Method and system for predicting permeability of coalbed methane reservoir
Through the cross-entropy algorithm, a support vector machine was optimized and a coalbed methane reservoir permeability prediction model was constructed, which solved the problem of low prediction accuracy in the existing technology, and achieved a more efficient coalbed methane reservoir permeability prediction.
Patent Information
- Application Number
- CN202510110256.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-23
- Publication Date
- 2025-05-27
AI Technical Summary
The prior art has problems in the prediction of coalbed methane reservoir permeability, which is difficult to determine the structure and local optimality, resulting in low prediction accuracy.
The support vector machine is optimized by cross-entropy algorithm to construct a coalbed methane reservoir permeability prediction model. By pre-treating and normalizing the permeability parameters of the coalbed methane reservoir, a permeability sample set is generated, and the model is trained to predict based on this sample set.
The accuracy of coalbed methane permeability prediction is improved, and the model has the advantages of high stability, fast convergence speed and short running time, which can effectively guide the exploration and development of coalbed methane.
Smart Images
Figure CN120046013A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical fields of coal mine safety and coalbed methane development, and particularly relates to a method and system for predicting the permeability of a coalbed methane reservoir. Background Art
[0002] As a replacement energy for conventional oil and gas resources, coalbed methane has become an important part of the world's energy structure. Permeability, as an index to measure the ability of a porous medium to allow fluid to pass through, is one of the key measurement indexes affecting the production of coalbed methane wells and determining the recoverability of coalbed methane reservoirs. Therefore, accurately predicting the coalbed permeability before development can guide the exploration and development of resources and effectively improve the development efficiency.
[0003] Currently, the main methods for predicting the permeability of coal reservoirs include descriptive techniques, mathematical analysis, etc. Descriptive techniques are to find characteristic parameters related to permeability based on coal samples, such as filling degree, cleat density, cleat wall distance, etc. Mathematical analysis refers to establishing corresponding mathematical models through some mathematical methods to predict permeability, and the feasibility of this method has not been analyzed in practical applications. Currently, more and more mathematical methods are applied to geological research, such as statistics, pattern recognition, and artificial neural networks. However, the models established using artificial neural networks have defects such as difficult-to-determine structure and easy occurrence of local optima.
[0004] Therefore, there is an urgent need to propose a method and system for predicting the permeability of a coalbed methane reservoir to predict the permeability of the coalbed methane reservoir. Summary of the Invention
[0005] The purpose of the present invention is to provide a method and system for predicting the permeability of a coalbed methane reservoir to predict the permeability of the coalbed methane reservoir.
[0006] To achieve the above purpose, the present invention adopts the following technical solutions: A method for predicting the permeability of a coalbed methane reservoir includes the following steps: S1. Collect the permeability parameters of the coalbed methane reservoir and perform preprocessing; S2. Normalize the permeability parameters to obtain a permeability sample set; S3. Optimize the support vector machine based on the cross-entropy algorithm to construct a permeability prediction model; S4. Train the permeability prediction model based on the permeability sample set; S5. Predict the permeability of the coalbed methane reservoir based on the trained permeability prediction model, and evaluate the trained permeability prediction model based on the prediction data.
[0007] Further, in step S1, the process of preprocessing the permeability parameter includes: performing missing value processing on the permeability parameter based on the interpolation method or the deletion method, performing outlier processing on the permeability parameter based on the data range check method or the statistical method, performing data consistency check on the permeability parameter based on the logical relationship check method or the time series consistency method, and performing data quality control on the permeability parameter based on the data verification method or the data review method.
[0008] Further, in step S2, after the normalization process, the obtained permeability sample set includes at least permeability, reservoir pressure, in-situ stress, effective stress, coal seam thickness, and burial depth.
[0009] Further, in step S3, the process of optimizing the support vector machine based on the cross-entropy algorithm includes: taking the radial basis width and penalty factor of the support vector machine as the optimization objectives, taking the cross-validation probability of the support vector machine as the fitness function, first sampling the initial values of the radial basis width and penalty factor to obtain a number of candidate sample matrices; obtaining a sequence of fitness function values, sorting them from smallest to largest to obtain a new matrix; calculating the number of the candidate sample matrices based on the new matrix, and then performing update processing and smoothing processing on the radial basis width and penalty factor until the termination condition is met, then the iteration ends, and the optimal solutions of the radial basis width and penalty factor are obtained.
[0010] Further, in step S5, the process of evaluating the trained permeability prediction model based on the prediction data includes: obtaining the measured data of the coal seam gas reservoir permeability, obtaining the prediction data based on the trained permeability prediction model, comparing and analyzing the measured data and the prediction data, and then evaluating the accuracy of the trained permeability prediction model.
