation software, the grid system of the basic model is established;

[0013] Among them, the fluid model adopts the black oil model, considering Lagmuir isotherm adsorption and Fick diffusion;

[0014] The grid number of the conceptual model is 150×51×51, the grid step size in the X and Y directions is 10m, the grid step size in the Z direction is 0.5m, the model is 1500m long and 510m wide;

[0015] It is divided into 8 small layers in total, of which the second and sixth layers are coal seams, and the rest are rock and soil layers other than coal seams;

[0016] The input parameters of the model are determined based on the coal core experimental data and mine data. The gas-water two-phase is taken as the research object, the double-porosity and double-permeability model is selected, and the Intersect simulator is used for numerical simulation.

[0017] Furthermore, the specific process of step 2 is as follows:

[0018] First, analyze the factors affecting the productivity of coalbed methane fracturing horizontal wells:

[0019] Including matrix porosity, gas saturation, artificial fracture conductivity, artificial fracture half-length, fracture height, fracture spacing, and horizontal section length;

[0020] The data range design of the main control factors:

[0021] Through parameter sensitivity analysis, four key influencing factors were identified, including the half-length of artificial fractures, the distance between fractures, the conductivity of fractures and their height;

[0022] The full factorial experimental design method involves conducting at least one experiment on all factors at all levels to estimate the main effects of each factor and their interactions.

[0023] Furthermore, the specific process of step three is as follows:

[0024] Automatically generate numerical simulation files:

[0025] Based on the data range of the main control factors screened in step 2, a simulation scheme is generated by random combination. The ixf file of the intersect simulator is used to assemble keywords through Python definition functions to achieve automatic filling of crack length, crack height, crack conductivity and crack spacing, and the design scheme is assembled using the AFI file.

[0026] Afterwards, the numerical simulator is automatically called in batches and the results are extracted:

[0027] Based on Python, the Intersect environment in the computer is called to batch process the automatically generated AFI files, and the required numerical simulation result files are extracted and saved in a new folder for later extraction and combination of simulator results;

[0028] Based on the numerical simulation result file (PRT file), by marking the coalbed methane geological reserves of the first time node and the coalbed methane geological reserves of the last time node of each numerical simulation result file, a loop function is defined, and the cumulative gas production of different schemes is automatically extracted and formatted and saved in a CSV file, thereby realizing the full automation of simulator result extraction and storage.

[0029] Furthermore, the specific process of step four includes: prediction sample set division, prediction model establishment and model prediction effect evaluation.

[0030] Furthermore, the process of dividing the prediction sample set is as follows: the prediction sample set obtained by the self-generation model of the numerical simulation sample is divided into a training set and a validation set according to a relevant ratio of 80% and 20%;

[0031] The training set is mainly used to train the prediction model;

[0032] The validation set is mainly used to adjust hyperparameters, including the number of decision trees, learning rate, and maximum depth of the tree;

[0033] Each set contains the input and output data of the model, where the input data includes fracture height, fracture length, fracture conductivity and fracture spacing, and the output data includes the cumulative gas production of coalbed methane.

[0034] Furthermore, the process of establishing the prediction model is as follows: using the XGBoost algorithm for modeling, the implementation process of the XGBoost algorithm is as follows:

[0035] (1) Order i represents the i-th sample in the training set (X, y)

[0036] (2) Satisfy 1≤t≤N k ;

[0037] ①Construct cost function:

[0038]

[0039] Among them, g i and h i They represent the first-order derivative and second-order derivative of the loss function l at sample i respectively; y is the true label value of the sample; represents the XGBoost prediction model, x i is the input feature parameter of the decision tree model; r i Residual, N k is the number of decision tree models; T represents the number of leaves of the tree model, γ, λ represent the regularization coefficients, ω j Indicates the leaf node value.

[0040] ②Use the data set (X, r) to train a tree model f with a maximum depth of d t

[0041] ③Update the prediction value of the XGBoost model

[0042]

[0043] ④ Update the residual term r

[0044]

[0045] (3) Obtain the final prediction model

[0046]

[0047] Here, ε represents the compression factor or learning rate.

