Ancient river channel sand body type discrimination method
By removing suspicious data in logging parameters and performing dimensionality reduction processing, a river sand body type discrimination model is constructed and optimized, and a problem of large errors and high cost in river sand body type identification in the existing technology is solved, and a sand body type discrimination with higher accuracy and lower cost is achieved.
Patent Information
- Application Number
- CN202510459401.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-14
- Publication Date
- 2025-07-04
AI Technical Summary
In the prior art, the identification of river-type sand body types has problems such as large errors in prediction results, time-consuming and labor-intensive and inaccurate prediction results.
By removing suspicious data in the logging parameters, dimensionality reduction processing is performed, a prediction model with core-free well segment parameter M is constructed, and a secondary fit is performed to optimize the prediction model, and finally the river channel type is determined based on the width-depth ratio and river curvature.
It improves the accuracy and efficiency of river sand body type identification, reduces exploration and development costs, is repeatable and verifiable, and is suitable for prediction environments equipped with Excel computers.
Smart Images
Figure CN120255018A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of oil and gas field development, and in particular to a method for discriminating ancient river sand body types. Background Art
[0002] Fluvial facies reservoirs are one of the most important continental clastic rock reservoirs. Underground ancient river sand bodies are important oil and gas reservoirs, and the distribution characteristics and continuity of sand bodies vary significantly for different river types. Understanding the development types of sand bodies can better understand the heterogeneity and its changes within the reservoir, which is an important factor affecting the planar distribution of remaining oil.
[0003] Currently, for the identification of sand body types, mainly geological outcrops, core observations, and logging curve analysis methods are used, and then the arithmetic mean method is used to map manually for artificial determination. The randomness is too large, and different geological staff will have different understandings when mapping, resulting in non-standard and inaccurate geological research, causing errors in subsequent well placement. Summary of the Invention
[0004] Aiming at the problems of large prediction result errors and time-consuming and laborious existing in the current identification method for channel sand body types, the present invention provides a method for discriminating ancient river sand body types.
[0005] The method of the present invention eliminates suspicious data in the original data, then reduces the required parameters by dimensionality reduction of the original parameters, simplifies the prediction model, fits with the measured M value to establish a prediction model of M, and can further optimize the prediction model by secondary fitting to obtain the final prediction model.
[0006] The method for discriminating ancient river sand body types provided by the present invention specifically comprises the following steps:
[0007] S1. Obtain the logging parameters corresponding to the depth of the sample, including the original data of acoustic travel time (AC), compensated neutron (CNL), compensated density (DEN), and natural gamma (GR), and identify and eliminate suspicious data from the original data.
[0008] The method for eliminating suspicious data is as follows:
[0009] S11. Calculate the average value of all the original data of each logging parameter
[0010] S12. Calculate the deviation D between the original data of each logging parameter and the average value ; Further calculate D 2 ;
[0011] S13. Calculate the probability error p of each logging parameter, and the calculation formula is as follows:
[0012]
[0013] Wherein, n is the total number of original data of each logging parameter;
[0014] S14. Calculate the ratio of the deviation D to the probability error p of each logging parameter, that is, the D / p value, for subsequent identification of suspicious data. Among the n original data, if the deviation of a certain original data is D i , when all deviations equal to or greater than D i appear with a probability less than 1 / (2n), then this original data is suspicious data and should be discarded.
[0015] S2. Perform dimensionality reduction on the four logging parameters after removing suspicious data to reduce the required parameters, and obtain the dimensionality-reduced parameter F i . The specific method of dimensionality reduction is as follows:
[0016] S21. Standardize the original data to eliminate the influence of variables in terms of level and dimension. The formula used for standardization is:
[0017]
[0018] Where: x ij is the parameter in the i-th row and j-th column, is the average value of all parameters in the j-th column.
[0019] S22. Calculate the correlation coefficient matrix R m×n based on the standardized data matrix. The formula for calculating the correlation coefficient is:
[0020]
[0021] Where: m is the total number of rows / columns of the matrix, k represents the k-th row / column, x ik is the parameter in the i-th row and k-th column of the matrix, x jk is the parameter in the j-th column and k-th row of the matrix, is the average value of all parameters in the i-th row of the matrix, is the average value of all parameters in the j-th column of the matrix, i, j = 1, 2,..., m.