[0011] A coal seam gas reservoir permeability prediction system includes a parameter acquisition module for collecting the permeability parameters of the coal seam gas reservoir, a parameter processing module for preprocessing and normalizing the permeability parameters to obtain a permeability sample set, the parameter processing module is connected to the parameter acquisition module, a model construction module for optimizing the support vector machine based on the cross-entropy algorithm, constructing a permeability prediction model, and then training the permeability prediction model based on the permeability sample set, the model construction module is connected to the parameter processing module, and a model evaluation module for predicting the permeability of the coal seam gas reservoir based on the trained permeability prediction model and evaluating the trained permeability prediction model based on the prediction data, the model evaluation module is connected to the model construction module.
[0012] Further, the parameter processing module includes a parameter preprocessing unit for performing missing value processing on the permeability parameter based on the interpolation method or the deletion method, performing outlier processing on the permeability parameter based on the data range check method or the statistical method, performing data consistency check on the permeability parameter based on the logical relationship check method or the time series consistency method, and performing data quality control on the permeability parameter based on the data verification method or the data review method, and a parameter normalization unit for performing normalization processing on the preprocessed permeability parameter.
[0013] Further, the model construction module includes a model construction unit for taking the radial basis width and penalty factor of the support vector machine as the optimization objectives, taking the cross-validation probability of the support vector machine as the fitness function, first sampling the initial values of the radial basis width and penalty factor to obtain a number of candidate sample matrices; obtaining a sequence of fitness function values and sorting them from small to large to obtain a new matrix; calculating the number of the candidate sample matrices based on the new matrix, and then performing update processing and smoothing processing on the radial basis width and penalty factor until the termination condition is met, then the iteration ends, obtaining the optimal solutions of the radial basis width and penalty factor, and constructing a permeability prediction model based on the optimal solutions of the radial basis width and penalty factor, and a model training unit for training the permeability prediction model based on the permeability sample set.
[0014] Further, the model evaluation module includes a model prediction unit for predicting the permeability of the coalbed methane reservoir based on the trained permeability prediction model to obtain prediction data, and a model evaluation unit for obtaining the measured data of the permeability of the coalbed methane reservoir, comparing and analyzing the measured data and the prediction data, and further evaluating the accuracy of the trained permeability prediction model.
[0015] The beneficial effects of the present invention are as follows: Based on the cross-entropy algorithm to optimize the support vector machine, the constructed permeability prediction model has the advantages of high stability, fast convergence speed, short running time, etc. compared with other intelligent algorithms, and can improve the accuracy of coalbed methane permeability prediction; the proposed method for predicting the permeability of the coalbed methane reservoir provides help for the exploration and development of coalbed methane and has a relatively broad application prospect. Description of the Drawings
[0016] Figure 1 It is a schematic flow chart of the method for predicting the permeability of the coalbed methane reservoir according to the embodiment of the present invention; Figure 2 It is a schematic structural diagram of the system for predicting the permeability of the coalbed methane reservoir according to the embodiment of the present invention. Detailed Embodiments
[0017] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0018] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.
[0019] Embodiment As Figure 1 shown, in this embodiment, a method for predicting the permeability of a coalbed methane reservoir is provided, including the following steps: collecting the permeability parameters of the coalbed methane reservoir and performing preprocessing; performing normalization processing on the permeability parameters to obtain a permeability sample set; optimizing the support vector machine based on the cross-entropy algorithm to construct a permeability prediction model; training the permeability prediction model based on the permeability sample set; predicting the permeability of the coalbed methane reservoir based on the trained permeability prediction model, and evaluating the trained permeability prediction model based on the prediction data.
[0020] Implementable, the process of preprocessing the permeability parameters includes: performing missing value processing on the permeability parameters based on the interpolation method or the deletion method, performing outlier processing on the permeability parameters based on the data range check method or the statistical method, performing data consistency check on the permeability parameters based on the logical relationship check method or the time series consistency method, and performing data quality control on the permeability parameters based on the data verification method or the data review method.
[0021] Implementable, the permeability sample set after normalization processing includes at least: permeability, reservoir pressure, in-situ stress, effective stress, coal seam thickness, and burial depth.
[0022] Implementable, the process of optimizing the support vector machine based on the cross-entropy algorithm includes: taking the radial basis width and penalty factor of the support vector machine as the optimization objectives, taking the cross-validation probability of the support vector machine as the fitness function, first sampling the initial values of the radial basis width and penalty factor to obtain a number of candidate sample matrices; obtaining a sequence of fitness function values and sorting them from smallest to largest to obtain a new matrix; calculating the number of candidate sample matrices based on the new matrix, and then performing update processing and smoothing processing on the radial basis width and penalty factor until the termination condition is met, then the iteration ends, and the optimal solutions of the radial basis width and penalty factor are obtained.