[0048] Furthermore, the process of building a model with XGBoost includes the following steps:

[0049] The first step is to perform feature classification by continuously adding trees, and gradually grow a complete tree by learning the new function f(x) for each newly generated tree. In this way, each added tree can fit the residual of the last prediction, continuously improve the classification model, and improve the accuracy of the prediction;

[0050] The second step is to predict the sample scores after training k trees. The samples are classified into leaf nodes of each tree according to their characteristics and the corresponding scores are obtained.

[0051] In the third step, you only need to sum up the scores calculated by each tree to get the final prediction value corresponding to the sample.

[0052] Furthermore, the specific process of evaluating the model prediction effect is as follows:

[0053] The prediction effect of the capacity forecasting model was evaluated and analyzed by using the determination coefficient (R2) and mean square error (MSE).

[0054]

[0055] In the formula, y i and are the actual value and the predicted value, is the average of the actual value, and N is the size of the actual value in the validation set. When R2 is larger and MSE is smaller, it means that the error between the predicted value and the actual value of the model is smaller, and the performance of the model is better.

[0056] Compared with the prior art, the present invention has the following advantages: the optimization design method for matching relationship of fracture parameters of horizontal wells with complex coal body structure solves the problem of high labor cost of manually modifying numerical simulation parameters for numerical calculation through the self-generation program of numerical simulation samples. Based on the self-generation of numerical simulation samples written, a large number of accurate sample sets for coalbed methane production capacity prediction can be quickly generated. In addition, the coalbed methane production capacity prediction agent model established can avoid the use of traditional optimization methods to only screen out local optimal solutions, reduce the time cost of simulation, and improve work efficiency, making the system more worthy of promotion and use. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a fine numerical simulation model diagram of a coal seam with a complex coal body structure according to the present invention;

[0058] Figure 2 It is a flow chart of establishing a self-generated model of numerical simulation samples of the present invention;

[0059] Figure 3 It is a schematic diagram of the fracture parameter optimization process based on the production capacity prediction agent model of the present invention;

[0060] Figure 4 This is a diagram showing the effect of the hydraulic fracture interval on the daily gas production and the cumulative gas production of the present invention;

[0061] Figure 5 This is a diagram showing the effect of the hydraulic fracture interval on the daily water production and the cumulative water production of the present invention;

[0062] Figure 6 This is a diagram showing the influence of the hydraulic fracture conductivity of the present invention on the daily gas production and the cumulative gas production;

[0063] Figure 7 This is a diagram showing the influence of the hydraulic fracture conductivity of the present invention on the daily water production and the cumulative water production;

[0064] Figure 8 This is a diagram showing the influence of hydraulic fracture height on cumulative gas production according to the present invention;

[0065] Fig. 9 This is a diagram showing the influence of hydraulic fracture height on cumulative gas production according to the present invention;

[0066] Fig.10 This is a graph comparing the production capacity of the optimization method of the present invention with that of the original solution. DETAILED DESCRIPTION

[0067] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given, but the protection scope of the present invention is not limited to the following embodiment.

[0068] like Figures 1 to 10 As shown, this embodiment provides a technical solution: a method for optimizing the design of fracture parameter matching relationship of a horizontal well with a complex coal structure, comprising the following steps:

[0069] The first step is to establish a fine numerical simulation model of coal seams with complex coal structure through the Petrel-Re numerical simulator

[0070] The grid system of the basic model was established using Petrel-Re reservoir numerical simulation software. The fluid model adopts the black oil model, considering Lagmuir isothermal adsorption and Fick diffusion. The number of grids in the conceptual model is 150×51×51, the grid step size in the X and Y directions is 10m, the grid step size in the Z direction is 0.5m, the model is 1500m long and 510m wide. It is divided into 8 small layers in total, of which the second and sixth layers are coal seams, and the rest are rock and soil layers other than coal seams. In the gas reservoir model, most of the well sections of the horizontal wells are mainly arranged in the lower coal seam, and the remaining well sections are located in the middle layer. The predicted production cycle is 15 years. The hydraulic fracture is set to expand along the direction of the maximum principal stress. The local grid encryption technology and the equivalent conductivity method are used to accurately simulate the artificial hydraulic fractures.