[0022] S23. Establish the characteristic matrix |R - λE| = 0, where R is the correlation coefficient matrix and E is the identity matrix, and calculate the eigenvalue λ i . Use the parameters whose sum of eigenvalues is greater than 85% of the total eigenvalues as the dimensionality-reduced parameters, and set the formula for the dimensionality-reduced parameter F i as:
[0023] F i = x i AC + x i+1 CNL + xi+2 DEN + x i+3 GR + x i
[0024] where: x i , x i+1 , x i+2 , x i+3 is the eigenvector corresponding to λ i , i = 1, 2, 3, ….
[0025] Establish the system of equations (R - λE)X = 0 for the obtained λ, and find the non - zero solutions of this system of equations. The obtained non - zero solutions are the eigenvectors x i corresponding to the eigenvalue λ i , x i+1 , x i+2 , x i+3 ; Substitute the eigenvector into the formula of parameter F i to obtain the parameter F after dimensionality reduction i .
[0026] S3. Construct a prediction model for the parameter M of the non - cored well section based on the parameter F after dimensionality reduction. The parameter M refers to the percentage content of components with a particle diameter less than 0.074 mm; The formula for the prediction model of the parameter M of the non - cored well section is as follows: i M1 = a1F1 + a2F2 + … + a
[0027] F i F i + c
[0028] where M1 is the predicted value of parameter M, dimensionless; F1, F2, F i are the parameters after dimensionality reduction, dimensionless; a1, a2, a i , c are fitting coefficients.
[0029] S4. Calculate the width - depth ratio F and the river sinuosity P according to M1 combined with the following two formulas:
[0030] F = 255×M1 -1.08
[0031] P = 3.5×F -0.27
[0032] where F is the width - depth ratio, that is, the ratio of the river width to the river depth, dimensionless; P is the river sinuosity, dimensionless.
[0033] If the width - depth ratio F ≥ 20 and the river sinuosity P ≤ 1.5, then the river is a braided river;
[0034] If the width - depth ratio F < 20 and the river sinuosity P > 1.5, then the river is one of meandering rivers or anastomosing rivers.
[0035] Preferably, the process of secondary fitting may also be included in step S3, and the specific method is as follows:
[0036] The predicted value M1 obtained is fitted with the measured value M of the particle size analysis data of the partial cored well section to obtain the secondary fitting coefficient, and the final particle size parameter M prediction model is obtained according to the secondary fitting coefficient. The model formula is as follows:
[0037] M2 = e(a1F1 + a2F2 + … + a i F i + c) + d
[0038] In the formula, a1, a2, a i , c, d, e are fitting coefficients; M2 is the final predicted value of parameter M; F1, F2, F i are the parameters after dimensionality reduction.
[0039] Then in step S4, according to M2, the width-depth ratio F and the river sinuosity P are calculated by combining the following two formulas:
[0040] F = 255 × M2 -1.08
[0041] P = 3.5 × F -0.27 .
[0042] Further preferably, the above method for discriminating ancient channel sand body types may further include step S5, and further identify the sand body according to the ratio of the thickness of the mudstone section to the thickness of the sandstone section:
[0043] ① To verify the accuracy of the braided river discrimination result in step S4, additional determination conditions are set: when the ratio of the thickness of the mudstone section to the thickness of the sandstone section within the same research period is less than 0.5, it is determined as a braided river channel; if this determination result is the same as the determination result in step S4, it is considered that the braided river discrimination result in step S4 is accurate; otherwise, return to step S4 to check whether there is an error in the calculation of the width-depth ratio and sinuosity data. If the data is correct, the result of the additional determination condition shall prevail;
[0044] ② To further distinguish the meandering river and the anastomosing river in step S4, additional determination conditions are set: when the ratio of the thickness of the mudstone section to the thickness of the sandstone section is between 0.5 and 0.7, it is discriminated as a meandering river sand body;
[0045] ③ When the ratio of the thickness of the mudstone section to the thickness of the sandstone section is between 0.7 and 0.9, it is discriminated as flood deposit.