[0023] As a specific embodiment, the process of optimizing the support vector machine based on the cross-entropy algorithm is as follows: Support Vector Machine (SVM) is simple in form and has strong superiority especially in small data problems. Among them, how to appropriately select parameters is the key to the algorithm, which will affect the accuracy of the constructed model. In order to coordinate the parameters of SVM, the method of Cross Entropy Support Vector Machine (CE-SVM) is selected to optimize two important parameters in SVM, namely the penalty factor c and the radial basis width g. Since the complexity of the non-linear optimization problem is determined by g, improper selection of the value of g will directly affect the adaptability of SVM; improper selection of the value of c will cause overfitting or underfitting.
[0024] Adopt the continuous cross-entropy algorithm, with the penalty factor c and the radial basis width g as the optimization objectives, and the cross-validation probability of SVM as the fitness function. The specific optimization steps are as follows: Step 1: Start: Assign initial values to the initial values of parameters c and g ( The dimension of is n), , the number of random samples M, the smoothing coefficient , and the quantile respectively, and let the iteration number t = 0.
[0025] Step 2: Sampling: Let t = t + 1, and generate M candidate sample matrices according to the distribution, where each is an n-dimensional vector, .
[0026] Step 3: Sorting: Sort the sequence of fitness function values in ascending order to obtain a new matrix , and then calculate the quantile of the sequence using the following formula.
[0027] Step 4: Update: Substitute the generated M random samples into the following formula to update the parameters and , where Step 5: Smoothing: For , calculate: In the formula: , and is the L-th element in the sequence after the t-th iteration.
[0028] Step 6: Stop: If the termination condition is met after the t-th iteration during the iterative process , terminate the iteration; otherwise, return to Step 1 and execute again.
[0029] Finally, obtain the optimal solution , penalty factor and radial basis width .
[0030] Implementable. The process of evaluating the trained permeability prediction model based on prediction data includes: obtaining the measured data of the permeability of the coalbed methane reservoir, obtaining the prediction data based on the trained permeability prediction model, comparing and analyzing the measured data and the prediction data, and then evaluating the accuracy of the trained permeability prediction model.
[0031] As Figure 2 shown, this embodiment also provides a coalbed methane reservoir permeability prediction system, including: a parameter acquisition module for collecting the permeability parameters of the coalbed methane reservoir; a parameter processing module connected to the parameter acquisition module for preprocessing and normalizing the permeability parameters to obtain a permeability sample set; a model construction module connected to the parameter processing module for optimizing the support vector machine based on the cross-entropy algorithm to construct a permeability prediction model, and then training the permeability prediction model based on the permeability sample set; a model evaluation module connected to the model construction module for predicting the permeability of the coalbed methane reservoir based on the trained permeability prediction model and evaluating the trained permeability prediction model based on the prediction data.
[0032] Implementable. The parameter processing module includes a parameter preprocessing unit and a parameter normalization unit; the parameter preprocessing unit is used to process the missing values of the permeability parameters based on the interpolation method or the deletion method, process the outliers of the permeability parameters based on the data range check method or the statistical method, check the data consistency of the permeability parameters based on the logical relationship check method or the time series consistency method, and control the data quality of the permeability parameters based on the data verification method or the data review method; the parameter normalization unit is used to normalize the preprocessed permeability parameters.
[0033] Implementable, the model construction module includes a model construction unit and a model training unit; the model construction unit is used to take the radial basis width and penalty factor of the support vector machine as the optimization objectives, and the cross-validation probability of the support vector machine as the fitness function. First, sample the initial values of the radial basis width and penalty factor to obtain a number of candidate sample matrices; obtain the sequence of fitness function values, sort them from small to large to obtain a new matrix; calculate the number of candidate sample matrices based on the new matrix, and then update and smooth the radial basis width and penalty factor until the termination condition is met, then the iteration ends, and the optimal solutions of the radial basis width and penalty factor are obtained. Based on the optimal solutions of the radial basis width and penalty factor, a permeability prediction model is constructed; the model training unit is used to train the permeability prediction model based on the permeability sample set.
[0034] Implementable, the model evaluation module includes a model prediction unit and a model evaluation unit; the model prediction unit is used to predict the permeability of the coalbed methane reservoir based on the trained permeability prediction model to obtain prediction data; the model evaluation unit is used to obtain the measured data of the permeability of the coalbed methane reservoir, compare and analyze the measured data and the prediction data, and then evaluate the accuracy of the trained permeability prediction model.
[0035] The above is only the preferred specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.
