A method for predicting the main controlling factors of tight gas production based on interactive data

Abstract

The present invention discloses a method for predicting the main controlling factors of tight gas production based on interaction data, including: collecting and organizing data of various characteristic factors affecting tight gas production according to geological research materials and fracturing construction reports, and performing data cleaning; calculating the maximum mutual information number between each pair of variables in the remaining sample data; normalizing the maximum mutual information value, selecting the maximum value of mutual information at different scales as the MIC value, and then converting it into a percentage score MIC i ; introducing the random forest regression algorithm to calculate the RF scores between each pair of variables 10 times in a loop, and then converting them into a percentage score RF i ; finally, taking the obtained MIC i value and the obtained RF I value as the mean value of the comprehensive score Total_Score i , and regarding Total_Score i as the main controlling factors affecting tight gas production. The present invention can effectively screen the main controlling factors affecting tight gas production, and the screened main controlling factors are very practical.

Description

Technical Field

[0001] The present invention relates to the technical field of gas reservoir development, and particularly relates to a method for predicting the main controlling factors of tight gas production based on interactive data. Background Art

[0002] Tight sandstone gas and tight gas, including tight volcanic rock gas and carbonate gas, are important unconventional oil and gas resources, and their production accounts for almost 70% of the global unconventional resource volume. Among the top 100 gas reservoirs in the United States by reserves, 58 are tight gas reservoirs. 80% of China's natural gas is stored in tight gas reservoirs, with extremely broad development potential. Therefore, accurately screening out the main controlling factors affecting tight gas production is of great significance for subsequent tight gas production prediction research.

[0003] According to the research results of domestic and foreign scholars, the production control factors of tight gas fracturing wells include geological factors and engineering factors. Geological factors generally include preservation conditions (such as burial depth and thickness of source rocks), gas saturation, and development of natural fractures; engineering factors generally include injection volume of fracturing fluid, injection amount of proppant, and fracturing and stimulation effect of the reservoir, etc. There are large deviations between the traditional methods for screening main controlling factors and actual production. For this reason, the method for screening main controlling factors based on mutual information is the latest and most practical method at present. Summary of the Invention

[0004] Aiming at the above problems, the present invention aims to provide a method for comprehensively screening the main controlling factors affecting tight gas production based on the maximum mutual information method and the random forest method, which is easier to implement and can make up for the deficiencies of the existing methods for screening the main controlling factors affecting tight gas production.

[0005] The technical solution of the present invention is as follows:

[0006] A method for predicting the main controlling factors of tight gas production based on interactive data, comprising the following steps:

[0007] Step1: According to geological research data and fracturing construction reports, collect and sort out various characteristic factor data affecting tight gas production, obtain the original data set, and perform data cleaning on the original data set to eliminate abnormal sample data;

[0008] Preferably, the obtained basic characteristic factor data includes 18 independent variables such as perforation depth (m), perforation interval length (m), source rock thickness (m), propped fracture width (mm), propped fracture height (m), fracture conductivity (md·m), total hydrocarbon peak (%)), resistivity (Ω·m), rock density (g / cm3), acoustic time difference (μs / m), porosity (%), gas saturation (%), total liquid volume (m3), percentage of preflush fluid (%), proppant volume (m3), average sand ratio (%), pumping rate (m3), and sand-carrying fluid volume (m3).

[0009] Preferably, the abnormal sample data is excluded from the basic data in Step 1, including the following sub-steps:

[0010] Step 101: Screen out the samples containing zeros in the basic sample data and exclude them;

[0011] Step 102: Calculate the increasing gradient of gas production, draw a scatter plot of the increasing gradient of gas production, and exclude the samples with too large changes in gas production;

[0012] Step 103: Exclude the samples with abnormal gas saturation.