[0046] The thickness of the sandstone section is obtained by the following method: according to the lithology interpretation result of the logging data, the formation is stratified into sandstone and mudstone according to lithology, and the demarcation point between the sandstone and the mudstone is distinguished by the half-amplitude point of the spontaneous potential logging curve to determine the thickness of the sandstone.
[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0048] (1) The method of the present invention preprocesses the original data, eliminates the suspicious data that may exist in the original data, and reduces the error. Through dimensionality reduction processing, the required parameters are reduced, and only a small number of parameters can contain the information of the previous multi-parameters and can make accurate predictions; the dimensionality-reduced prediction model is secondarily fitted to further reduce the error and obtain a higher prediction accuracy.
[0049] (2) The method of the present invention establishes a prediction model that can be applied to non-cored well M through the logging parameters of the cored well, reduces the difficulty of obtaining sand bodies, and reduces the exploration and development costs; it has repeatability and testability, and the prediction can be realized on a computer equipped with Excel. It solves the technical problems such as the too high cost of identifying channel-type sand bodies, limited quantity and distribution, and inaccurate prediction accuracy at present.
[0050] Other advantages, objectives and features of the present invention will be partially reflected by the following description, and partially will be understood by those skilled in the art through the research and practice of the present invention. Description of the Drawings
[0051] Figure 1 It is the dimensionality reduction gravel graph in the embodiment. Detailed Embodiment
[0052] The following describes the preferred embodiments of the present invention with reference to the drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0053] In this embodiment, the measured grain size analysis parameters M and the corresponding logging parameter data of the channel-type cored well are collected through investigation. After preprocessing the obtained logging parameters, they are substituted into the grain size analysis parameter M prediction model after dimensionality reduction, the fitting coefficients in the model are calculated and determined, the grain size analysis parameter M calculation prediction model is obtained, and then the predicted values are secondarily fitted to obtain the final prediction model of the grain size analysis parameter M and the final predicted values. Calculate the relative error between the measured value and the final predicted value, and prove the prediction accuracy of the method of the present invention according to the relative error.
[0054] Using the original data of a certain channel-type cored well sample (Table 1), the fitting coefficients are calculated through mathematical formulas, and thus the grain size analysis data M prediction model is obtained. The calculation process is as follows:
[0055] Table 1 Original data of relevant parameters of a certain channel-type sand body sample and the type of the cored location
[0056]
[0057]
[0058] First, identify the suspicious original data required: successively calculate the average value of each parameter The difference D between each observed value and the average value and its D 2 ; and calculate the probability error p value of each parameter; The calculation data results are shown in Table 2. And compare all D / p values with the standard values (Table 3). If the value is greater than the standard value, it is an outlier and should be discarded.
[0059] Table 2 Discriminant analysis data
[0060]
[0061]
[0062] Table 3 Standard table of D / p values corresponding to the number n of original data
[0063] n 5 10 15 20 50 100 D / p 2.5 2.9 3.2 3.3 3.8 4.2
[0064] Note: Table 3 is a standard table in a statistics book and is an empirical table based on probability control and extreme values. If n takes other values, such as the numbers 6, 7, 8, 9, etc., the D / p values corresponding to different n values can be determined by linear interpolation. For example, when n is equal to 6, 7, 8, 9, the corresponding D / p values are 2.6, 2.7, 2.7, 2.8 respectively.
[0065] The number of original data n obtained in this prediction is 20, so the D / p standard value of 3.3 is taken. After calculation and screening, it is obtained that only the AC data in samples 5 and 13 are abnormal.
[0066] Among them, for sample 5, AC = 54.491, D = 8.45245, D / p = 3.42692129 > 3.3.
[0067] For sample 13, AC = 54.432, D = 8.39345, D / p = 3.40300061 > 3.3.
[0068] Therefore, the AC in samples 5 and 13 are outliers. To reduce the error of the prediction equation, samples 5 and 13 are discarded.
