CN113987929A - Coal seam permeability change prediction method based on FA-SSA-SVM algorithm - Google Patents

Abstract

The invention proposes a coal seam permeability change prediction method based on the FA-SSA-SVM algorithm, which belongs to the field of CO ₂ storage. The method is to build a new hybrid intelligent model by integrating support vector machine SVM, sparrow search algorithm SSA, firefly algorithm FA, in which SVM is used to explore the relationship between the change of coal reservoir permeability and its influencing variables, SSA and FA are used to optimize SVM hyperparameters. Using FA to perturb the SSA, by perturbing the general solution and the optimal solution in the SSA, the searchability of the global solution is improved, and the global optimal solution of the SVM hyperparameter is obtained, so as to realize nonlinear prediction of coal reservoir permeability change and its influence in high-dimensional space. relationship between variables. The FA‑SSA‑SVM prediction model can achieve a good prediction effect on the basis of considering various complex factors in the CO ₂ sequestration process, and has the advantages of low cost, high prediction accuracy, and strong generalization ability.

Description

A Coal Seam Permeability Change Prediction Method Based on FA-SSA-SVM Algorithm

technical field

The invention relates to a method for predicting changes in coal seam permeability, in particular to a method for predicting changes in coal seam permeability based on FA-SSA-SVM algorithm used in the change of coal seam permeability in the process of CO ₂ sequestration, and belongs to coal seam CO ₂ geological storage field.

Background technique

_CO2 is one of the greenhouse gases, and a large amount of emissions can cause the greenhouse effect. my country has proposed to achieve carbon peaking by 2030 and carbon neutrality by 2060. The capture and storage of CO ₂ is an effective method for carbon emission reduction, and the injection of CO ₂ into deep coal seams can achieve the dual effects of CH ₄ displacement and CO ₂ sequestration, which has the dual significance of protecting the environment and developing energy.

The field test found that the injectability of CO 2 was significantly reduced after the injection of CO ₂ into the coal seam, which was caused by the reduction of crack opening or the phenomenon of crack closure caused by the adsorption and expansion of the coal matrix, which inhibited the gas permeability.

Studying the dynamic change of coal reservoir permeability during CO ₂ storage is related to the success or failure of storage. At present, the methods of obtaining coal seam permeability mainly include laboratory tests and theoretical models. The former has the disadvantages of complicated testing process, long time consumption, and expensive testing fees. The theoretical model has too many assumptions, and it is difficult to consider the complex changes of coal seam CO ₂ geological storage. Artificial intelligence models have been well used in engineering prediction, but there is still a lack of research on using artificial intelligence models to predict changes in coal seam permeability during CO ₂ sequestration.

SUMMARY OF THE INVENTION

Aiming at the shortcomings of the prior art, a method for predicting the change of coal seam permeability based on the FA-SSA-SVM algorithm is provided. The method is simple in steps and convenient to use, and can quickly predict the change of coal seam permeability in the process of CO ₂ sequestration in coarse coal seams. Variety.

In order to realize the above-mentioned technical purpose, the coal seam permeability change prediction method based on the FA-SSA-SVM algorithm of the present invention is characterized in that: comprising the following steps:

Step 1. Based on the coal seam adsorption expansion permeability test test and the coal seam CO ₂ sequestration industrial test, analyze the key geological parameters and construction parameters that affect the change of coal seam permeability, and optimize the key geological parameters and construction parameters to determine the model input variables;

Step 2: Sort out and collect laboratory test and literature retrieval data, determine the data covering multiple mines in different regions and various coal ranks through screening and de-false, collect all the data into a data set, and divide the data set into Training set and test set; the data set includes different mine data and the corresponding coal reservoir CO ₂ injection pressure, coal body effective stress, coal rank, coal body temperature, coal seam burial depth, and coal seam permeability change information;

Step 3: Build the FA-SSA-SVM prediction model, use the support vector machine SVM to build the relationship between the change of coal reservoir permeability and the input variables of the model, connect the sparrow search algorithm SSA and the firefly algorithm FA in series, and then jointly optimize the super performance of the SVM. parameters, optimize and determine the initialization parameter settings of the FA-SSA-SVM prediction model;

Step 4: Use the training set to repeatedly train the FA-SSA-SVM prediction model, use the test set to verify the prediction performance of the trained FA-SSA-SVM prediction model, and use the correlation coefficient R, mean absolute error MAE, root mean square error RMSE and The mean absolute percentage error MAPE verifies and evaluates the prediction performance of the trained FA-SSA-SVM prediction model, and continues training if the evaluation fails to meet the standard;

Step 5. Use the FA-SSA-SVM prediction model that has reached the evaluation standard to predict the change of coal reservoir permeability during CO ₂ sequestration.

