A prediction method for TOC logging curves of source rocks based on adaptive time windows

Abstract

The present invention discloses a method for predicting TOC logging curves of hydrocarbon source rocks using an adaptive time window. The method comprises the following steps: selecting target logging curves and parameters; establishing a TOC multivariate linear regression formula and a neural network model; obtaining logging curves of different frequencies; calculating time windows for specific frequency lithologic curves; calculating TOC logging curves using a multivariate regression method; calculating time windows for specific frequency physical property curves; calculating TOC logging curves using a multivariate regression method; calculating TOC logging curves using a neural network model; and finally calculating time window divisions and TOC logging curve prediction results. The present invention utilizes a time-frequency analysis method to implement adaptive time window calculations at different scales, achieving accurate prediction of TOC logging curves of hydrocarbon source rocks, with the predicted results highly consistent with the measured TOC.

Description

Technical Field

[0001] The present invention relates to a method for predicting a well logging curve, in particular to a method for predicting a TOC well logging curve of a self-adaptive time window hydrocarbon source rock. Background Art

[0002] Total organic carbon (TOC) is a key parameter of organic matter abundance and can be used to evaluate source rock quality. TOC data can generally only be obtained through geochemical laboratory measurements. However, offshore core data is limited and sampling is discontinuous, making it difficult to accurately assess source rock quality. Therefore, obtaining a continuous and accurate TOC curve is crucial for geophysical evaluation of source rocks. Because there is a certain inherent relationship between well logging parameters and source rock TOC, well logging curves can be used to establish corresponding prediction models, thereby generating TOC curves.

[0003] Currently, methods such as multivariate linear regression, ΔlgR, and neural networks are used to predict source rock TOC log curves based on well logging data. These methods differ in their underlying principles and applicable conditions, suited to different geological conditions and data sources. When calculating source rock TOC log curves, these methods typically use uniform calculation parameters for the entire target interval. However, due to variations in sedimentary environments, this makes it difficult to guarantee TOC log prediction accuracy across different depth intervals. Summary of the Invention

[0004] The present invention aims to address at least one of the technical problems existing in the prior art. To this end, the present invention provides a method for predicting TOC log curves of hydrocarbon source rocks using an adaptive time window. This method utilizes time-frequency analysis techniques to calculate time windows of varying scales at a given frequency, and optimizes the calculation method for TOC log curve prediction, thereby ensuring TOC log curve prediction accuracy.

[0005] To achieve the above object, the present invention adopts the following technical solution: a method for predicting TOC logging curves of source rocks using an adaptive time window, comprising the following steps:

[0006] S100. Selecting target logging curves and parameters: Selecting lithologic and physical property logging curves that can reflect sequence stratigraphic changes in the target logging curves, and setting calculation parameters for the TOC logging curves;

[0007] S200. Establishing a TOC multiple linear regression formula and neural network model: Selecting well logging curves with a high linear correlation with measured TOC, and establishing a TOC multiple linear regression formula and neural network model;

[0008] S300. Obtaining logging curves of different frequencies: Obtaining logging curves of different dominant frequencies using time-frequency analysis technology;

[0009] S400. Calculation time window division of specific frequency lithology curve: Calculation time window division of specific frequency lithology curve using activity stratification method;

[0010] S500. Calculate TOC logging curve using multiple regression method: Calculate TOC logging curve using multiple regression formula based on calculation time window division of lithologic curve, record calculation time window that meets error conditions and output corresponding TOC logging curve;

[0011] S600. Calculation window division for specific frequency property curves: Calculation window division for specific frequency property curves using activity stratification method;

[0012] S700. Calculate TOC log curves using a multivariate regression method: Calculate TOC log curves using a multivariate regression formula based on the calculation time window of the physical property curves. Record the calculation time window that meets the error condition and output the corresponding TOC log curve. If the specific frequency does not reach the maximum frequency value, increase the main frequency and return to step S400. Otherwise, proceed to the next calculation step.

[0013] S800. Calculate TOC logging curves using a neural network model: For calculation time windows that do not meet the error conditions, calculate the TOC logging curve using a neural network model, compare it with the TOC logging curve calculated using the multivariate linear regression formula, and output the result;

[0014] S900. Calculation time window division and TOC logging curve prediction results: Merge the calculation time windows that meet the error conditions in step S400 and step S600 and the calculation time window in step S800 as the final calculation time window division scheme, and merge the TOC logging curves calculated and output in step S500, step S700 and step S800 as the final TOC logging curve prediction results.