[0013] Step 2: Calculate the maximum mutual information number between each pair of variables (including each independent variable and the dependent variable of tight gas production) in the remaining sample data;

[0014] Preferably, the calculation rule of the maximum mutual information number in Step 2 is as follows:

[0015] In information theory, according to the chain rule of entropy, there is

[0016] H(X,Y) = H(X) + H(Y|X) = H(Y) + H(X|Y) (1)

[0017] Therefore

[0018] H(X) - H(X|Y) = H(Y) - H(Y|X) (2)

[0019] The difference in formula (2) is called the mutual information of X and Y, denoted as I(X;Y).

[0020] The calculation formula of I(x;y) is as follows:

[0021]

[0022] In the formula: p(x), p(y) are the marginal distributions of two random variables (X,Y), and p(x,y) is the joint distribution of two random variables

[0023] (X,Y0

[0024] Step 3: Normalize the maximum mutual information value, select the maximum value of the mutual information at different scales as the MIC value, and then convert it to a percentage score MIC i ;

[0025] Preferably, Step 3 specifically includes the following sub-steps:

[0026] Step301: Divide the maximum mutual information obtained in Step2 by log2(min(a, b)) for normalization;

[0027] Step302: Take the maximum value among all the obtained normalized maximum mutual information values as the MIC value;

[0028] The specific calculation formula for the MIC value is as follows: Given i and j, grid the scatter plots formed by each independent variable and the dependent variable into i columns and j rows, and find the maximum mutual information value. The formula for calculating the mutual information value is as follows:

[0029]

[0030] In the formula: a represents the number of grids divided along the x direction; b represents the number of grids divided along the y direction; the empirical coefficient B ≈ n 0.6 , where n is the number of samples;

[0031] Step303: Convert the obtained MIC value score into a percentage-based MIC i , and the formula is as follows:

[0032]

[0033] Preferably, MIC has many advantages, and its most prominent advantages are as follows:

[0034] MIC has universality and fairness.

[0035] The so-called universality means that when the sample size is large enough (covering most of the information of the samples), it can capture various interesting associations, not limited to specific function types (such as linear functions, exponential functions, or periodic functions), or in other words, it can evenly cover all function relationships. Generally, the complex relationships between variables cannot be modeled by a single function alone, but need to be represented by superimposed functions. For functions with better universality, the starting points of different types of association relationships should be close and close to one.

[0036] The so-called fairness means that when the sample size is large enough, it can give similar coefficients for different types of correlation relationships with similar single-noise levels. For example, for a linear relationship and a sine relationship filled with the same noise, a good evaluation algorithm should give the same or similar correlation coefficients. And for a comparison method with better fairness, as the noise increases, the function changes of different types of association relationships should be similar.

[0037] Step4: Introduce the random forest regression algorithm to loop and calculate the RF scores between each pair of variables (including each independent variable and the dependent variable of tight gas production) 10 times, and then convert them into a percentage-based score RF i, and take the average of 10 times of RF i as the final RF score, namely RF I ;

[0038] Preferably, Step4 specifically includes the following sub-steps:

[0039] Step401: Use the random forest regression algorithm to calculate the RF scores between each pair of variables (including each independent variable and the dependent variable of tight gas production) in a loop for 10 times;

[0040] Step402: Convert the 10 times of RF scores into a percentage score RF i , and the formula is as follows:

[0041]

[0042] Step403: Take the average of 10 times of RF i as the final RF score RF I .

[0043] Step5: Take the average of the MIC i value obtained in Step3 and the RF I value obtained in Step4 as the comprehensive score Total_Score i , and sort the Total_Score i value to screen out the optimal independent variable, and use it as the main control factor affecting tight gas production.

[0044] Preferably, Step5 specifically includes the following steps:

[0045] Step501: Calculate the comprehensive score respectively after obtaining the MIC i and RF I values:

[0046]

[0047] Step502: According to the obtained comprehensive score values Total_Score i of each independent variable and the dependent variable, sort them, compare to obtain the optimal independent variable, and use it as the main control factor affecting tight gas production.