Key geological parameters and construction parameters include effective stress of coal body, coal rank, elastic modulus of coal body, Poisson's ratio of coal body, porosity of coal body, initial permeability coefficient of coal body, coal body temperature, and coal seam burial depth; the key The construction parameters include CO ₂ injection pressure and CO ₂ injection temperature; the optimally determined model input variables include CO ₂ injection pressure, effective coal body stress, coal rank, coal body temperature and coal seam burial depth.

In the FA-SSA-SVM intelligent model, because SSA is easy to fall into the local optimum, it cannot obtain the global optimal solution, so the FA is used to perturb the SSA, and the searchability of the global solution is improved by perturbing the general solution and the optimal solution in the SSA. The global optimal solution is updated; the FA-SSA combination algorithm is used to optimize the hyperparameters of the SVM, the hyperparameters of the SVM include the penalty factor C and the kernel parameter g, and obtain the nonlinear prediction of coal reservoir permeability change and its influencing variables in higher dimensional space The relationship between;

The optimization is as follows:

a Construct the SSA algorithm optimization framework, including randomly initializing the sparrow population, and setting the maximum number of iterations, the proportion of discoverers, the proportion of joiners and the proportion of alerters; use the SSA algorithm to optimize the hyperparameters of the SVM; calculate the fitness value of the initial population, and carry out Sort to determine the initial optimal value; the fitness value calculation formula is as follows:

Among them, f is the fitness value, n is the number of sparrows, and d is the dimension of the problem variable to be optimized;

b Update the positions of discoverers, joiners, and alerters, and recalculate the fitness value. If the current optimal value is better than the last iteration, update the position of the sparrow, otherwise do not update the position of the sparrow, and continue to iterate until the conditions are met until;

Where the finder's location update description:

Among them, t is the current number of iterations, iter _max is the maximum number of iterations, X _i,j is the position information of the ith sparrow in the jth dimension, α∈(0,1] is a random number, R ₂ (R ₂ ∈ [0,1) and ST(ST∈(0.5,1) are the warning value and the safety value respectively, Q is a random number obeying the normal distribution, L is a 1×d matrix, each element in the matrix is 1;

Joiner's location update description:

Among them, X _p is the optimal position currently occupied by the discoverer, X _worst is the current global worst position, A is a 1 × d matrix, each element in the matrix is randomly assigned to 1 or -1, and A ⁺ = ^AT (AA ^T ) ^-1 ; when i>n/2, it indicates that the i-th participant with a lower fitness value has not obtained food and is in a very hungry state. At this time, it needs to fly to other places for food. to get more energy;

The vigilante's location update description:

Among them, X _best is the current global optimal position, β is the step size control parameter, which obeys the random number of normal distribution with mean 0 and variance 1, K∈[-1,1] is a random number, f _i is the fitness value of the current sparrow individual; f _g and f _w are the current global best and worst fitness values, respectively, and ε is the smallest constant to avoid zero in the denominator;

c When the SSA algorithm optimizes the hyperparameters of the SVM, use the FA algorithm to continue to optimize the hyperparameters of the SVM: build the FA algorithm optimization framework, and initialize the number of fireflies, firefly step factor, maximum attraction, and light intensity attraction coefficient;

d Calculate the fitness value of each firefly by using the firefly’s position. The firefly with better fitness value has higher brightness; all fireflies fly to individuals with higher brightness than themselves, and the position update formula is:

分别为萤火虫i和j的位置，β₀为萤火虫的吸引度，α为步长因子，rand为[0,1]上服从均匀分布的随机数；in,

are the positions of fireflies i and j, respectively, β ₀ is the attraction of fireflies, α is the step factor, and rand is a random number that obeys a uniform distribution on [0,1];

e Calculate the fitness value of the firefly's new position. If the position is better than the position before the flight, update the firefly position to the new position, otherwise do not update the firefly position, and continue the iterative operation until the conditions are met.

The parameters of the mixed model are set as follows: the number of sparrows is 50, the maximum number of iterations is 100, the ratio of discoverers is 0.6, the ratio of joiners is 0.4, the ratio of alerters is 0.2, the firefly step factor is 0.1, and the maximum attraction is 2, the light intensity attraction coefficient is 1.

Data set division: The division ratio of the training set and the test set in the entire data set is realized by the k-fold cross-validation method, and the division ratio is 7:3 or 8:2 to divide the training set and the test set.