[0015] The prediction method preferably selects the lithology curve that can reflect the change of sequence stratigraphy in the target logging curve as the gamma curve, the physical property curve as the density curve, and the calculation parameters set include the maximum value of the filter frequency f max , minimum value f min and frequency interval h.

[0016] The prediction method preferably selects well logging curves with a high linear correlation with the measured TOC, including gamma γ, acoustic wave time difference t, density ρ, resistivity R and neutron porosity Φ, and establishes the TOC multivariate linear regression formula as follows:

[0017]

[0018]

[0019]

[0020] Among them, formula (1) is the regression formula for semi-deep-deep lake phase; formula (2) is the regression formula for delta phase; formula (3) is the regression formula for other phases;

[0021] The TOC neural network model was obtained by BP neural network training using five logging curves including gamma γ, acoustic time difference t, density ρ, resistivity R and neutron porosity Φ.

[0022] The prediction method preferably uses time-frequency analysis technology to obtain logging curves of different main frequencies. Specifically, the lithology curve and physical property curve used for time window calculation are firstly transformed by fast Fourier transform to obtain the curves of the corresponding frequency domain, and then the frequency minimum value f is obtained by bandpass filtering. min To the maximum frequency f max , the frequency domain data of the logging curve with a frequency interval of h, and finally the filtered logging curves with different main frequencies are obtained by inverse Fourier transform; among which the initialization parameters are the initialization parameter calculation times n = 1, the average error threshold value δ = 25%, the initial calculation frequency f0 = f min .

[0023] The prediction method, preferably, uses the activity layering method to calculate the specific frequency lithology curve calculation time window division specifically as follows: calculate f n =f0+n*h, select frequency as f n The gamma curve of the well logging curve is used to obtain the calculation time window division scheme using the activity layer method. In order to express the dynamic properties of the layer curve, the activity of the well logging curve is defined as:

[0024]

[0025]

[0026] Where, E(n0) represents the activity value of the logging curve at data point n0; n0 represents the location where the activity value is calculated; t represents the time at which the activity value is calculated. The location of each well logging curve data point within the range; X(t) represents the value of the original well logging curve; d represents the time window length for calculating activity; Indicates that the logging curve is The average value of the logging curve within the range;

[0027] The corresponding discretization formula is:

[0028]

[0029]

[0030] Using equations (6) and (7), we can get f0=f minThe activity curve of the gamma curve has an extreme point which is the boundary point between different strata. The extreme point is calculated using the following formula:

[0031] d i =E(i+1)-E(i);i=1,2,…,n-1 (8)

[0032] Where i represents the i-th data point; E(i) represents the activity value of the logging curve at the i-th data point; d i represents the extreme value of the activity value at the i-th data point;

[0033] If d i》 >0, and d i+1 <0, then the i-th point is the peak point, that is, the time window interface, thus obtaining the time window division scheme.

[0034] The prediction method preferably uses a multivariate regression formula to calculate the TOC logging curve, records the calculation time window that meets the error condition, and outputs the corresponding TOC logging curve. Specifically, first, based on the obtained time window division scheme, use formulas (1) to (3) to perform TOC logging curve prediction calculation; then, calculate the average error between the TOC logging curve and the measured TOC in each time window; finally, select the calculation time window with the smallest average error value that is less than 25% of the threshold value, and output the corresponding TOC logging curve.

[0035] The prediction method preferably uses a neural network model to calculate the TOC logging curve within the calculation time window that does not meet the error condition, and outputs it after comparison with the TOC logging curve calculated by the multivariate linear regression formula: if the average error of the TOC logging curve calculated by the neural network model is less than the TOC logging curve calculated by the multivariate linear regression formula, the TOC logging curve is output as the calculation result; otherwise, the TOC logging curve calculated by the multivariate logging regression formula is output as the calculation result.