[0071] The input parameters of the model are determined by coal core experimental data, mine data, etc. The gas-water two-phase is taken as the research object, the double-porosity and double-permeability model is selected, and the Intersect simulator is used for numerical simulation.

[0072] The second step is to screen the main controlling factors of coalbed methane fracturing horizontal well development effect through numerical simulation method, and analyze the factors affecting the production capacity of coalbed methane fracturing horizontal well.

[0073] There are many factors that affect coalbed methane production, such as matrix porosity, gas saturation, artificial fracture conductivity, artificial fracture half-length, fracture height, fracture spacing, horizontal section length, etc. This invention starts from the factors directly related to coalbed methane production capacity and finds the main controlling factors that affect the coalbed methane development effect. Using numerical simulation methods, sensitivity schemes are set, and sensitivity studies of different parameters are conducted to obtain the main factors affecting coalbed methane production capacity.

[0074] Data range design of main control factors

[0075] Through detailed parameter sensitivity analysis, four key influencing factors were identified, including the half-length of artificial fractures, the distance between fractures, the conductivity of fractures and their height. The full factorial experimental design method involves the combination of all factors at all levels, and at least one experiment is conducted to estimate the main effect of each factor and its interaction. This design method can obtain a rich amount of information and accurately estimate the main influence of experimental factors and the interaction between factors. The value range of each influencing factor is based on the mine data and parameter sensitivity analysis in the study area. Specifically, the variation range of the half-length of the fracture is set to a range of 60m-150m for the half-length of the fracture, 10-110m for the distance between fractures, and 50mD·m-300mD·m for the conductivity of the fracture. There are two ways to design the fracture height. One is to penetrate two coal seams with a fracture height of 25.5m, and the other is to penetrate only the lower coal seam, and the fracture extends vertically to the floor of the upper coal seam. The designed fracture height is 20m. This methodology aims to deeply explore and quantify the impact of artificial fracture characteristics on coalbed methane production capacity to guide fracture design and optimization in practice.

[0076] The third step is to build a self-generated model for numerical simulation samples

[0077] 1. Automatic generation of numerical simulation files

[0078] Based on the data range of the main control factors screened in the second step, the simulation scheme is randomly combined. Using the ixf file of the intersect simulator, the keywords are assembled through Python definition functions to realize the automatic filling of crack length, crack height, crack conductivity and crack spacing, and the design scheme is assembled with AFI files.

[0079] 2. Automatic batch call and result extraction of numerical simulator calculations

[0080] Based on Python calling the Intersect environment in the computer, the automatically generated AFI files are batch processed, and the required numerical simulation result files (PRT files) are extracted and saved in a new folder for the extraction and combination of the simulator results in the later stage. Based on the numerical simulation result files (PRT files), by marking the coalbed methane geological reserves at the first time node and the coalbed methane geological reserves at the last time node of each numerical simulation result file, a loop function is defined to automatically extract and format the cumulative gas production of different schemes and save them in a CSV file, so as to realize the full automation of the simulator result extraction and preservation, so that the multi-threaded processing of a single model significantly improves the operation efficiency. This ensures the automation of the entire process from the deployment of design parameters in numerical simulation, the parallel calculation of multiple schemes to the extraction and preservation of simulation results, and realizes the rapid and accurate construction process of the database.

[0081] The fourth step is to establish a coalbed methane production capacity prediction agent model based on the XGBoost algorithm

[0082] 1. Prediction sample set division

[0083] The prediction sample set obtained by the self-generated model of numerical simulation samples is divided into a training set and a validation set according to the relevant ratio of 80% and 20%. The training set is mainly used to train the prediction model; the validation set is mainly used to adjust hyperparameters, such as the number of decision trees, learning rate, maximum depth of the tree, etc. Each set contains the input and output data of the model, where the input data includes fracture height, fracture length, fracture conductivity and fracture spacing, and the output data includes the cumulative gas production of coalbed methane.