[0069] Then, perform dimensionality reduction on the four logging parameters of AC, CNL, DEN, and GR required. The specific method is as follows:
[0070] In the first step, standardize the parameter data to eliminate the influence of the parameters in terms of level and dimension; the formula used for standardization is:
[0071]
[0072] In the formula: xij is the parameter in the \(i\)-th row and \(j\)-th column, is the average value of all parameters in the \(j\)-th column.
[0073] Calculate the correlation coefficient matrix \(R\) of the standardized data m×n , and the formula for calculating the correlation coefficient is:
[0074]
[0075] In the formula: \(x\) i are all parameters in the \(i\)-th row of the matrix, \(x\) j are all parameters in the \(j\)-th column of the matrix, \(x\) ik is the parameter in the \(i\)-th row and \(k\)-th column of the matrix, \(x\) jk is the parameter in the \(j\)-th column and \(k\)-th row of the matrix, \(m\) is the total number of rows / columns of the matrix, and \(k\) represents the \(k\)-th row / column, is the average value of all parameters in the \(i\)-th row of the matrix, is the average value of all parameters in the \(j\)-th column of the matrix.
[0076] The results of the correlation coefficient matrices of AC, CNL, DEN, and GR are shown in Table 4.
[0077] Table 4 Correlation Coefficient Matrices of AC, CNL, DEN, and GR
[0078]
[0079]
[0080] In the second step, establish the characteristic matrix \(|R - \lambda E| = 0\) (\(R\) is the correlation matrix, \(E\) is the identity matrix), find the eigenvalues \(\lambda\) of the matrix, and establish the characteristic matrix:
[0081]
[0082] The eigenvalues obtained are \(\lambda_1 = 1.987\); \(\lambda_2 = 1.405\); \(\lambda_3 = 0.432\); \(\lambda\) 4＝ 0.176. By observing Figure 1 the scree plot shown, it can be seen from the elbow plot that the broken line changes from steep to flat, and it is more appropriate to extract two principal components.
[0083] It can be known that selecting the first two parameters after dimensionality reduction can represent the original four parameters of AC, CNL, DEN, and GR.
[0084] In the third step, find the eigenvectors to characterize the parameters after dimensionality reduction. Substitute the eigenvalues into the characteristic matrix and establish a system of equations to find the eigenvectors of the eigenvalues:
[0085]
[0086] The eigenvectors are obtained as follows: x1 = -0.903; x2 = -0.008; x3 = 0.955; x4 = 0.131; x5 = 0.290; x6 = 0.899; x7 = 0.100; x8 = -0.864.
[0087] The parameters after dimensionality reduction are obtained as: F1 = -0.903GR - 0.008AC + 0.955CNL + 0.131DEN; F2 = 0.290GR + 0.899AC + 0.100CNL - 0.864DEN.
[0088] The parameters F after dimensionality reduction i are fitted with the parameters M of the particle size analysis data (the percentage content of the component with a particle diameter less than 0.074 mm) to construct a prediction model for the parameters M in the non-cored well section. The mathematical formula of the fitted curve used is:
[0089] Y = a1X1 + a2X2 + a
[0090] where
[0091]
[0092] in the formula: a1, a2, c are fitting coefficients; Y is the dependent variable parameter; X1, X2 are independent variable parameters; X 1i , X 2i are respectively the i-th data of the first and second independent variable parameters; are respectively the average values of the dependent variable and the first and second independent variable parameters.
[0093] The prediction model for the parameters M in the non-cored well section constructed is:
[0094] M1 = a1F1 + a2F2 + c
[0095] In the formula, M1 is the predicted value of the M parameter, dimensionless; F1, F2 are the parameters after dimensionality reduction, dimensionless; a1, a2, c are fitting coefficients.
[0096] According to the above fitting method, the prediction model for the parameters M in the non-cored well section obtained in this embodiment is specifically as follows:
[0097] M1 = -59.472 - 5.753×F1 - 1.204×F2.