Beneficial effects:

This method uses the hybrid intelligent model of the integrated support vector machine SVM, the sparrow search algorithm SSA, and the firefly algorithm FA to predict the change of coal reservoir permeability during the _{CO 2 storage} process. Good prediction effect has the advantages of low cost, high prediction accuracy, and strong generalization ability. Compared with traditional theoretical models, this method can consider various complex conditions; compared with laboratory experiments, this method has the advantages of simple test process, short time-consuming time, and low test cost.

Description of drawings

Fig. 1 is a flow chart of the coal seam permeability change prediction method based on the FA-SSA-SVM algorithm of the present invention.

Fig. 2 is the flow chart of the sparrow search algorithm SSA-firefly algorithm FA of the present invention.

FIG. 3 is a graph of the prediction result of the training set of the calculation example of the present invention.

FIG. 4 is a graph of the prediction result of the test set of the calculation example of the present invention.

Detailed ways

The implementation of the present invention will be further described below in conjunction with the accompanying drawings:

As shown in Figure 1 and Figure 2, a coal seam permeability change prediction method based on the FA-SSA-SVM algorithm, a new hybrid intelligent model is constructed by integrating the support vector machine SVM, the sparrow search algorithm SSA, and the firefly algorithm FA. To explore the relationship between changes in coal reservoir permeability and its influencing variables, SSA and FA are used to optimize the hyperparameters of SVM. Using FA to perturb the SSA, by perturbing the general solution and the optimal solution in the SSA, the searchability of the global solution is improved, and the global optimal solution of the SVM hyperparameter is obtained, so as to realize nonlinear prediction of coal reservoir permeability change and its influence in high-dimensional space. relationship between variables. The FA-SSA-SVM prediction model can achieve a good prediction effect on the basis of considering various complex factors in the CO ₂ sequestration process, and has the advantages of low cost, high prediction accuracy and strong generalization ability.

The specific steps are:

Step 1. Based on the coal seam adsorption expansion permeability test test and the coal seam CO ₂ sequestration industrial test, analyze the key geological parameters and construction parameters that affect the change of coal seam permeability, and optimize the key geological parameters and construction parameters to determine the model input variables;

Step 2: Sort out and collect laboratory test and literature retrieval data, determine the data covering multiple mines in different regions and various coal ranks through screening and de-false, collect all the data into a data set, and divide the data set into Training set and test set; the data set includes different mine data and the corresponding coal reservoir CO ₂ injection pressure, coal body effective stress, coal rank, coal body temperature, coal seam burial depth, and coal seam permeability change information;

Step 3. Build a hybrid intelligent model integrating support vector machine SVM, sparrow search algorithm SSA, and firefly algorithm FA, namely FA-SSA-SVM prediction model, in which SVM is used to construct the relationship between the change of coal reservoir permeability and the input variables of the model. relationship, SSA and FA are connected in series and jointly optimize the hyperparameter penalty factor C and kernel parameter g of SVM; at the same time, optimize and determine the initialization parameter settings of the FA-SSA-SVM prediction model;

Step 4: Use the training set to repeatedly train the FA-SSA-SVM prediction model, use the test set to verify the prediction performance of the trained FA-SSA-SVM prediction model, and use the correlation coefficient R, mean absolute error MAE, root mean square error RMSE and The mean absolute percentage error MAPE verifies and evaluates the prediction performance of the trained FA-SSA-SVM prediction model;

Step 5: Input the relevant input variables of the field test area into the FA-SSA-SVM prediction model to predict the coal seam permeability, and collect the field measured coal seam permeability in real time to enrich the model training set, so as to continuously modify the FA-SSA-SVM prediction model.

Specific embodiment one,

Step 1: Based on the coal seam adsorption expansion permeability test test and the coal seam CO ₂ sequestration industrial test, analyze the key geological parameters and construction parameters that affect the coal seam permeability change, and preferably determine the model input variables including the following parameters: CO ₂ injection pressure, coal mass Effective stress, coal rank, coal body temperature and coal seam burial depth.

In the second step, by sorting out the data of laboratory tests and literature search, a total of 254 sets of data were collected, covering mines in China, Japan, the United Kingdom and other countries, involving different coal ranks such as lignite, bituminous coal, and anthracite. Its basic parameter characteristics are shown in Table 1. Since the algorithm model needs to be trained and learned, after the original data set is filtered and de-pseudo-preprocessed, the data set is divided into training set and test set.