[0036] The present invention has the following advantages due to the adoption of the above technical solution:

[0037] The present invention uses the time-frequency analysis method to realize adaptive time window calculation of different scales, realizes accurate prediction of source rock TOC logging curve, and the prediction result has a high degree of consistency with the measured TOC. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic flow chart of the method for predicting TOC logging curves of source rocks using an adaptive time window provided by the present invention;

[0039] Figure 2 This is the result of calculating the Pearson correlation coefficient between measured TOC and logging data;

[0040] Figure 3 f n = Lithology curve calculation time window division scheme and TOC prediction result diagram at 10Hz;

[0041] Figure 4 f n = Lithology curve calculation time window division scheme and TOC prediction result diagram at 10Hz;

[0042] Figure 5 This is the final calculation time window division scheme and TOC prediction result diagram for L-1 well from 2900 meters to 3300 meters. DETAILED DESCRIPTION

[0043] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0044] In the description of the present invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the systems or components referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limitations on the present invention. Furthermore, the use of terms such as "first" and "second" to define components is solely for the purpose of distinguishing those components. Unless otherwise stated, these terms have no special meanings and should not be construed as indicating or implying relative importance.

[0045] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "installed," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0046] The adaptive time-window TOC logging curve prediction method for hydrocarbon source rocks provided by the present invention includes the following steps: selecting target logging curves and parameters; establishing a TOC multivariate linear regression formula and neural network model; obtaining logging curves at different frequencies; calculating time windows for specific frequency lithologic curves; calculating TOC logging curves using a multivariate regression method; calculating time windows for specific frequency physical property curves; calculating TOC logging curves using a multivariate regression method; calculating TOC logging curves using a neural network model; and finally calculating time window divisions and TOC logging curve prediction results. The present invention utilizes a time-frequency analysis method to achieve adaptive time window calculations at different scales, enabling accurate prediction of source rock TOC logging curves, with the predicted results highly consistent with measured TOC.

[0047] The following describes in detail the adaptive time window source rock TOC logging curve prediction method provided by the embodiment of the present invention with reference to the accompanying drawings.

[0048] In this example, a total of 5 wells with measured TOC data and complete logging curves were selected in the study area, which were respectively recorded as L-1, L-2, L-3, L-4 and L-5 wells, among which L-1 well was used as a verification well.

[0049] like Figure 1 As shown, the adaptive time window source rock TOC logging curve prediction method provided by the present invention mainly includes the following steps:

[0050] S100. Selection of target well logging curve and parameters: Select the lithology curve that can reflect the change of sequence stratigraphy in the L-1 well logging curve as the gamma curve, the physical property curve as the density curve, and set the maximum frequency f max , minimum value f min The maximum and minimum values ​​of the frequency and the frequency interval h can be determined according to the thickness and velocity of the target layer. In this case, the minimum frequency is set to 10 Hz, the maximum frequency is set to 40 Hz, and the frequency interval is set to 10 Hz.

[0051] S200. Establishing a TOC multiple regression formula and neural network model: Based on the well logging curves and measured TOC of wells L-2, L-3, L-4, and L-5, and according to the correlation calculation results (see Table 1 below), gamma γ, acoustic wave travel time t, density ρ, resistivity R, and neutron porosity Φ, which have a high linear correlation with the measured TOC, are selected to establish a TOC multiple linear regression formula:

[0052]

[0053]

[0054]

[0055] Among them, formula (1) is the regression formula for the semi-deep-deep lake phase; formula (2) is the regression formula for the delta phase; formula (3) is the regression formula for other phases.

[0056] Table 1 Calculation results of Pearson correlation coefficient between measured TOC and logging curve

[0057]

[0058] Well L-2 was used as a verification well for the calculation accuracy of the neural network model. Five well logging curves with high correlation, namely gamma, acoustic time difference, density, resistivity and neutron porosity, were selected from three wells, L-3, L-4 and L-5. BP (Back Propagation) neural network training was used to obtain the TOC neural network model. Figure 2 This is the TOC curve prediction result of Well L-2, with an average error of 4.7% compared with the measured data.

[0059] S300. Obtaining different frequency logging curves: First, the lithology curve and physical property curve used for time window calculation are transformed into the corresponding frequency domain curve by fast Fourier transform, and then the f is obtained by bandpass filtering. min =10Hz to f max =30Hz, the frequency interval is h=10Hz, and the frequency domain data of the logging curve is obtained. Finally, the logging curves with different main frequencies after filtering are obtained by inverse Fourier transform. Among them, the initialization parameters are the initialization parameter calculation times n=1, the average error threshold value δ=25%, the initial calculation frequency f0=f min .