[0098] The R-squared value of the model is 0.891, which means that F1 and F2 can explain 89.1% of the variation in M, and the predicted value M1 of M after dimensionality reduction is obtained. The comparison between the measured value of M and the predicted value M1 is shown in Table 5. Further, the width-depth ratio and sinuosity of the river are calculated based on the predicted value M1:
[0099] F = 255 × M1 -1.08
[0100] P = 3.5 × F -0.27
[0101] In the formula, F is the width-depth ratio, that is, the ratio of the river width to the river depth, dimensionless; M is the percentage content of the component with a particle diameter less than 0.074 mm, dimensionless; P is the sinuosity of the river, dimensionless.
[0102] If the width-depth ratio is greater than 20 and the sinuosity is less than 1.5, then the river is a braided river.
[0103] If the width-depth ratio is less than 20 and the sinuosity is greater than 1.5, then the river is one of meandering rivers and anastomosing rivers.
[0104] Except for the above two cases, other situations of the width-depth ratio and sinuosity are uniformly determined as anastomosing rivers.
[0105] Table 5 Measured and predicted values of M for core well channel sandstone samples
[0106]
[0107]
[0108] It can be seen from Table 5 that the predicted value is in good agreement with the measured value, the error is small, and the correct judgment rate is 94.44%, which can meet the general working requirements.
[0109] As can be seen from Table 5, there is a certain error between the measured value of M and the predicted value M1. In order to further reduce the error, the present invention can also perform a second-order fitting on the model formula M1 = -59.472 - 5.753*F1 - 1.204*F2 to obtain a more accurate parameter M prediction model formula. The calculation process of the second-order fitting is as follows:
[0110] Plotting the measured value M and the first predicted value M1 gives the correlation diagram of M - M1. It can be seen that the measured value M and the predicted value M1 show an obvious linear relationship, and the R-squared value of the model is 0.962, which means that Q1 can explain 96.2% of the variation in M (%). Therefore, the following formula is used for linear fitting of the measured value M and the predicted value M1 to further improve the prediction accuracy:
[0111] Y = d + eX
[0112] Among them,
[0113] Where: e and d are fitting coefficients, Y is the dependent variable parameter; X is the independent variable parameter; X i is the i-th independent variable parameter, is the average value of the independent variable parameters, Y i is the i-th dependent variable parameter, is the average value of the dependent variable parameters, and n is the total number of independent variable parameters.
[0114] Let the fitting curve be y = d1 + e1x, where x represents the first predicted value M1 and y represents the second predicted value M2.
[0115] Obtain the coefficients d1 and e1. Among them, first calculate and obtain Then from the formula we can get d1 = 0.226 and e1 = 1.031, that is, y = 1.031x + 0.226; Substitute e1 and d1 into it, and the final particle size parameter M prediction model is:
[0116] M2 = 1.031×(-59.472 - 5.753×F1 - 1.204×F2) + 0.226
[0117] Then calculate the river width-depth ratio and curvature according to the predicted value M2, and determine the river channel type. Table 6 conducts statistical analysis on the measured value of M, the two predicted values, and the correct judgment rate.
[0118] Table 6 Analysis of the measured value of M, the two predicted values, and the correct judgment rate
[0119] Measured value M <![CDATA[First predicted value M1]]> Is it judged correctly <![CDATA[Second predicted value M2]]> Is it judged correctly 69.33 68.36 Yes 70.70 Yes 62.25 51.96 Yes 53.79 Yes 95.32 93.54 Yes 96.67 Yes 81.11 79.29 Yes 81.97 Yes 38.32 36.78 Yes 38.15 Yes 24.89 23.66 Yes 24.62 Yes 11.32 10.61 Yes 11.16 Yes 7.42 6.93 Yes 7.37 Yes 3.86 3.51 Yes 3.84 Yes 76.23 73.65 Yes 76.16 Yes 45.44 43.99 Yes 45.58 Yes 30.36 30.02 Yes 31.18 Yes 87.23 83.87 Yes 86.70 Yes 72.32 68.72 Yes 71.08 Yes 93.65 90.14 Yes 93.16 Yes 82.78 80.08 Yes 82.79 Yes 94.24 90.96 Yes 94.00 Yes 99.03 96.88 Yes 99.11 Yes
[0120] It can be seen from Table 6 that the error between the predicted value obtained by the second fitting and the measured value is smaller and closer, and the correct judgment rate is higher. The present invention performs a secondary fitting on the dimension-reduced prediction model, further reducing the error and obtaining a higher prediction accuracy.