[0048] The beneficial effects of the present invention are:

[0049] On the basis of ensuring easy implementation, the present invention is more practical than the existing methods in screening the main control factors affecting tight gas production, and has extremely profound significance for the subsequent research on tight gas production prediction. Description of the Drawings

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0051] Figure 1 This is a schematic flowchart of the method for comprehensively screening the main controlling factors affecting the tight gas production based on the maximum mutual information method and the random forest method of the present invention;

[0052] Figure 2 This is a heat map of the correlation coefficient of the maximum information number of MIC calculated for Example 1;

[0053] Figure 3 This is a heat map of the correlation coefficient of the random forest of RF calculated for Example 1; Detailed implementation manners

[0054] The following further illustrates the present invention with reference to the drawings and embodiments. It should be noted that, without conflict, the embodiments in this application and the technical features in the embodiments can be combined with each other. It should be pointed out that unless otherwise specified, all technical and scientific terms used in this application have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs. The term "including" or "comprising" used in the disclosure of the present invention means that the elements or objects appearing before this term cover the elements or objects listed after this term and their equivalents, without excluding other elements or objects.

[0055] Example 1

[0056] As Figure 1 shown, a method for comprehensively screening the main controlling factors affecting the tight gas production based on the maximum mutual information method and the random forest method includes the following steps:

[0057] Step1: According to geological research data and fracturing construction reports, collect and sort out various characteristic factor data that affect the tight gas production, obtain the original data set, and perform data cleaning on the original data set, excluding sample data with zeros, excluding samples with overly large fluctuations in gas production, and excluding samples with abnormal gas saturation. Some results are shown in Table 1:

[0058] Table 1 Part of the sample data after cleaning

[0059]

[0060]

[0061] Note: Due to a large amount of sample data, only some sample data are listed in Table 1, where "…" indicates that there are unlisted data.

[0062] Step2: Calculate the maximum mutual information number between each pair of variables (including each independent variable and the dependent variable of tight gas production) in the remaining sample data;

[0063] The calculation rule of the maximum mutual information number in Step2 is as follows:

[0064] In information theory, according to the chain rule of entropy, we have

[0065] H(X,Y) = H(X) + H(Y|X) = H(Y) + H(X|Y) (1)

[0066] Therefore

[0067] H(X) - H(X|Y) = H(Y) - H(Y|X) (2)

[0068] The difference in Equation (2) is called the mutual information between X and Y, denoted as I(X;Y).

[0069] The calculation formula of I(x;y) is as follows:

[0070]

[0071] In the formula: p(x), p(y) are the marginal distributions of two random variables (X,Y), and p(x,y) is the joint distribution of two random variables

[0072] (X,Y).

[0073] Step3: Normalize the maximum mutual information value, select the maximum value of the mutual information under different scales as the MIC value, and then convert it to a percentage score MIC i , where Step3 specifically contains the following sub-

[0074] steps:

[0075] ① Divide the maximum mutual information obtained in Step2 by log2(min(a,b)) for normalization;

[0076] ② Take the maximum value of all the obtained normalized maximum mutual information values as the MIC value;

[0077] The specific calculation formula for the MIC value is: Given i, j, grid the scatter plot formed by each independent variable and the dependent variable into i columns and j rows, and find the maximum mutual information value. The formula for calculating the mutual information value is as follows:

[0078]

[0079] In the formula: a represents the number of grids divided along the x - direction; b represents the number of grids divided along the y - direction; the empirical coefficient B≈n 0.6 , where n is the number of samples;

[0080] The MIC correlation coefficient matrix is obtained as shown in Table 2 below:

[0081] Table 2 Partial MIC Correlation Coefficient Matrix

[0082]

[0083]

[0084] Note: Since there is a large amount of data in the MIC correlation coefficient matrix, only part of the MIC correlation coefficient matrix data is listed in Table 2, where "…" indicates that there are unlisted data.

[0085] Where: y represents the gas production rate (m3 / h), and x1 to x18 represent the perforation depth (m), perforated interval length (m), source rock thickness (m), proppant fracture width (mm), proppant fracture height (m), fracture conductivity (md·m), total hydrocarbon peak (%)), resistivity (Ω·m), rock density (g / cm3), acoustic travel time (μs / m), porosity (%), gas saturation (%), total liquid volume (m3), pre - flush percentage (%), proppant volume (m3), average sand ratio (%), pumping rate (m3), and sand - carrying fluid volume (m3), respectively.