[0084] 2. Prediction model establishment

[0085] XGBoost is a machine learning system that uses decision trees as base learners and has both linear scale solvers and classification and regression tree learning algorithms. The training process of XGBoost is to continuously add trees, and use new trees to fit the prediction residuals in each process of adding trees, thereby improving the prediction accuracy. The characteristics of the XGBoost tree model allow us to quantify the importance of each feature and perform feature selection. The basic implementation process of the XGBoost algorithm is as follows:

[0086] (1) Order i represents the i-th sample in the training set (X, y)

[0087] (2) Satisfy 1≤t≤N k

[0088] ①Construct cost function:

[0089]

[0090] Among them, g i and h i They represent the first-order derivative and second-order derivative of the loss function l at sample i respectively; y is the true label value of the sample; represents the XGBoost prediction model, x i is the input feature parameter of the decision tree model; r i Residual, N k is the number of decision tree models; T represents the number of leaves of the tree model, γ, λ represent the regularization coefficients, ω j Indicates the leaf node value.

[0091] ②Use the data set (X, r) to train a tree model with a maximum depth of d

[0092] ③Update the prediction value of the XGBoost model

[0093]

[0094] ④ Update the residual term r

[0095]

[0096] (3) Obtain the final prediction model

[0097]

[0098] Here, ε represents the compression factor or learning rate.

[0099] The process of the XGBoost prediction model is as follows:

[0100] The first step is to perform feature classification by continuously adding trees, and gradually grow a complete tree by learning the new function f(x) for each newly generated tree. In this way, each added tree can fit the residual of the last prediction, continuously improve the classification model, and improve the accuracy of the prediction;

[0101] The second step is to predict the sample scores after training k trees. The samples are classified into leaf nodes of each tree according to their characteristics and the corresponding scores are obtained.

[0102] In the third step, you only need to sum up the scores calculated by each tree to get the final prediction value corresponding to the sample.

[0103] 3. Model prediction effect evaluation

[0104] The prediction effect of the capacity forecasting model was evaluated and analyzed by using the determination coefficient (R2) and mean square error (MSE).

[0105]

[0106] In the formula, y i and are the actual value and the predicted value, is the average of the actual value, and N is the size of the actual value in the validation set. When R2 is larger and MSE is smaller, it means that the error between the predicted value and the actual value of the model is smaller, and the performance of the model is better.

[0107] 4. Model prediction effect evaluation

[0108] Output is based on the crack parameters, crack spacing, etc. corresponding to the optimal development plan.

[0109] Take a coal mine block in Liupanshui, Guizhou as an example. The coalbed methane field mainly uses fracturing horizontal wells for coalbed methane development. There are two target coal seams, namely 10 coal and 12 coal. The main body of the horizontal well is designed to be located in the 12 coal seam. The coal seam is transformed through hydraulic fracturing transformation, and the two coal seams are vertically connected through hydraulic fractures. Simulate what combination of fracture parameters can maximize the coalbed methane production capacity resources.

[0110] The specific steps of optimizing fracture parameters of horizontal wells by using the present invention are as follows:

[0111] Step 1: Using Petrel-Re reservoir numerical simulation software, based on the block coring data and reservoir physical property data as the data basis, the Petrel-Re commercial numerical simulation software black oil simulator was used to establish a complex superimposed coal-gas reservoir model. The depth of the middle reservoir is 743m, the formation temperature is 29℃, the reservoir pressure gradient of 10 coal is 0.88MPa / 100m, the reservoir pressure gradient of 12 coal is 0.89MPa / 100m, and the water saturation in the fracture is 99%.

[0112] Based on the development of natural fractures in coal seams, and the extremely low matrix permeability and porosity, the fracture permeability is much greater than the matrix permeability, and the gas mainly flows in the fractures. Therefore, a dual medium model is used to characterize the coal seams. The model uses a Cartesian grid with a grid size of 10×10×0.5m and a model size of 1500×510×25.5m. The hydraulic fracture is set to expand along the direction of the maximum principal stress. The local grid encryption technology and the equivalent conductivity method are used to accurately simulate the artificial hydraulic fractures.

[0113] Step 2: Use numerical simulation methods to select the main controlling factors of coalbed methane fracturing horizontal well development effects, set sensitivity schemes, and conduct sensitivity studies of different parameters to obtain the main factors affecting coalbed productivity. The main factors selected are fracture spacing, fracture height, fracture conductivity, and fracture length.