Table 1 Data Statistics of the Dataset

The third step is to establish the prediction model of the FA-SSA-SVM hybrid algorithm, in which SVM is used to explore the relationship between the change of coal reservoir permeability and its influencing variables, and FA-SSA is used to optimize the SVM. Because SSA is easy to fall into local optimum, it cannot obtain the global optimum solution. As shown in Figure 2, using FA to perturb the SSA, by perturbing the general solution and the optimum solution in the SSA, the searchability of the global solution is improved, and the global optimum solution is updated. The optimal solution is to optimize the hyperparameters (penalty factor C and kernel parameter g) of the SVM by using the FA-SSA combination algorithm to achieve nonlinear prediction of the relationship between the change of coal reservoir permeability and its influencing variables in high-dimensional space. According to the simulation experience and data set characteristics, the parameters of the mixed model are set as follows: the number of sparrows is 50, the maximum number of iterations is 100, the ratio of discoverers is 0.6, the ratio of joiners is 0.4, the ratio of alerters is 0.2, and the firefly step size is 0.6. The factor is 0.1, the maximum attraction is 2, and the light intensity attraction coefficient is 1.

Step 4: In order to analyze the relationship between the predicted value of the model and the actual value, R, MAE, RMSE and MAPE are selected for evaluation. The closer R is to 1, the smaller the MAE, RMSE, and MAPE, indicating that the predicted value is better correlated with the true value. In the model establishment and evaluation, the prediction effect of the training set is shown in Figure 3. The R value of the entire training set is 0.9794, the MAE is 0.0003, the RMSE is 0.0008, and the MAPE is 0.2385, which indicates that the prediction of the FA-SSA-SVM model training set is 0.2385. The value has a good correlation with the true value and the error is small.

After the model is trained, the FA-SSA-SVM model is tested on the test set. Figure 4 shows the prediction effect of the model test set. It can be found that the test sample points are basically distributed around the ideal fitting line "measured value = predicted value", the R value is 0.9315, and the MAE, RMSE, and MAPE are 0.0008, 0.0011, and 0.2375, respectively, which indicates that the FA-SSA-SVM prediction model The test results are very good, and the error is not large. Regardless of the test set or the training set, the predicted value of the FA-SSA-SVM model is basically consistent with the real experimental value, which indicates that the FA-SSA-SVM prediction model proposed in this patent can effectively predict the permeability of coal reservoirs during CO ₂ storage. rate change.

Step 5: Use the prediction results of the model to guide the design engineering practice, collect the measured engineering data and input it into the model to correct it, enrich the input data set of the model, and further improve the prediction accuracy and generalization ability of the model.

Matters not addressed in the present invention are known in the art.

The above-mentioned embodiments are only intended to illustrate the technical concept and characteristics of the present invention, and the purpose thereof is to enable those who are familiar with the art to understand the content of the present invention and implement them accordingly, and cannot limit the protection scope of the present invention. All equivalent changes or modifications made according to the spirit of the present invention should be included within the protection scope of the present invention.

Claims (5)

1. a coal seam permeability variation prediction method based on FA-SSA-SVM algorithm, is characterized in that: comprise the following steps: Step 1. Based on the coal seam adsorption expansion permeability test test and the coal seam CO ₂ sequestration industrial test, analyze the key geological parameters and construction parameters that affect the change of coal seam permeability, and optimize the key geological parameters and construction parameters to determine the model input variables; Step 2: Sort out and collect laboratory test and literature retrieval data, determine the data covering multiple mines in different regions and various coal ranks through screening and de-false, collect all the data into a data set, and divide the data set into Training set and test set; the data set includes different mine data and the corresponding key geological parameters and construction parameters that affect the change of coal seam permeability; Step 3: Build the FA-SSA-SVM prediction model, use the support vector machine SVM to build the relationship between the change of coal reservoir permeability and the input variables of the model, connect the sparrow search algorithm SSA and the firefly algorithm FA in series, and then jointly optimize the super performance of the SVM. parameters, optimize and determine the initialization parameter settings of the FA-SSA-SVM prediction model; Step 4: Use the training set to repeatedly train the FA-SSA-SVM prediction model, use the test set to verify the prediction performance of the trained FA-SSA-SVM prediction model, and use the correlation coefficient R, mean absolute error MAE, root mean square error RMSE and The mean absolute percentage error MAPE verifies and evaluates the prediction performance of the trained FA-SSA-SVM prediction model, and continues training if the evaluation fails to meet the standard; Step 5. Use the FA-SSA-SVM prediction model that has reached the evaluation standard to predict the change of coal reservoir permeability during CO ₂ sequestration. 2. the coal seam permeability change prediction method based on FA-SSA-SVM algorithm according to claim 1, is characterized in that: key geological parameter and construction parameter comprise coal body effective stress, coal rank, coal body elastic modulus, coal body Body Poisson's ratio, coal body porosity, coal body initial permeability coefficient, coal body temperature, coal seam burial depth; the key construction parameters include CO ₂ injection pressure, CO ₂ injection temperature; preferably determined model input variables include CO ₂ Injection pressure, effective stress of coal body, coal rank, coal body temperature and coal seam burial depth. 3. the coal seam permeability change prediction method based on FA-SSA-SVM algorithm according to claim 1, is characterized in that: in FA-SSA-SVM intelligent model, because SSA is easy to fall into local optimum, it cannot obtain global optimum. Therefore, FA is used to perturb the SSA, and the searchability of the global solution is improved by perturbing the general solution and the optimal solution in the SSA, so as to update the global optimal solution; the FA-SSA combination algorithm is used to optimize the hyperparameters of the SVM, and the hyperparameters of the SVM are The parameters include the penalty factor C and the kernel parameter g to obtain the nonlinear prediction of the relationship between the change of coal reservoir permeability and its influencing variables in higher dimensional space; The optimization is as follows: a Construct the SSA algorithm optimization framework, including randomly initializing the sparrow population, and setting the maximum number of iterations, the proportion of discoverers, the proportion of joiners and the proportion of alerters; use the SSA algorithm to optimize the hyperparameters of the SVM; calculate the fitness value of the initial population, and carry out Sort to determine the initial optimal value; the fitness value calculation formula is as follows:

Among them, f is the fitness value, n is the number of sparrows, and d is the dimension of the problem variable to be optimized; b Update the positions of discoverers, joiners, and alerters, and recalculate the fitness value. If the current optimal value is better than the last iteration, update the position of the sparrow, otherwise do not update the position of the sparrow, and continue to iterate until the conditions are met until; Where the finder's location update description:

Among them, t is the current number of iterations, iter _max is the maximum number of iterations, X _i,j is the position information of the ith sparrow in the jth dimension, α∈(0,1] is a random number, R ₂ (R ₂ ∈ [0,1) and ST(ST∈(0.5,1) are the warning value and the safety value respectively, Q is a random number obeying the normal distribution, L is a 1×d matrix, each element in the matrix is 1; Joiner's location update description:

Among them, X _p is the optimal position currently occupied by the discoverer, X _worst is the current global worst position, A is a 1×d matrix, each element in the matrix is randomly assigned to 1 or -1, and A ⁺ =A ^T (AA ^T ) ^-1 ; when i>n/2, it indicates that the i-th participant with a lower fitness value has not obtained food and is in a very hungry state, and needs to fly to other places for food at this time, to get more energy; The vigilante's location update description:

Among them, X _best is the current global optimal position, β is the step size control parameter, which obeys the random number of normal distribution with mean 0 and variance 1, K∈[-1,1] is a random number, f _i is the fitness value of the current sparrow individual; f _g and f _w are the current global best and worst fitness values, respectively, and ε is the smallest constant to avoid zero in the denominator; c When the SSA algorithm optimizes the hyperparameters of the SVM, use the FA algorithm to continue to optimize the hyperparameters of the SVM: build the FA algorithm optimization framework, and initialize the number of fireflies, firefly step factor, maximum attraction, and light intensity attraction coefficient; d Calculate the fitness value of each firefly by using the firefly’s position. The firefly with better fitness value has higher brightness; all fireflies fly to individuals with higher brightness than themselves, and the position update formula is:

in,

are the positions of fireflies i and j, respectively, β ₀ is the attraction of fireflies, α is the step factor, and rand is a random number that obeys a uniform distribution on [0,1]; e Calculate the fitness value of the firefly's new position. If the position is better than the position before the flight, update the firefly position to the new position, otherwise do not update the firefly position, and continue the iterative operation until the conditions are met. 4. the coal seam permeability change prediction method based on FA-SSA-SVM algorithm according to claim 1, it is characterized in that the input parameter of FA-SSA-SVM mixed model is: the number of sparrows is 50, and the maximum number of iterations is 100, The ratio of discoverers is 0.6, the ratio of joiners is 0.4, the ratio of alerters is 0.2, the firefly step factor is 0.1, the maximum attraction is 2, and the light intensity attraction coefficient is 1. 5. the coal seam permeability variation prediction method based on FA-SSA-SVM algorithm according to claim 1, is characterized in that the data set is divided: the division ratio of training set and test set in the whole data set is realized by k-fold cross-validation method Yes, the division ratio is 7:3 or 8:2 to divide the training set and the test set.

Images