[0086] Step303: Convert the obtained MIC value score into a percentage - based MIC i , and the formula is as follows:

[0087]

[0088] The MIC correlation coefficient matrices between each independent variable and the dependent variable of tight gas production are obtained i , as shown in Table 3 below:

[0089] Table 3 MIC i Correlation Coefficient Matrix

[0090]

[0091]

[0092] Where: y represents the gas production rate (m3 / h), and x1 to x18 respectively represent the perforation depth (m), perforation interval length (m), thickness of source rock (m), propped fracture width (mm), propped fracture height (m), fracture conductivity (md·m), peak value of total hydrocarbon (%), resistivity (Ω·m), rock density (g / cm3), acoustic travel time (μs / m), porosity (%), gas saturation (%), total liquid volume (m3), percentage of preflush fluid (%), proppant volume (m3), average sand ratio (%), pumping rate (m3), and carrier fluid volume (m3).

[0093] Step4: Introduce the random forest regression algorithm to calculate the RF scores between each pair of variables (including each independent variable and the dependent variable of tight gas production) in a loop for 10 times, and then convert them into a percentage-based score RF i , and take the mean of the 10 times of RF i as the final RF score, namely RF I , where Step4 specifically includes the following sub-steps:

[0094] ① Use the random forest regression algorithm to calculate the RF scores between each pair of variables (including each independent variable and the dependent variable of tight gas production) in a loop for 10 times;

[0095] ② Convert the 10 times of RF scores into a percentage-based score RF i , and the formula is as follows:

[0096]

[0097] ③ Take the mean of the 10 times of RF i as the final RF score RF I .

[0098] Obtain the RF I correlation coefficient matrix between each independent variable and the dependent variable of tight gas production, as shown in Table 4 below:

[0099] Table 4 RF I correlation coefficient matrix

[0100]

[0101]

[0102] Where: y represents the gas production rate (m3 / h), and x1 to x18 represent the perforation depth (m), perforation interval length (m), thickness of source rock (m), propped fracture width (mm), propped fracture height (m), fracture conductivity (md·m), peak value of total hydrocarbon (%), resistivity (Ω·m), rock density (g / cm3), acoustic travel time (μs / m), porosity (%), gas saturation (%), total liquid volume (m3), percentage of preflush fluid (%), proppant volume (m3), average sand ratio (%), pumping rate (m3), and sand-carrying fluid volume (m3), respectively.

[0103] Step5: Take the MIC i value obtained in Step3 and the RF I value obtained in Step4, and take their average as the comprehensive score Total_Score i ,

[0104]

[0105] to obtain the comprehensive score Total_Score i , as shown in Table 5 below:

[0106] Table 5 Comprehensive score Total_Score i

[0107]

[0108]

[0109] And sort the Total_Score i values to screen out the optimal independent variables, and take them as the main controlling factors affecting the tight gas production. According to the comparison, the following optimal independent variables, that is, the main controlling factors, are shown in Table 6:

[0110] Table 6 Main controlling factors

[0111] Variable Variable name Variable Variable name <![CDATA[x4]]> Proppant fracture width <![CDATA[x 12 > Gas saturation <![CDATA[x5]]> Proppant fracture height <![CDATA[x 13 > Total liquid volume <![CDATA[x6]]> Conductivity <![CDATA[x 17 > Treatment rate

[0112] In this example, based on the method of comprehensively screening the main controlling factors affecting the tight gas production by the maximum mutual information method and the random forest method, six main controlling factors, namely, propped fracture width, propped fracture height, conductivity, gas saturation, total liquid volume, and pumping rate, are obtained.