[0114] Table 1 Parameter settings of influencing factors

[0115]

[0116] Step 3: Through the self-generated model of the constructed fracture parameter numerical simulation samples, the intersect numerical simulator is called based on Python to automatically construct a coalbed methane production capacity prediction sample set. The input parameters of the model are fracture height, fracture length, fracture conductivity and fracture spacing, and the output data is the cumulative gas production.

[0117] Step 4: Based on the XGBoost algorithm, a coalbed methane production capacity prediction proxy model is established. The prediction sample set obtained by the self-generated model of numerical simulation samples is divided into a training set and a validation set according to the relevant ratio of 80% and 20%. The total number of self-generated samples is 500, and the number of samples in the training set and the test set is 400 and 100 respectively. The comparison between the prediction results on the training set and the validation set and the actual results is shown in the figure below. Fig. 9 As shown in the figure, R2 is 0.935. When R2 is larger and MSE is smaller, it means that the error between the predicted value and the actual value of the model is smaller, and the performance of the model is better. Based on the optimized capacity prediction model, the optimal capacity solution is found, thereby determining the global optimal solution for the fracture parameters.

[0118] Table 2 Effect diagram of the model performance on evaluation indicators

[0119]

[0120] The actual application proves that this method can effectively solve the coalbed methane fracture parameter optimization problem in the target block, and the solution obtained by the optimization calculation has good performance in development effect, indicating that the production capacity prediction agent model based on numerical simulation has high credibility and can well assist on-site fracturing decision-making.

[0121] This invention combines the results of various tests on coal rock cores and uses the Petrel-Re numerical simulation software to construct a numerical simulation model that takes into account coalbed methane adsorption-desorption-diffusion, Lagmuir isothermal adsorption, interlayer heterogeneity, and other characteristics of multi-layered coal seams. At the same time, artificial fractures are used to construct a fracturing horizontal well model, and numerical simulation methods are used to screen the main control factors affecting coalbed methane production capacity. At the same time, this invention uses Python to call a numerical simulator to realize the self-generation of numerical simulation samples, and uses the XGboost algorithm to realize the construction of a production prediction proxy model, thereby obtaining the optimal solution for the fracture parameter combination of coalbed methane fracturing horizontal wells.

[0122] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0123] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0124] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for optimizing the matching relationship of fracture parameters of horizontal wells with complex coal structures, characterized in that: The following steps are involved: Step 1: Establish a fine numerical simulation model of coal seams with complex coal structure through the Petrel-Re numerical simulator; Step 2: Screen the main controlling factors of coalbed methane fracturing horizontal well development effect through numerical simulation method; Step 3: Construct a self-generated model of numerical simulation samples; Step 4: Establish a coalbed methane production capacity prediction agent model based on the XGBoost algorithm.

2. The method for optimizing the design of fracture parameter matching relationship of a horizontal well with a complex coal structure according to claim 1, characterized in that: The specific process of step one is as follows: Using Petrel-Re reservoir numerical simulation software, the grid system of the basic model is established; The fluid model adopts the black oil model, taking into account Lagmuir temperature adsorption and Fick diffusion; The grid number of the conceptual model is 150×51×51, the grid step size in the X and Y directions is 10m, the grid step size in the Z direction is 0.5m, the model is 1500m long and 510m wide; It is divided into 8 small layers in total, of which the second and sixth layers are coal seams, and the rest are rock and soil layers other than coal seams; The input parameters of the model are determined based on the coal core experimental data and mine data. The gas-water two-phase is taken as the research object, the double-porosity and double-permeability model is selected, and the Intersect simulator is used for numerical simulation.

3. The method for optimizing the design of fracture parameter matching relationship of a horizontal well with a complex coal structure according to claim 1, characterized in that: The specific process of step 2 is as follows: First, analyze the factors affecting the productivity of coalbed methane fracturing horizontal wells: Including matrix porosity, gas saturation, artificial fracture conductivity, artificial fracture half-length, fracture height, fracture spacing, and horizontal section length; The data range design of the main control factors: Through parameter sensitivity analysis, four key influencing factors were identified, including the half-length of artificial fractures, the distance between fractures, the conductivity of fractures and their height; The full factorial experimental design method involves conducting at least one experiment on all factors at all levels to estimate the main effects of each factor and their interactions.