[0121] Verify the above-established M prediction model with another 20 known particle size analysis data M and samples of logging curves. The relevant data and results are shown in Tables 7 and 8.
[0122] Table 7 M and logging parameter values of sand body samples from a certain river channel type coring well
[0123]
[0124]
[0125] Table 8 Analysis of predicted values of M logging parameters and correct judgment rate
[0126] Sample Measured value of M Predicted value of M River channel type Judgment type Is it judged correctly 21 77.56 78.54 Meandering river Meandering river Yes 22 95.43 96.33 Meandering river Meandering river Yes 23 80.66 80.66 Anastomosing river Anastomosing river Yes 24 96.07 93.32 Meandering river Meandering river Yes 25 89.66 89.66 Anastomosing river Anastomosing river Yes 26 4.21 9.21 Meandering river Meandering river Yes 27 11.32 10.32 Braided river Braided river Yes 28 15.21 15.21 Braided river Braided river Yes 29 12.43 12.33 Braided river Braided river Yes 30 12.55 12.34 Meandering river Braided river No 31 30.57 30.32 Meandering river Meandering river Yes 32 13.88 13.32 Anastomosing river Anastomosing river Yes 33 30.34 30.21 Braided river Braided river Yes 34 13.55 13.21 Anastomosing river Anastomosing river Yes 35 39.37 39.32 Braided river Braided river Yes 36 16.57 16.33 Anastomosing river Anastomosing river Yes 37 24.09 24.36 Braided river Braided river Yes 38 24.42 25.21 Meandering river Meandering river Yes 39 9.34 9.55 Anastomosing river Anastomosing river Yes 40 5.33 5.85 Meandering river Meandering river Yes
[0127] The data results in Table 7 and Table 8 further prove the high accuracy of the determination results obtained by the method for discriminating river channel sand body types of the present invention.
[0128] 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 technical content disclosed above 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 according to 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 discriminating ancient river channel sand body types, characterized in that, Including the following steps: S1. Obtain the logging parameters corresponding to the depth of the sample, including the original data of acoustic travel time, compensated neutron, compensated density, and natural gamma, and identify and eliminate suspicious data from the original data; The method for eliminating suspicious data is as follows: S11. Calculate the average value of all the original data of each logging parameter S12. Calculate the deviation D between the original data of each logging parameter and the average value ; Further calculate D 2 ; S13. Calculate the probability error p of each logging parameter. The calculation formula is as follows: In the formula, n is the total number of original data of each logging parameter; S14. Calculate the ratio of the deviation D of each logging parameter to the probability error p, that is, the D / p value, for subsequent identification of suspicious data; among the n original data, if the deviation of a certain original data is D i , when all deviations equal to or greater than D i occur with a probability less than 1 / (2n), then this original data is suspicious data and should be discarded; S2. Perform dimensionality reduction on the four logging parameters after removing suspicious data to obtain the parameter F after dimensionality reduction i ; S3. According to the parameters F after dimensionality reduction i Construct a prediction model for the parameters M of the non-cored well section, where the parameter M refers to the percentage content of components with a particle diameter less than 0.074 mm; the formula for the prediction model of the parameters M of the non-cored well section is as follows: M1 = a1F1 + a2F2 + … + a i F i + c In the formula, M1 is the predicted value of parameter M, dimensionless; F1, F2, F i are the parameters after dimensionality reduction, dimensionless; a1, a2, a i , and c are fitting coefficients; S4. Calculate the width-depth ratio F and river sinuosity P according to M1 in combination with the following two formulas: F = 255 × M1 -1.08 P = 3.5×F -0.27 In the formula, F is the width-depth ratio, that is, the ratio of river width to river depth, dimensionless; P is the river sinuosity, dimensionless; If the width-depth ratio F≥20 and the river sinuosity P≤1.5, then the river is a braided river; If the width-depth ratio F<20 and the river sinuosity P>1.5, then the river is one of meandering rivers or anastomosing rivers.