[0113] The above are only the preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Although the present invention has been disclosed above with the preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments by using the above-disclosed technical content within the scope of the technical solution of the present invention. However, as long as it does not depart from the content of the technical solution of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A method for comprehensively screening the main controlling factors affecting the production of tight gas based on the maximum mutual information method and the random forest method, comprising the following steps: Step1: According to geological research data and fracturing construction reports, collect and organize data on various characteristic factors affecting the production of tight gas, obtain the original data set, and clean the original data set to eliminate abnormal sample data; Step2: Calculate the maximum mutual information number between each pair of variables in the remaining sample data, including each independent variable and the dependent variable of tight gas production. The corresponding calculation method is: Where: MIC(x; y) represents the mutual information value; a represents the number of grids divided along the x direction; b represents the number of grids divided along the y direction; the empirical coefficient B ≈ n 0.6 , where n is the number of samples; the calculation formula of I(x; y) is as follows: In the formula: p(x) and p(y) are the marginal distributions of two random variables (X, Y), and is the joint distribution of two random variables (X, Y); Step 3: Normalize the maximum mutual information value, select the maximum mutual information value at different scales as the MIC value, and then convert it into a percentage score MIC i ; Step 4: Introduce the random forest regression algorithm to calculate each variable 10 times in a loop, including the RF scores between each independent variable and the dependent variable of tight gas production, and then convert them into a percentage-based score RF i , and take the average of the 10 RF i as the final RF score, namely RF I ; Step5: Obtain the MIC obtained in Step3 i value and the RF obtained in Step4 I The mean value of the values is used as the comprehensive score Total_Score i , and for Total_Score i Sort the values to screen out the optimal independent variables, and use them as the main controlling factors affecting the tight gas production.

2. The method for comprehensively screening the main controlling factors affecting the tight gas production based on the maximum mutual information method and the random forest method according to claim 1, wherein The obtained basic characteristic factor data includes perforation depth, perforation interval length, hydrocarbon source rock thickness, propped fracture width, propped fracture height, fracture conductivity, total hydrocarbon peak value, resistivity, rock density, acoustic time difference, porosity, gas saturation, total liquid volume, percentage of preflush fluid, proppant dosage, average sand ratio, pumping rate, and sand-carrying fluid volume.

3. The method for comprehensively screening the main controlling factors affecting the tight gas production based on the maximum mutual information method and the random forest method according to claim 1, characterized in that, The sub-steps for eliminating abnormal sample data from the basic data in Step1 include the following: Step101: Screen out the samples containing zeros in the basic sample data and eliminate them; Step102: Calculate the increasing gradient of gas production, draw a scatter plot of the increasing gradient of gas production, and eliminate the samples with too large changes in gas production; Step103: Eliminate the samples with abnormal gas saturation.

4. The method for comprehensively screening the main controlling factors affecting the tight gas production based on the maximum mutual information method and the random forest method according to claim 1, wherein Step3 specifically includes the following sub-steps: Step301: Divide the maximum mutual information obtained in Step2 by log2(min(a, b)) for normalization; Step302: Take the maximum value of all the obtained normalized maximum mutual information values as the MIC value; Step303: Given i and j, grid the scatter plot formed by each independent variable and dependent variable into a grid with i columns and j rows, and find the maximum mutual information value. Convert the obtained MIC value score into a percentage-based MIC. i , and the formula is as follows:

5. The method for comprehensively screening the main controlling factors affecting the tight gas production based on the maximum mutual information method and the random forest method according to claim 4, characterized in that Step4 specifically includes the following sub-steps: Step401: Use the random forest regression algorithm to loop and calculate the RF scores between each pair of variables 10 times, including each independent variable and the dependent variable of tight gas production; Step402: Convert the 10 - time RF score to a percentage - system score RF i , and the formula is as follows: Step403: Take the average of 10 RFs i as the final RF score RF I .

6. The method for comprehensively screening the main controlling factors affecting the tight gas production based on the maximum mutual information method and the random forest method according to claim 5, wherein Step5 specifically includes the following steps: Step501: Obtain the MIC values separately i and the RF I values, and then calculate the comprehensive score: Step502: According to the comprehensive score value Total_Score of each independent variable and dependent variable obtained i , sort them, compare to obtain the optimal independent variable, and use it as the main controlling factor affecting the tight gas production.

Images