4. The method for optimizing the design of fracture parameter matching relationship of a horizontal well with a complex coal structure according to claim 1, characterized in that: The specific process of step three is as follows: Automatically generate numerical simulation files: Based on the data range of the main control factors screened in step 2, a simulation scheme is generated by random combination. The ixf file of the intersect simulator is used to assemble keywords through Python definition functions to achieve automatic filling of crack length, crack height, crack conductivity and crack spacing, and the design scheme is assembled using the AFI file. Afterwards, the numerical simulator is automatically called in batches and the results are extracted: Based on Python, the Intersect environment in the computer is called to batch process the automatically generated AFI files, and the required numerical simulation result files are extracted and saved in a new folder for later extraction and combination of simulator results; Based on the numerical simulation result file (PRT file), by marking the coalbed methane geological reserves of the first time node and the coalbed methane geological reserves of the last time node of each numerical simulation result file, a loop function is defined, and the cumulative gas production of different schemes is automatically extracted and formatted and saved in a CSV file, thereby realizing the full automation of simulator result extraction and storage.

5. The method for optimizing the design of fracture parameter matching relationship of a horizontal well with a complex coal structure according to claim 1, characterized in that: The specific process of step 4 includes: prediction sample set division, prediction model establishment and model prediction effect evaluation.

6. The method for optimizing the design of fracture parameter matching relationship of a horizontal well with a complex coal structure according to claim 5, characterized in that: The process of dividing the prediction sample set is as follows: the prediction sample set obtained by the self-generation model of the numerical simulation sample is divided into a training set and a validation set according to the relevant ratio of 80% and 20%; The training set is mainly used to train the prediction model; The validation set is mainly used to adjust hyperparameters, including the number of decision trees, learning rate, and maximum depth of the tree; Each set contains the input and output data of the model, where the input data includes fracture height, fracture length, fracture conductivity and fracture spacing, and the output data includes the cumulative gas production of coalbed methane.

7. The method for optimizing the design of fracture parameter matching relationship of a horizontal well with a complex coal structure according to claim 5, characterized in that: The process of establishing the prediction model is as follows: using the XGBoost algorithm for modeling, the implementation process of the XGBoost algorithm is as follows: (1) Order i represents the i-th sample in the training set (X, y) (2) Satisfy 1≤t≤N k ; ①Construct cost function: Among them, g i and h i They represent the first-order derivative and second-order derivative of the loss function l at sample i respectively; y is the true label value of the sample; represents the XGBoost prediction model, x i is the input feature parameter of the decision tree model; r i Residual, N k is the number of decision tree models; T represents the number of leaves of the tree model, γ, λ represent the regularization coefficients, ω j Indicates the leaf node value; ②Use the data set (X, r) to train a tree model with a maximum depth of d ③Update the prediction value of the XGBoost model ④ Update the residual term r (3) Obtain the final prediction model Here, ε represents the compression factor or learning rate.

8. The method for optimizing the design of fracture parameter matching relationship of a horizontal well with a complex coal structure according to claim 7, characterized in that: The process of XGBoost model building includes the following steps: The first step is to perform feature classification by continuously adding trees, and each newly generated tree learns a new function f(x) to gradually grow a complete tree; The second step is to predict the sample scores after training k trees. According to the characteristics of the sample, the sample is classified into the leaf nodes of each tree and the corresponding scores are obtained. In the third step, you only need to sum up the scores calculated by each tree to get the final prediction value corresponding to the sample.

9. The method for optimizing the design of fracture parameter matching relationship of a horizontal well with a complex coal structure according to claim 5, characterized in that: The specific process of evaluating the model prediction effect is as follows: The prediction effect of the capacity prediction model was evaluated and analyzed by using the determination coefficient (R2) and mean square error (MSE); In the formula, y i and are the actual value and the predicted value, is the average of the actual values, and N is the size of the actual values ​​in the validation set. The larger the R2 and the smaller the MSE, the smaller the error between the predicted value and the actual value of the model, and the better the performance of the model.

Images