[0060] S400. Calculation time window division of specific frequency lithologic curve: Calculation of f n =f0+n*h, select the filter frequency as f n The gamma curve of the well logging curve is obtained by using the activity layering method to obtain the calculation time window division scheme. In order to express the dynamic properties of the layer curve, the activity of the well logging curve is defined as:

[0061]

[0062]

[0063] Where, E(n0) represents the activity value of the logging curve at data point n0; n0 represents the location where the activity value is calculated; t represents the time at which the activity value is calculated. The location of each well logging curve data point within the range; X(t) represents the value of the original well logging curve; d represents the time window length for calculating activity; Indicates that the logging curve is The average value of the logging curve within the value range.

[0064] The corresponding discretization formula is:

[0065]

[0066]

[0067] Using equations (6) and (7), we can obtain the activity curve of the gamma curve with f0 = 10 Hz. The extreme point is the boundary point between different strata. The extreme point is calculated using the following formula:

[0068] d i =E(i+1)-E(i);i=1,2,…,n-1 (8)

[0069] Where i represents the i-th data point; E(i) represents the activity value of the logging curve at the i-th data point; d i represents the extreme value of the activity value at the i-th data point;

[0070] If d i》 >0, and d i+1 <0, then the i-th point is the peak point, that is, the time window interface, thus obtaining the time window division scheme.

[0071] S500. Calculate TOC logging curve using multiple regression method: First, based on the time window division scheme obtained in step S400, use multiple linear regression formulas (1)-(3) to predict and calculate TOC logging curves; then, calculate the average error between the TOC logging curve and the measured TOC in each time window; finally, select the calculation time window with the smallest average error value that is less than 25% of the threshold value, and output the corresponding TOC logging curve. Figure 3 f n =10Hz lithologic curve (gamma curve) calculation time window division scheme and TOC prediction results.

[0072] S600. Calculation window division for specific frequency physical property curve: For the calculation window with the average error greater than 25% in the previous step, the frequency is selected as f n The density curve of is obtained, and the calculation time window division scheme is obtained using the activity stratification method in step S400.

[0073] S700. Calculate TOC logging curve using multiple regression method: Calculate TOC logging curve using multiple linear regression formulas (1)-(3) respectively, record the time window with the minimum average error and less than the threshold value of 25%, and output the corresponding TOC logging curve. If f n If the maximum frequency value is not reached, the number of steps n is set to n+1 and the process returns to step S400; otherwise, the next step of calculation is performed; Figure 4 f n=10Hz physical property curve (density curve) calculation time window division scheme and TOC prediction results.

[0074] S800. Calculate TOC logging curves using the neural network method: For calculation time windows with an average error greater than 25%, use the neural network model to calculate the TOC logging curve; if the average error of the TOC logging curve calculated by the neural network model is less than that of the TOC logging curve calculated by the multivariate linear regression formula, output the TOC logging curve as the calculation result; otherwise, output the TOC logging curve calculated by the multivariate linear regression formula as the calculation result.

[0075] S900. Calculation time window division and TOC logging curve prediction results: Combine the calculation time windows that meet the error conditions in steps S400 and S600 and the calculation time window in step S800 as the final calculation time window division scheme, and combine the TOC logging curves calculated and output in steps S500, S700 and S800 as the final TOC logging curve prediction results. Figure 5 The final time window division scheme and TOC logging curve results for the depth section from 2900 m to 3280 m in Well L-1. It can be seen that there is a certain correspondence between the time window division scheme and the sedimentary facies, and the overall error between the TOC prediction curve and the measured TOC data is 24%.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for predicting TOC logging curves of source rocks using an adaptive time window, characterized in that: The following steps are involved: S100. Selecting target logging curves and parameters: Selecting lithologic and physical property logging curves that can reflect sequence stratigraphic changes in the target logging curves, and setting calculation parameters for the TOC logging curves; S200. Establishing a TOC multiple linear regression formula and neural network model: Selecting well logging curves with a high linear correlation with measured TOC, and establishing a TOC multiple linear regression formula and neural network model; S300. Obtaining logging curves of different frequencies: Obtaining logging curves of different dominant frequencies using time-frequency analysis technology; S400. Calculation time window division of specific frequency lithology curve: Calculation time window division of specific frequency lithology curve using activity stratification method; S500. Calculate TOC logging curve using multiple regression method: Calculate TOC logging curve using multiple regression formula based on calculation time window division of lithologic curve, record calculation time window that meets error conditions and output corresponding TOC logging curve; S600. Calculation window division for specific frequency property curves: Calculation window division for specific frequency property curves using activity stratification method; S700. Calculate TOC log curves using a multivariate regression method: Calculate TOC log curves using a multivariate regression formula based on the calculation time window of the physical property curves. Record the calculation time window that meets the error condition and output the corresponding TOC log curve. If the specific frequency does not reach the maximum frequency value, increase the main frequency and return to step S400. Otherwise, proceed to the next calculation step. S800. Calculate TOC logging curves using a neural network model: For calculation time windows that do not meet the error conditions, calculate the TOC logging curve using a neural network model, compare it with the TOC logging curve calculated using the multivariate linear regression formula, and output the result; S900. Calculation time window division and TOC logging curve prediction results: Merge the calculation time windows that meet the error conditions in step S400 and step S600 and the calculation time window in step S800 as the final calculation time window division scheme, and merge the TOC logging curves calculated and output in step S500, step S700 and step S800 as the final TOC logging curve prediction results.