2. The method for discriminating ancient river channel sand body types according to claim 1, characterized in that, The specific method of step S2 is as follows: S21. Standardize the original data; S22. Calculate the correlation coefficient matrix R based on the standardized data matrix m×n , and the formula for calculating the correlation coefficient is as follows: Where: m is the total number of rows / columns of the matrix, k represents the k-th row / column, x ik is the parameter at the i-th row and k-th column in the matrix, x jk is the parameter at the j-th column and k-th row in the matrix, is the average value of all parameters in the i-th row of the matrix, is the average value of all parameters in the j-th column of the matrix, i, j = 1, 2, …, m; S23. Establish the characteristic matrix |R - λE| = 0, where R is the correlation coefficient matrix and E is the identity matrix, and obtain the eigenvalues λ i , and use the parameter whose sum of eigenvalues is greater than 85% of the total eigenvalues as the parameter after dimensionality reduction, and set the parameter after dimensionality reduction as F i The formula for is: F i = x i AC + x i+1 CNL + x i+2 DEN + x i+3 GR + x i where: x i , x i+1 , x i+2 , x i+3 are the eigenvectors corresponding to λ i , i = 1, 2, 3...; Establish the system of equations \((R - \lambda E)X = 0\) for the obtained \(\lambda\), find the non - zero solutions of this system of equations, and the obtained non - zero solutions are the eigenvectors \(x\) corresponding to the eigenvalue \(\lambda\). i of i 、\(x\) i+1 、\(x\) i+2 、\(x\) i+3 ; Substitute the eigenvector into the formula of the parameter \(F\). i to obtain the parameter \(F\) after dimensionality reduction i .
3. The method for discriminating ancient river channel sand body types according to claim 1, wherein, Step S3 also includes a process of quadratic fitting. The specific method is as follows: Fit the obtained predicted value M1 with the measured value M of the particle size analysis data of some cored well sections to obtain the quadratic fitting coefficient, and obtain the final particle size parameter M prediction model according to the quadratic fitting coefficient. The model formula is as follows: M2 = e(a1F1 + a2F2 + … + a i F i + c) + d Where a1, a2, a i , c, d, e are fitting coefficients; M2 is the final predicted value of parameter M; F1, F2, F i are the parameters after dimensionality reduction.
4. The method for discriminating ancient river channel sand body types according to claim 3, wherein In step S4, calculate the width-depth ratio F and river sinuosity P according to M2 in combination with the following two formulas: F = 255 × M2 -1.08 P = 3.5×F -0.27 .
5. The method for discriminating ancient river channel sand body types according to claim 1, wherein, It also includes step S5, and further identify the sand body according to the ratio of the thickness of the mudstone section to the thickness of the sandstone section: ① To verify the accuracy of the braided river discrimination result in step S4, set an additional determination condition: when the ratio of the thickness of the mudstone section to the thickness of the sandstone section within the same research period is less than 0.5, it is determined as a braided river channel; if this determination result is the same as the determination result in step S4, it is considered that the braided river discrimination result in step S4 is accurate; otherwise, return to step S4 to check whether there is an error in the calculation of the width-depth ratio and sinuosity data. If the data is correct, take the result of the additional determination condition as the standard; ② To further distinguish the meandering river and anastomosing river in step S4, set an additional determination condition: when the ratio of the thickness of the mudstone section to the thickness of the sandstone section is between 0.5 and 0.7, it is discriminated as a meandering river sand body; ③ When the ratio of the thickness of the mudstone section to the thickness of the sandstone section is between 0.7 and 0.9, it is discriminated as flood deposition.
6. The method for discriminating ancient river channel sand body types according to claim 5, characterized in that, The thickness of the sandstone section is obtained by the following method: According to the lithology interpretation results of logging data, the formation is stratified into sandstone and mudstone according to lithology, and the demarcation point between sandstone and mudstone is distinguished through the half-amplitude point of the spontaneous potential logging curve, and the thickness of the sandstone is determined.