Claims
1. A method for predicting coalbed methane reservoir permeability, characterized in that: The steps include: S1. Collect and pre-process the permeability parameters of coalbed methane reservoirs; S2. normalizing the permeability parameters to obtain a permeability sample set; S3, optimizing the support vector machine based on the cross entropy algorithm and building a permeability prediction model; S4, training the permeability prediction model based on the permeability sample set; S5. Predict the permeability of the coalbed methane reservoir based on the trained permeability prediction model, and evaluate the trained permeability prediction model based on the prediction data.
2. A method for predicting coalbed methane reservoir permeability according to claim 1, characterized in that: In step S1, the process of preprocessing the permeability parameters includes: processing missing values of the permeability parameters based on interpolation or deletion, processing outliers of the permeability parameters based on data range check or statistical method, checking data consistency of the permeability parameters based on logical relationship check or time series consistency method, and controlling data quality of the permeability parameters based on data verification or data review method.
3. A method for predicting coalbed methane reservoir permeability according to claim 1, characterized in that: In step S2, after normalization processing, the permeability sample set obtained includes at least permeability, reservoir pressure, ground stress, effective stress, coal seam thickness and burial depth.
4. A method for predicting coalbed methane reservoir permeability according to claim 1, characterized in that: In step S3, the process of optimizing the support vector machine based on the cross entropy algorithm includes: taking the radial basis width and penalty factor of the support vector machine as the optimization target, taking the cross-validation probability of the support vector machine as the fitness function, first sampling the initial values of the radial basis width and the penalty factor to obtain a number of candidate sample matrices; obtaining a sequence of fitness function values, and sorting them from small to large to obtain a new matrix; calculating the number of the candidate sample matrices based on the new matrix, and then updating and smoothing the radial basis width and the penalty factor until the termination condition is met, then the iteration ends and the optimal solution of the radial basis width and the penalty factor is obtained.
5. A method for predicting coalbed methane reservoir permeability according to claim 1, characterized in that: In step S5, the process of evaluating the trained permeability prediction model based on the predicted data includes: obtaining measured data of the permeability of the coalbed gas reservoir, obtaining predicted data based on the trained permeability prediction model, comparing and analyzing the measured data and the predicted data, and then evaluating the accuracy of the trained permeability prediction model.
6. A coalbed methane reservoir permeability prediction system, characterized in that: It includes a parameter acquisition module for collecting permeability parameters of coalbed methane reservoirs, a parameter processing module for preprocessing and normalizing the permeability parameters to obtain a permeability sample set, the parameter processing module is connected to the parameter acquisition module, a model construction module is used to optimize the support vector machine based on the cross entropy algorithm, construct a permeability prediction model, and then train the permeability prediction model based on the permeability sample set, the model construction module is connected to the parameter processing module, a model evaluation module is used to predict the permeability of the coalbed methane reservoir based on the trained permeability prediction model, and evaluate the trained permeability prediction model based on the prediction data, and the model evaluation module is connected to the model construction module.
7. A coalbed methane reservoir permeability prediction system according to claim 6, characterized in that: The parameter processing module includes a parameter preprocessing unit for processing missing values of permeability parameters based on interpolation or deletion, processing outliers of permeability parameters based on data range check or statistical method, checking data consistency of permeability parameters based on logical relationship check or time series consistency method, and controlling data quality of permeability parameters based on data verification or data review method, and a parameter normalization unit for normalizing the preprocessed permeability parameters.
8. A coalbed methane reservoir permeability prediction system according to claim 6, characterized in that: The model building module includes a method for taking the radial basis width and penalty factor of the support vector machine as optimization targets and the cross-validation probability of the support vector machine as a fitness function, firstly sampling the initial values of the radial basis width and the penalty factor to obtain a plurality of candidate sample matrices; obtaining a sequence of fitness function values and sorting them from small to large to obtain a new matrix; The number of the candidate sample matrices is calculated based on the new matrix, and then the radial basis width and the penalty factor are updated and smoothed until the termination condition is met. The iteration ends and the optimal solution of the radial basis width and the penalty factor is obtained. A model construction unit for constructing a permeability prediction model based on the optimal solution of the radial basis width and the penalty factor and a model training unit for training the permeability prediction model based on the permeability sample set are constructed.
9. A coalbed methane reservoir permeability prediction system according to claim 6, characterized in that: The model evaluation module includes a model prediction unit for predicting the permeability of the coalbed methane reservoir based on the trained permeability prediction model to obtain prediction data and a model evaluation unit for obtaining measured data of the permeability of the coalbed methane reservoir, comparing and analyzing the measured data and the predicted data, and then evaluating the accuracy of the trained permeability prediction model.