2. The prediction method according to claim 1, characterized in that The lithologic logging curve that can reflect the sequence stratigraphic changes in the target logging curve is selected as the gamma curve, and the physical property logging curve is selected as the density curve. The calculation parameters set include the maximum value of the filter frequency f max , minimum value f min and frequency interval h.

3. The prediction method according to claim 2, characterized in that The well logging curves with high linear correlation with measured TOC, including gamma γ, acoustic time difference t, density ρ, resistivity R and neutron porosity Φ, were selected to establish the TOC multivariate linear regression formula: （1） （2） （3） Among them, formula (1) is the regression formula for semi-deep-deep lake phase; formula (2) is the regression formula for delta phase; formula (3) is the regression formula for other phases; The TOC neural network model was obtained by BP neural network training using five logging curves including gamma γ, acoustic time difference t, density ρ, resistivity R and neutron porosity Φ.

4. The prediction method according to claim 3, characterized in that The time-frequency analysis technology is used to obtain logging curves with different main frequencies. First, the lithology curve and physical property curve used for time window calculation are transformed into the corresponding frequency domain curve by fast Fourier transform, and then the minimum frequency f is obtained by bandpass filtering. min To the maximum frequency f max , the frequency domain data of the logging curve with a frequency interval of h, and finally the filtered logging curves with different main frequencies are obtained by inverse Fourier transform; among them, the initialization parameters are the initialization parameter calculation times n=1, the average error threshold value δ=25%, the initial calculation frequency f0=f min .

5. The prediction method according to claim 4, characterized in that The calculation time window of the specific frequency lithology curve calculated by the activity layering method is divided into the following parts: n =f0+n*h, select frequency as f n The gamma curve of the well logging curve is used to obtain the calculation time window division scheme using the activity layer method. In order to express the dynamic properties of the layer curve, the activity of the well logging curve is defined as: (4) (5) Where, E(n0) represents the activity value of the logging curve at data point n0; n0 represents the location of the calculated activity value; t represents the time in [ , ] represents the location of each logging curve data point within the range of the value range; X(t) represents the value of the original logging curve; d represents the time window length for calculating the activity; Indicates that the logging curve is in [ , ]The average value of the logging curve within the range of the value; The corresponding discretization formula is: （6） （7） Using equations (6) and (7), we can get f0=f min The activity curve of the gamma curve has an extreme point which is the boundary point between different strata. The extreme point is calculated using the following formula: （8） Where i represents the i-th data point; E(i) represents the activity value of the logging curve at the i-th data point; d i represents the extreme value of the activity value at the i-th data point; If d i >0, and d i+1 <0, then the i-th point is the peak point, that is, the time window interface, thus obtaining the time window division scheme.

6. The prediction method according to claim 5, characterized in that Use the multivariate regression formula to calculate the TOC logging curve, record the calculation time window that meets the error conditions, and output the corresponding TOC logging curve: First, based on the obtained time window division scheme, the TOC logging curve prediction calculation is performed using equations (1) to (3). Then, the average error between the TOC logging curve and the measured TOC in each time window is calculated. Finally, the calculation time window with the smallest average error value that is less than 25% of the threshold value is selected, and the corresponding TOC logging curve is output.

7. The prediction method according to claim 6, characterized in that For the TOC logging curve calculated by the neural network model within the calculation time window that does not meet the error conditions, the output after comparison with the TOC logging curve calculated by the multivariate linear regression formula is as follows: If the average error of the TOC logging curve calculated by the neural network model is less than that of the TOC logging curve calculated by the multiple linear regression formula, the TOC logging curve is output as the calculation result; otherwise, the TOC logging curve calculated by the multiple linear regression formula is output as the calculation result.

Images